Welcome to Global Data Center Hub. Join investors, operators, and innovators reading to stay ahead of the latest trends in the data center sector in developed and emerging markets globally.

Join Global Data Center Hub

Edition No. 73

THE WEEK’S THESIS: BRING YOUR OWN POWER

This week, the data centers that moved forward brought their own power plant. The ones that leaned on someone else’s grid, or on a thin energy plan, stalled at the planning vote.

The evidence runs across three countries. AEP is building a power plant for Google in Wyoming. Fermi’s Texas campus runs on its own private grid. The Taylor campus near Austin will burn gas behind the meter. Solaria’s Uruguay park comes with its own solar, and Echelon’s Irish campus taps offshore wind through a substation it shares with the wind farm.

The projects that stalled lacked that answer. The UK turned down a growth zone bid built around a Drax campus and named its energy plan as a gap. In Waterford, gas turbines on site weren’t enough to win consent. Owning the power is now the entry ticket, and the buyer behind it decides who gets financed.

The structure behind AEP’s deal is in The Mechanism below.

MARKET SIGNALS

Fink puts a recession on the table. BlackRock CEO Larry Fink spoke to CNBC on Oct. 14. “If we have a moratorium of some sort in this country, we will be in a recession,” he said. He blamed most of the opposition on misinformation, three weeks before the Nov. 3 midterms. The data center fight has moved from zoning boards to national politics, and the money is now arguing its side in public.

Microsoft prices chips in tokens per megawatt. Microsoft unveiled Maia 300 on Oct. 12 and says it expects 40% more tokens per megawatt than leading GPU-based systems. Production deployments start in 2027. That is the company’s own projection. When power is the scarce input, output per megawatt decides which chip fills the hall.

Digital Realty pays the neighbors. Digital Realty is offering every household in De Soto, Kansas, up to $1,000 a year for 10 years, KSHB reported. The credits cover property taxes or utility bills from a $20 million community fund. The city council holds a hearing on the project’s tax district on Nov. 19. Local consent now has a price per household, and it sits in the developer’s budget.

Malaysia’s approvals pile up. Malaysia approved 27 data center projects worth RM225.97 billion between February 2025 and July 2026, Bernama reported. Johor holds 16 of them and Selangor 10. Each must meet a 30% local content floor. The government is also reviewing how much power the sector actually needs, and that review sets how many more get through.

Humain goes to tender in Riyadh. Humain has issued a construction tender for a 100MW AI data center in Riyadh, MEED reported, with an option to reach 200MW. MEED puts the contract at ~$1 billion. The design calls for direct-to-chip liquid cooling and racks of 240 to 250kW. Bids are due Nov. 8, and they will show what Gulf contractors charge for halls at that density.

CAPITAL WATCH

Solaria and Uruguay. President Yamandú Orsi signed an agreement with Spain’s Solaria on Oct. 14 for a 500MW data center park and 350MW of solar. The government puts the total at US$5.7 billion, with the first 200MW phase due by 2029. Bloomberg reported before the signing that the deal would take the form of a memorandum of understanding, and no tenant is named. So the figure describes the park’s planned size, and no buyer has put cash behind it yet.

Hillhouse and TeraNova. Hillhouse Investment launched TeraNova DC, an India data center platform, the Economic Times reported. It plans to invest ~$12 to $15 billion and build more than 2GW by 2032. No sites or committed capital have been named. The first land deal will show whether the number is a budget or a headline.

Princeton Digital Group and its lenders. PDG is seeking a syndicated loan of up to $1.2 billion for its 198.8MW Johor campus, Bloomberg reported. The three-year facility also refinances existing debt, with HSBC, Maybank, OCBC, Standard Chartered and UOB as lead arrangers. A three-year loan on a long-lived asset means PDG will be back in the loan market once the campus fills.

Nscale and its Norway site. Nscale’s flagship site at Glomfjord, Norway, runs 6.4MW of a planned 30MW and sits about six months behind schedule, Bloomberg reported. Bloomberg cited cooling leaks, local rules, construction delays and a fire. The company is reportedly preparing an IPO at a valuation of up to ~$35 billion. Public buyers will price that listing on megawatts delivered, and Glomfjord is the first number they’ll ask for.

GLOBAL LENS: THIS WEEK BY REGION

North America: the county signs, the state still decides

Mason County, Kentucky, approved development and bond agreements on Oct. 13 for a ~3,000-acre campus, and Meta was named as the company behind it. The first phase is reportedly ~$10 billion, though the county judge-executive has cited $15 billion, WVXU reported. The deal runs through an industrial revenue bond lease-back, so the county holds the deed and Meta makes fixed payments in place of property tax. State incentive approval is still pending, and that vote now decides when the first phase starts.

Big Watt Digital and PowerHouse Data Centers won approval from the Taylor City Council for a campus in Williamson County, Texas. It covers 664 acres and up to 1GW. It will run on natural gas behind the meter, and the developers asked for no tax incentives. First power is targeted for the first half of 2028, and no tenant is named. Gas on site gets the campus around the grid queue, but it still has to find a buyer for the space.

Europe: own the power, then sign the tenant

Echelon Data Centres broke ground on what it calls Ireland’s first green energy park, at its Arklow campus in Wicklow. The €4.2 billion plan covers two data centers and a 220kV substation shared with SSE’s 800MW Arklow Bank Wind Park 2. On-site generation runs on HVO, a renewable diesel. Completion is set for 2028, but the wind farm is still in development. The campus now runs on a wind farm timeline it doesn’t control.

Nebius signed a letter of intent to be the first customer at Volt’s Port of Rotterdam AI campus. Volt plans 600MW across six buildings, and its first phase of two 100MW buildings costs ~€7 billion including compute. Eneco will supply power from offshore wind. The LOI isn’t a lease, and Nebius hasn’t said how many megawatts it will take. A signed lease is what Volt needs to finance phase one.

Asia-Pacific: India builds, but only one has a buyer

Colt DCS and RMZ began construction on a third data center at their Navi Mumbai campus. The 64MW building is fully pre-let to an unnamed hyperscaler. Contracted power and a 220kV substation are already secured. A full pre-let before the building goes up moves the leasing risk off the developers. The next test is whether the substation energizes on the hyperscaler’s schedule.

CapitaLand unveiled a 53MW data center in Chennai, with a planned investment of ~INR 1,940 crore. It offers 34MW of IT capacity across 13 halls and is due by the end of 2026. Its listed India trust and a dedicated data center fund own it jointly. The site was first planned for a 2025 opening, so it’s running about a year late. No tenant has been named, and the first lease will set what Chennai capacity rents for.

Share

CONSTRAINT OF THE WEEK: THE PLANNING VOTE

The planning vote decided more projects this week than the power supply did. The UK government rejected the York and North Yorkshire AI Growth Zone bid, which centered on a Drax campus of at least 1.2GW. It said the bid fell short on credible energy delivery, environmental impact and local benefit. Drax now plans a 100MW first phase on the national grid, with a planning submission due in the coming months. That first phase is a twelfth of what the bid promised.

Waterford City and County Council refused permission for a data center and energy campus at Kilmeaden. The plan carried three data center buildings, 16 gas turbines, solar panels and a battery. The Irish Examiner reported that Echelon is also behind the project. On-site gas didn’t win consent, and any appeal now goes to An Coimisiún Pleanála.

The 150MW Norwich Apex Data Centre drew objections from local MP Ben Goldsborough and broadcaster Jake Humphrey within a week of its application. The Keswick site sits beside the protected Marston Marsh, and the plans include 72 rooftop chiller units. South Norfolk Council will decide. Once an MP objects in week one, the council vote turns political before the first officer report lands.

The next bottleneck is the ballot box. Amazon Web Services held an open house for an 18-building, 600-acre campus outside Trenton, Ohio. It also said it will stop asking local officials to sign NDAs. On Nov. 3, Trenton votes on Issue 17, a data center ban. A town vote can now stop a campus before any planning board sees it.

THE MECHANISM

A utility builds off its map for one buyer

American Electric Power is one of the largest US utilities, with 5.6 million customers in 11 states and about 33,000MW of generation. On Oct. 14 it committed ~$4 billion to a 702MW fuel cell plant near Cheyenne, Wyoming. Google will buy the output under a 20-year power purchase agreement. CEO Bill Fehrman said the plant serves needs “beyond our traditional service territory.”

The seller is Cheyenne Fuel Cell, LLC. AEP owns about 92%, and funds advised by Basalt Infrastructure Partners hold nearly 8%. Bloom Energy supplies fuel cells that make power from natural gas without burning it, and the plan adds carbon capture. Construction is planned to start this year, with full operation by the end of 2028.

That contract is what makes the plant financeable. No ratepayers stand behind this plant, so AEP’s return rests on one customer’s signature. For Google, a dedicated plant puts its power supply on a contract it controls.

The release gives no contract price. It doesn’t say who carries the gas cost for 20 years, or which permits the plant still needs. So, the risk sits with one buyer, one fuel and one fuel cell vendor, and the disclosures leave the fuel side blank.

Leave a comment

NOTABLE TRANSACTIONS

Fermi: AMD credit support behind a $6.5 billion TensorWave lease

Fermi said on Oct. 13 that AMD signed a credit support agreement for its Project Matador campus in Carson County, Texas. The support backs TensorWave’s lease of 222MW of power and 156MW of critical IT. The lease carries $6.5 billion of contracted revenue over a 15-year initial term. The campus runs on its own private grid behind the meter.

What makes it notable is that a chipmaker now stands behind its customer’s rent. TensorWave runs a GPU cloud built on AMD chips, so the chipmaker is backing demand for its own product. It is rated A by S&P and A1 by Moody’s.

AMD’s obligations begin only after construction is complete and rent starts on the final phase. The support ends if TensorWave or its guarantor holds an investment-grade rating for 12 straight months. The amount isn’t disclosed. Until rent starts, Fermi and its lenders carry the construction risk.

What to watch is the project financing. Fermi calls the agreement a step toward it and lists financing as a risk. The first construction loan will show what a backstop that starts after completion is worth to lenders.

NeoAI Cloud: KRW600 billion financing for a 40MW Pohang AI data center

NeoAI Cloud closed KRW600 billion (~$445 million) of financing for a 40MW AI data center in Pohang, South Korea. SC Capital Partners joined as a key equity investor. The project loan closed on Oct. 7, and service is targeted for the fourth quarter of 2027.

What makes it notable is how much support sits under one small building. A pre-lease, a contractor’s completion commitment and regional public money all back the first phase.

The money splits into a KRW450 billion project loan and KRW150 billion of equity. Woori Bank and KB Kookmin Bank arranged the loan. North Gyeongsang put in KRW4 billion and helped bring KRW131.5 billion from the Regional Revitalization Investment Fund. NHN Cloud has pre-leased 16MW, half the IT load, and Hyundai E&C committed to complete construction.

What to watch is the 260MW second phase. Phase one leased half its IT load before completion. Phase two will test whether lenders fund a build more than six times larger on the same terms.

EdgeConneX and Stockland: 50-50 Australian venture with 450MW of power secured

EdgeConneX entered Australia through a 50-50 joint venture with Stockland, announced Oct. 14. The venture has secured 450MW of power for hyperscale campuses in Sydney and Melbourne. No capital figure, tenant or timeline was disclosed.

What makes it notable is that a listed property company takes half of an operating platform. Stockland brings land, development and project management. EdgeConneX brings hyperscale delivery.

The power came before the tenant. Further power applications and planning work are in progress. Chris Maher was appointed managing director for Australia.

What to watch is the first planning application and the first signed tenant. Either one will show what a half share in land plus power is worth to Stockland’s shareholders.

THE FRAMEWORK: WHO BRINGS THE POWER

Every project that moved this week had a power plan it controlled. The table sets each one against its buyer and where it stands.

THE QUESTION FOR NEXT WEEK

Does Fermi close the TensorWave lease by Oct. 31? That is the deadline the two companies set in September. A close would put the AMD backstop to work and open the door to project debt. Another extension would tell lenders the guaranty still isn’t settled.

If you are new to Global Data Center Hub, visit our Start Here page for more information about the newsletter.

If you found this newsletter useful, share it with a colleague.

Have a great week.

- Obinna

Share Global Data Center Hub

Welcome to Global Data Center Hub. Join investors, operators, and innovators reading to stay ahead of the latest trends in the data center sector in developed and emerging markets globally.

This article is the 18th article in the series: Ground to Grid: A Free 21-Lesson Guide to Mastering Data Center Development

Join Global Data Center Hub

Why Handover Is Where Development Becomes Operation

A commissioned facility earns nothing until a tenant is live inside it. Handover is the moment development ends and the operating business begins. A clean handover is what makes the asset start performing.

Handover sits at the end of the development lifecycle. Before it, a project team builds and tests the facility. After it, an operations team runs that facility for the life of the lease. Handover is the bridge between the two.

The bridge carries real money. At the end of 2025, Digital Realty reported $817 million of annualized GAAP base rent (yearly rent under standard accounting rules) from leases signed but not yet commenced, at its share.

So, onboarding a tenant is an operational discipline. The lease milestone is only one part of it.

The tenant has to install equipment safely, receive a working service, and start paying on a date both sides can prove.

The quality of that work also shapes the tenant relationship for the life of the lease.

The Steps of Onboarding and Handover

Handover runs in four steps. They overlap, but each has its own owner.

1. Tenant Fit-Out and Deployment

The tenant installs and commissions its own equipment. That means racks, network gear and, in AI halls, liquid cooling connections.

The operator and the tenant have to coordinate closely to bring that load online safely. Every person, vendor and shipment is approved before arrival.

NIST SP 800-53 Rev. 5 sets out the access controls: authorize access, issue and remove credentials, log entry, escort visitors.

All work inside a live facility runs under an approved method of procedure (a written, step-by-step work plan), a change ticket and a back-out plan.

Network service is often the critical path. A cross-connect is complete when optical levels, continuity and speed are recorded and the customer or carrier accepts it.

AI tenants raise the bar. Uptime Institute notes that direct liquid cooling physically links IT and facility systems. Coolant distribution units need factory witness tests and load simulation. Practice is still maturing, so the test scripts belong in the contract.

2. Service-Level Activation

At this step the SLA (service level agreement: the promised performance and the remedy for missing it) moves from a contract to a live operating commitment.

An SLA works only when someone can measure it at an agreed point.

Verizon’s public U.S. colocation SLA defines power unavailability at the customer cabinet. It ties verified downtime to service credits.

So monitoring, reporting and response processes go into effect here. Each SLA line needs a named object, a handoff point, a metric, a data source, start and stop rules, exclusions and a remedy.

3. Documentation and Turnover

The operating team needs a full record of what it is about to run. That includes as-built drawings, commissioning records, operating procedures, equipment manuals, warranties, maintenance plans and training records.

ASHRAE/IES Standard 202-2024, the current published minimum commissioning standard for new buildings and systems, sets a consistent approach to documentation and training. ASHRAE’s guideline written for data centers, Guideline 1.6, is still only a proposal.

The site operations leader should formally accept these materials. When operations inherits them by default, nobody has checked that they are complete.

4. Operational Handover

The last step moves accountability from the development and construction team to the operations team. From this point, long-term management owns the facility, the tenant and the SLA.

The cleanest way to mark that point is a signed commencement record. It states the effective date, the services, the committed capacity, any accepted open items, the SLA start, the billing start and the opening meter readings.

A short stabilization period follows, with closer monitoring and a final reconciliation of alarms, access lists, circuits, meters and invoices. Then the service moves to normal review.

Why Handover Quality Matters

A poorly documented handover leaves the operations team blind to the facility it runs.

The team cannot maintain equipment it does not know about or follow a control sequence nobody gave it.

The cost of that blindness shows up in outages. In Uptime Institute’s 2026 outage analysis, 57% of respondents to its 2025 survey said their most recent major outage cost more than $100,000. One in five said their most recent impactful outage cost more than $1 million.

Uptime counts human error as a contributing factor in nine out of ten outages.

Uptime also cautions that outage data has limits. Those figures cover all outages, but procedures, training and configuration are set up at handover.

A rushed tenant onboarding spoils the first impression, and that impression sets the relationship.

The SLA becomes real at handover. The operator has to meet it from day one, including on the days the team is still learning the site.

Gaps between development and operations then surface as early operating failures.

What Separates a Clean Handover

The strongest handovers share four features.

First, complete documentation goes to an operations team that is ready for it.

The team was staffed and trained before the tenant arrived, and it signed for what it received.

Second, tenant onboarding is coordinated so load comes online without incident.

Work runs under approved procedures, and capacity is reserved in the company’s systems of record before the tenant’s equipment ships.

Third, SLA processes are tested and active before the tenant depends on them.

Strong operators raise a test alarm, confirm the customer is notified, calculate a sample credit and produce a sample monthly report.

Fourth, the transition is deliberate. Development and operations overlap for a planned period, with one accountable owner from contract signature to stabilization.

The contract, the operating records and the first invoice all show the same start date.

A Real-World Example: How Digital Realty Reports the Wait

Digital Realty shows what this interval looks like in public numbers. It defines backlog as annualized GAAP base rent from signed leases not yet commenced.

At year-end 2025, backlog stood at $817 million at Digital Realty’s share. Its investor materials put total backlog at about $1.4 billion at 100% share.

In the fourth quarter of 2025, Digital Realty booked $400 million of annualized GAAP rental revenue at 100% share, or $175 million at its own share. For those fourth-quarter leases, the average wait from signature to contractual start was eight months.

That lag reflects normal delivery schedules in a growing business. Each of those leases still has to pass through the four steps above before its rent begins.

Strategic Implications

For developers: Your reputation is set at handover. A clean transition is what earns the next deployment from the same tenant.

Plan the documentation package and the operations team’s training from the first design review.

For investors: Handover quality predicts the early operating performance your returns depend on.

Ask what share of signed rent has not commenced, how the company defines “commenced,” and who signs for the handover on the operations side.

For operators: Everything you inherit at handover defines your starting position. That includes the documentation, the SLA commitments and the tenant relationship.

Refuse to start the SLA clock until you have tested the reporting.

For policymakers: Reliable operation from day one is what makes the sector a stable part of the digital economy you are trying to build.

EU Delegated Regulation 2024/1364 applies to data centers with at least 500 kW of installed IT power and lets colocation operators gather required data from customers.

Give operators time to write such duties into tenant contracts.

Key Takeaway

Development ends and the business begin at handover. Everything before that day is cost. From that day on, the asset has to earn.

The tenant does not experience the years of development. It experiences the day it goes live: whether its racks went in safely and whether the first alarm reached the right person. A clean handover makes that day the start of a working relationship.

That day also sets the numbers. Rent starts on the commencement date, and the SLA clock starts with it. Every week of slip moves both further out. Digital Realty’s eight-month lag shows how much rides on that interval.

So, treat handover as the first day of operation. Staff the operations team before the tenant arrives.

Test the SLA before anyone depends on it. If you leave handover to the end, the first months of the lease become a repair job.

In your experience, what is the handover detail most often overlooked, and how does that oversight show up in the first months of operation?

Join Global Data Center Hub

Welcome to Global Data Center Hub. Join investors, operators, and innovators reading to stay ahead of the latest trends in the data center sector in developed and emerging markets globally.

Join Global Data Center Hub

In late 2022, Meta wrote off $1.3 billion of data center work to redesign for AI heat.

The newest AI racks are cooled with liquid, while most data centers run far cooler machines.

Every ChatGPT answer ends the same way. It becomes heat.

The chips that write the answer run on electricity, and almost all of that electricity leaves the building as warmth. For decades, data centers moved that warmth out with fans and cold air.

The newest AI chips run hot enough that the buildings around them need pipes.

Meta, the company behind Facebook and Instagram, learned this in the middle of a building spree.

Data Center Frontier, a trade publication, reported in April 2023 that Meta had dozens of data centers under construction. Many had been planned around ordinary computer chips.

In December 2022, Meta paused construction in Huntsville, Alabama; Temple, Texas; and Kuna, Idaho. It also dropped two of the three data centers it had planned near Odense, Denmark. On February 1, 2023, Meta told shareholders it had shifted to “a next-generation data center design” that could support AI.

Its results for the last three months of 2022 included $1.3 billion in charges for abandoned data center assets. In April 2023, Reuters reported that the new buildings would need liquid cooling to handle the heat of AI chips.

The redesign cost time and money. Temple and Kuna restarted about ten months after the pause, DatacenterDynamics reported on October 19, 2023.

The problem has kept growing since. A data center is planned around the chips of its day and built to run for decades, but AI chips change much faster.

The International Energy Agency reported in 2026 that AI server power density rose 11 times between 2020 and 2025.

A Shelf of AI Chips Gives Off 30 Boilers of Heat

The heat starts in the rack. A rack is a tall shelf, about the size of a refrigerator, that holds dozens of servers. A server is a desktop computer with the screen removed, flattened into a tray.

For most of the internet’s history, racks were cooled the way a person cools off in front of a fan.

Cold air blows across the front of the servers, picks up their heat, and leaves out the back. Machines outside the building then push that heat into the open air.

That works as long as the heat stays modest. Uptime Institute, a group that surveys data center operators every year, found in July 2025 that the typical rack in its survey drew a little under 9 kilowatts.

AI racks sit on another scale. NVIDIA, the company whose chips power the AI boom, designed its GB200 NVL72 rack to need 120 kilowatts of cooling, according to a design note it published on October 15, 2024.

That is more than 13 times the typical rack in Uptime’s survey.

The IEA put the same scale in plainer terms in its 2026 report: AI data centers must remove the heat of about 30 natural gas boilers from each rack.

At that level, NVIDIA cools the rack with liquid, much as a car radiator cools an engine.

Metal plates sit directly on the chips. Coolant flows through the plates, soaks up the heat, and carries it through pipes to equipment that cools the fluid and sends it back.

Liquid holds far more heat than the same amount of air, so a thin pipe can carry much more of it.

So, a building made for AI needs pipes running to every rack, pumps to push the coolant, and equipment to pull heat out of the fluid. A building made for air has fans, open floors, and none of that plumbing.

Most Operators Still Run Racks Under 30 Kilowatts

A newcomer could fairly assume that a data center is a data center: a secure building full of computers, ready for whatever chips come next.

The assumption has history behind it. For years, upgrading a data center mostly meant sliding newer servers onto the same shelves, cooled by the same air.

News coverage also measures data centers in megawatts, a number that shows how much electricity a building can draw and leaves out how it removes the heat.

Two buildings with the same megawatts can look identical in a headline.

Uptime’s 2025 survey shows how far most buildings sit from the AI frontier. More than 80% of respondents said their facility had no racks above 30 kilowatts.

Next to NVIDIA’s 120-kilowatt rack, most of the buildings in that survey were set up for a different kind of machine.

The clearer picture of the data center world has two parts. A large group of buildings runs ordinary computing on air. A newer group is being built around liquid for AI.

Meta’s 2022 pause shows what happens when a company finds its plans sitting in the first group while its future needs the second.

Old Buildings Stay in Service for Decades

The fix takes years because buildings last much longer than chips. In Uptime’s 2025 survey, 27.5% of respondents worked in data centers 16 years old or older.

Those buildings opened in 2009 or earlier, long before today’s AI chips existed, and they are still running.

Adding liquid cooling to a working building is slow work. Pipes have to run through halls full of live computers.

Pumps and new heat equipment need floor space and power of their own. Every change has to happen while customers’ computers keep running inside.

Even new campuses lose time adjusting. Meta’s redesign delayed Temple and Kuna by about ten months.

Microsoft, the company behind Windows and a major AI cloud, paused early work on the second phase of its Mount Pleasant, Wisconsin, campus in January 2025.

It told The Register on January 3, 2025, that it was evaluating “scope and recent changes in technology” and how they “might impact the design of our facilities.”

On the same Wisconsin campus, Microsoft said in September 2025 that Fairwater would cool over 90% of the facility using a closed-loop liquid system. On June 23, 2026, it announced that its first Mount Pleasant data center was fully operational.

Why the Heat Problem Reaches the AI on Your Phone

If you follow AI news, the heat problem explains why a new chip takes time to reach you. A faster chip helps only once a building can cool it.

ChatGPT launched on November 30, 2022, and a few weeks later Meta stopped building to redraw its data centers for AI.

The next AI tools will run in buildings that can carry away their heat.

If you live near a planned data center, cooling shapes how much water it uses. Some cooling systems shed heat by evaporating water, like sweat cooling skin.

Closed loops keep the same fluid moving around a sealed circuit.

Microsoft said in September 2025 that Fairwater’s loop is filled once during construction and then reused, and that the building would use roughly the water a typical restaurant uses in a year. That figure is Microsoft’s own.

Many operators publish far less: Uptime found in 2025 that only 47% of operators collected water data at all. At a public hearing for a new data center, it's fair to ask about the cooling design.

Four Clues That Show a Data Center’s Cooling

Start with the cooling words in an announcement. Terms like liquid-cooled, direct-to-chip, and closed-loop mean the building was designed for AI-grade heat.

Microsoft used that language for Fairwater in its September 2025 post, and company blogs and press releases are where these words show up first.

Next, watch for a pause to change the design. Meta used those words for Huntsville in December 2022, and Microsoft pointed to “recent changes in technology” in Wisconsin in January 2025.

Both statements tied a construction pause to the building's design, and in Meta’s case, the new design was built for AI. Local news and company statements carry these pauses first.

Then look for a charge for abandoned data center assets in a company’s quarterly results. Meta’s $1.3 billion line for late 2022 showed in one number how much unfinished building it was leaving behind.

These lines sit in earnings releases and in filings with the US Securities and Exchange Commission.

Finally, compare the rack number in a chip launch with what buildings run today.

NVIDIA’s 120-kilowatt rack sits far above the typical rack, which runs at a little under 9 kilowatts, in Uptime’s 2025 survey. Chipmakers publish their rack designs, and Uptime publishes its survey every year.

The AI people use every day feels weightless. Underneath it sits a shelf giving off the heat of 30 boilers, inside a building that has to carry that heat away every second.

The buildings going up in 2026 are being planned for the chips of 2026. Will they still be able to cool the chips of 2030?

Join Global Data Center Hub

Welcome to Global Data Center Hub. Join investors, operators, and innovators reading to stay ahead of the latest trends in the data center sector in developed and emerging markets globally.

Join Global Data Center Hub

Higher interest rates hit data centers through one gap: the time between when rent resets and when debt resets.

An owner whose rent moves first can absorb this cycle. An owner whose debt moves first carries it.

The rates are real. On September 16, the Federal Reserve raised its target range to 3.75% to 4.00%, its first hike in three years (Federal Reserve).

The 10-year Treasury closed at 5.29% on September 30, up from 4.19% on January 2 (FRED). The Fed’s median projection holds the policy rate at 4.1% through the end of 2027 (Federal Reserve).

The money is still moving. Bonds and securitizations backed by data center leases reached $17 billion in the first half of 2026, up 29%, JLL reported (JLL).

Vacancy in major U.S. markets sits at a record low of 1.4%, CBRE reported (CBRE).

Lenders are still lending. They are pricing the calendar.

The Buildout Is One of the Pressures the Fed Is Watching

Fed officials have tied part of the rate picture to AI spending. In July, Governor Lisa Cook named “increased capital expenditures tied to the buildout of AI infrastructure” as one of two unanticipated shocks of 2026 (Federal Reserve).

In September, Chair Kevin Warsh said hyperscalers “are out in the market raising funding, and so the competition for capital is real” (Reuters).

Keep the weight right. TD Economics estimates that about 50 of the roughly 80 basis points the 10-year has risen since late February came from expectations of more Fed hikes (TD Economics).

Reuters lists tariffs and higher energy prices beside AI (Reuters). AI spending is one of several forces the Fed is leaning against.

For an owner, the shape of the pressure matters most.

The Fed’s median shows no cut in 2027. Plan around rates near today’s level and find which side of your balance sheet feels them first.

Debt Reprices First

Construction loans feel a rate move almost at once. They float over SOFR, the overnight rate most construction lending prices from, which stood at 3.90% on September 28 (FRED).

JLL reports construction loans for investment-grade hyperscale tenants at spreads in the low 200s over that rate, for up to 85% of cost (JLL). That puts the all-in cost near 6%.

Bonds show the same pressure. In April, a Microsoft-leased QTS campus priced bonds at 5.70%. In August, another Microsoft-leased QTS deal priced at 7.228%, according to the CCIR deal ledger, which draws on press reports (CCIR).

The 10-year rose about 40 basis points between those dates. Most of the remaining gap came from wider spreads on AI-linked debt. The two bonds differ in structure and term, so read the comparison as directional.

Fixed-rate debt delays the hit. It does not remove it. Digital Realty carries a weighted-average coupon of 2.9%, and 90% of its debt is fixed, with maturities peaking in 2029 and 2030 (Investing.com).

Its guidance for new long-term debt moved from 4.0% to 4.5% in February to 4.5% to 5.5% in July (Digital Realty). The higher cost arrives one maturity at a time.

Leases Reprice on Their Own Schedule

Rent moves on two tracks: the yearly escalator and the renewal.

The escalator is the slow track. The three Northern Virginia hyperscale buildings Digital Realty agreed to buy on June 29 carry 15-year leases with rent rising 3.6% a year (Digital Realty 8-K).

The Fed projects 3.7% PCE inflation for 2026 and 2.3% for 2027 (Federal Reserve). This year, that escalator roughly keeps pace with prices.

If the 2027 forecast holds, it pulls ahead. A fixed escalator protects the owner as long as inflation stays near it.

The renewal is the fast track. Digital Realty’s second-quarter renewals came in 25.4% higher on a cash basis (Digital Realty). Leases above one megawatt renewed 66.7% higher. Leases under one megawatt renewed 5.2% higher (Investing.com).

CFO Matt Mercier put it plainly: “our expiring rates continue to drop over the next several years, while market rates continue to march up” (earnings call).

Short contracts reset most often. Equinix’s revenue contracts carry initial terms of one to five years, then renew automatically in one-year increments, its quarterly filing shows (Equinix 10-Q). Each renewal is a chance to price at today’s market.

New lease terms move cost too. CBRE reports take-or-pay minimums of 60% to 85% of allocated power. It also reports colocation operators beginning to adopt triple-net leases, where tenants pay taxes, insurance and upkeep directly (CBRE).

Both shift cost inflation from the owner to the tenant.

Four Positions on Two Clocks

Set the lease clock beside the debt clock and four positions appear.

The steadiest pairs short or frequently renewing contracts with long fixed-rate debt.

Rent resets to market often. Interest cost stays put. Higher rates arrive late, and renewals can outrun them.

Next is a long lease with escalators, funded with long fixed debt. Income and interest are both locked. This position barely notices a rate move until refinancing, when a new coupon meets an escalator set years earlier.

Tighter is a long lease with a fixed escalator, funded with floating construction debt.

Rent is set for 15 years. Interest moves with every Fed decision. The escalator has to cover the gap until the project refinances into fixed debt.

The most exposed is construction without a credit tenant. JLL puts those loans 200 to 300 basis points wider than investment-grade deals, at 70% to 80% of cost (JLL).

With SOFR near 3.90%, that implies 8% to 9% or more all-in. Until a lease is signed, only the debt clock runs.

Most real portfolios hold all four. What matters is the weight on each.

Value Shows the Gap First

Prices move before cash does. Digital Realty expects an initial stabilized cap rate above 6.5% on its June purchase (Digital Realty 8-K).

A cap rate is a building’s yearly income divided by its price. On signing day, the 10-year was 4.38%, a cushion of at least 212 basis points. On September 30, the cushion was at least 121.

If cap rates rise to rebuild that cushion while income holds flat, values fall. A move from 6.5% to 7.0% cuts value by about 7%. A move to 7.5% cuts it by about 13%.

Rent growth is the offset, and here the two clocks matter again. An owner who can re-sign leases 25% higher can grow income into a higher cap rate. An owner on a 3.6% escalator waits for the years to add up.

New building still pays. Digital Realty’s 1.4-gigawatt development pipeline is expected to deliver an 11.5% stabilized yield (earnings call).

That is about 620 basis points over the September 30 10-year. Higher rates thin the margin on existing buildings. On new ones, margins are still wide.

Outside the U.S., the Clocks Run on Other Rates

The same test applies abroad with different inputs.

Gulf central banks hiked with the Fed because their currencies are pegged to the dollar.

Saudi Arabia’s repo rate rose to 4.50% (Reuters). Dollar debt in the Gulf feels U.S. moves directly.

In India, dollar loans price off SOFR. AdaniConneX’s roughly $800 million construction loan priced at about SOFR plus 200 basis points, Bloomberg reported (Bloomberg).

Part of the revenue there comes in rupees. Two clocks become three: rent, debt and currency.

Nigeria cut its policy rate 350 basis points in September, to 23% (Central Banking).

For many operators there, energy is more than 35% of operating cost, Nairametrics reported (Nairametrics). Power sets the math more than the rate does.

What to Check Before the Next Deal

The rate picture will not settle by year-end. The Fed meets again on October 27 and 28. Investors are watching 6% on the 10-year as a pain point, Reuters reported (Reuters).

So put every rent reset and every debt reset on one calendar and find where debt resets first. Compare each escalator with the Fed’s inflation path for the next two years.

Then ask whether the lease moves power and operating cost to the tenant, through take-or-pay floors or triple-net terms.

One question stays open. Renewal spreads this wide depend on tight supply.

CBRE counts less than 1,500 megawatts of future capacity available across primary U.S. markets, about six months of leasing (CBRE).

If that supply loosens, the fast clock slows. Then the debt clock you set today is the protection you have left.

Join Global Data Center Hub

Welcome to Global Data Center Hub. Join investors, operators, and innovators reading to stay ahead of the latest trends in the data center sector in developed and emerging markets globally.

Join Global Data Center Hub

In Q3 2026, AI labs signed some of the largest data center contracts of the year. The company signing for capacity was often not the one owning the campus, raising the debt or backing the payments. If you build, lend to or invest in this market, that changes the questions you need to ask.

New Buyers, More Backers

AI labs became direct buyers of data center capacity in Q3 2026. Anthropic agreed to about $45 billion of capacity over six years with Nscale.

OpenAI announced a $20 billion campus of its own in Georgia and is the named customer for 8 GW at SB Energy’s PORTS-Pike campus in Ohio.

The money around them came through more hands than before. Banks lined up $15 billion for an Anthropic campus in Texas, with Google’s credit rating standing behind Anthropic.

Nvidia gave credit support at PORTS-Pike and set a $500 billion financing target with six of the largest investment firms. In El Paso, BlackRock became majority owner of a Meta campus while Meta kept a guarantee on part of its value.

I read this as good news for capital formation. More partners make money easier to raise. The open task is to widen the base of demand behind that money, because many of these structures still lean on a small group of labs and suppliers.

Most of the capital in this quarter’s list is also attached to capacity that is still being built, powered or contracted. For you, the opportunity is real, and so is the work of checking what each commitment actually covers.

How to Read This Quarter

Five Structural Trends

1. AI Labs Became Direct Buyers

Anthropic with Nscale in West Virginia. OpenAI in Effingham County and at PORTS-Pike. OpenAI again through Firmus in Asia-Pacific. The labs signed for capacity directly, and one report in the dataset put Anthropic’s compute deals at $517 billion over 11 months.

They were not the only buyers: Google, Meta, ByteDance and SK Telecom all committed capital too.

The labs also want capacity sooner than gigawatt campuses can deliver it, and both OpenAI and Anthropic were reported to be seeking 20 to 30 MW sites in the meantime. If you own ready capacity at that size, you are in a good position this year.

2. More Partners Now Stand Between the Tenant and the Money

The company using a data center is often not the one owning it, borrowing for it or backing its payments, and those roles are not interchangeable. Nvidia’s $500 billion platforms are a fundraising target.

At PORTS-Pike, Nvidia gave formal credit support on land, power and shell. On the Nexus loan, Google backstopped Anthropic with its credit rating. In El Paso, BlackRock became the majority owner while Meta kept a residual value guarantee.

Each of these moves a different piece of risk. If you lend into this market, “who is the tenant?” is no longer enough. Your question is which payments are supported, by whom, on what conditions, and which risks stay with the project.

3. Bitcoin Miners Became AI Landlords

Hut 8, Bitdeer and CleanSpark signed long-term AI leases. Riot reached a deal with Anthropic, IREN added $2.8 billion in contracts, and Core Scientific agreed to give AMD access to more than 500 MW.

The miners spent years assembling power and land, and in Q3 several turned those sites into AI leases running 16 to 20 years.

The value depends on conversion: cooling, network links, reliability, and the cost of meeting AI customers’ specifications. Where those line up, this is one of the quickest routes to powered capacity.

4. Local Policy Moved Into the Financing

States are setting clearer rules for large data centers, and the approaches differ. New York paused new state permits for hyperscale projects of 50 MW or more for one year.

Texas paused grid-connection approvals for new data centers while regulators review a queue of about 474 GW of proposed demand, roughly 90% of it from data centers.

Pennsylvania made its development standards mandatory and tied state permits to local approval, without a moratorium. These rules give developers a clearer path, and they also feed into financing.

Some lenders are already factoring permitting timelines into how they price construction debt. If you are planning a campus, your permitting plan is now part of your financing case.

5. Resilience Became Part of Site Planning

The Gulf remains one of the fastest-growing data center regions, and this year it also showed why resilience planning matters. Regional security events affected some facilities, and AWS advised some customers in Abu Dhabi and Bahrain to consider alternatives while repairs continue.

Investment kept coming. Microsoft framed its $10 billion Gulf strategy around resilience, Microsoft confirmed its Saudi Arabia East region for November 2026, AWS and Humain committed to 50 MW in a Saudi AI zone by 2028, and Humain is seeking $2.5 billion for new capacity.

If you are building or financing in the region, redundancy and physical security now sit next to power and land in your plan.

Where the 12 Announcements Stand

Capacity moves through stages: announced, site controlled, customer committed, power contracted, financed and under construction, energized, and finally accepted and earning revenue. The dollar figures in this list sit at very different points on that path, and knowing which point helps you compare them.

Where the Openings Are

1. Lending Where a Strong Backer Supports Specific Payments

The Google-backstopped Nexus loan moved toward bank sell-down within weeks, and an Anthropic-linked data center raised $1.3 billion in private credit. The opening is for lenders who read the support documents, not only the press release: what is backed, by whom, for how long, and what stays with the project.

2. Mining Sites That Can Be Converted at the Right Cost

Hut 8, Bitdeer, CleanSpark, Riot, IREN and Core Scientific all signed AI deals in the quarter. Your opportunity is in the sites that can meet AI customers’ needs at a conversion cost and on a timetable the lease supports.

3. Public Listings for Builders

Nscale filed for a US IPO. Switch lined up a $10 billion listing at a potential $80 billion valuation, SB Energy filed publicly, Firmus courted investors for an ASX listing, and Blue Owl planned a listed data center REIT. These are filings and plans, not completed raises. As they price, early backers gain a route to liquidity and builders a new source of equity.

4. Emerging Markets With New Rules and Funded Projects

Brazil signed its Redata tax regime into law. Indonesia drew a $3.1 billion loan for Zankore’s 1 GW project, plus CoreWeave’s first APAC entry. Johor kept raising multi-billion-dollar financing, and Egypt signed a $1 billion data center and sovereign AI deal. In each case, policy arrived with money attached.

5. Mid-Size Capacity That Can Deliver Soon

OpenAI and Anthropic were reported to be seeking 20 to 30 MW facilities while larger campuses are built. If you own ready capacity at that size, you may have pricing power, provided your site meets the labs’ needs on power density, cooling, connectivity and location.

What to Plan For

1. Widening the Base of Demand

Different developers, lenders and financing structures can still depend on the same few labs and suppliers. More partners make capital easier to raise. Spreading the demand behind it is the next step, and it is worth tracking in your own portfolio.

2. Power Costs Became a Policy Question

California signed seven bills requiring data centers to disclose power and water use and to keep the cost of new power off other customers’ bills. The US House passed the Ratepayer Protection Act 417-3. It asks states to consider a standard that makes large users pay the full cost of new generation and transmission, and it now goes to the Senate.

Virginia ordered a review of backup generators and named Eastern Virginia a cooling-water-scarcity area. Thailand asked operators to pause 49 buildouts while it completes a legal framework.

Clear cost rules help you plan. They also mean your project should budget for its own power and grid costs from the start.

3. Credit Terms Tied to Ratings and Permits

Some utilities now ask large data center customers below a set credit rating to post collateral sized to the power plants and lines built for them, which can run into the billions for the largest projects.

The El Paso debt priced at about 7.5%, and some lenders are factoring permitting timelines into construction loans. If you are raising money, your credit rating and your permits now shape your terms as much as your tenant does.

4. Resilience in Fast-Growing Regions

Recovery timelines for some affected Gulf facilities are still being set, and the UAE has revised its AI data center plans. Redundancy and site choice in the region now carry a planning cost that earlier budgets did not include.

5. Gigawatt Campuses Need Time

Large campuses in several markets have taken longer than planned, and grid connection timing has shaped schedules in parts of the UK, where Ofgem has proposed reforms to prioritize ready projects.

The labs are signing faster than big campuses can be energized, which is why smaller, ready sites are in demand.

What Would Change My View Next Quarter

• Delivery dates. Whether Nscale’s West Virginia campus, Meta’s El Paso campus and OpenAI’s Effingham site hold their schedules.

• The terms behind the backing. Whether the Nexus loan closes, and what Google’s support covers once the terms are public.

• The state rules. What the Texas grid review concludes, whether the Ratepayer Protection Act passes the Senate, and how New York’s one-year permit pause ends.

• Demand breadth. Whether new lab and enterprise customers join the contracts, widening the base behind the money.

• Gulf recovery. Whether recovery dates are set and how new capacity in the region is designed for resilience.

The Full Ranked List: Top 12 Global Announcements (Q3 2026)

The 12 announcements below are ranked by capital scale, strategic signal and structural impact. The dollar figures measure different things, from contract value to financing targets, so each entry says which one it is.

1. Anthropic Agrees $45 Billion AI Data Center Deal With Nscale

Anthropic agreed to rent capacity from Nscale, a UK start-up, in a deal worth about $45 billion over six years. That is contract value, not money invested. It covers 460 MW at Nscale’s flagship campus in Mason County, West Virginia, with Anthropic’s access starting when the chips come online in late 2027. About a month later, Nscale filed for a US IPO, which gives public investors a view of the contract and the delivery plan behind it. Nscale’s next chapter centers on delivering this campus on time. [Read here]

2. Nvidia Builds $500 Billion AI Compute Financing Platforms With Six Major Investors

Nvidia partnered with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR on financing platforms that aim to mobilize more than $500 billions of third-party capital. That figure is a target, not committed money. The firmer step came the same month at SB Energy’s PORTS-Pike campus in Ohio.

There, Nvidia agreed to give credit support on the land, power and shell buildout for an initial 4.25 GW of IT capacity, with an option on the rest, and to be the site’s AI compute supplier. OpenAI is the named customer for 8 GW. The chip supplier is now helping fund the buildings its chips go into. [Read here]

3. Banks Line Up $15 Billion for Google-Backed Anthropic Data Center

In late July, Morgan Stanley was leading advanced talks to lend $15 billion to Nexus Data Centers for an Anthropic campus in Hubbard, Texas. Google agreed to backstop Anthropic with its investment-grade credit rating. By early August, the banks were reported to be preparing to sell the debt on to other investors. If you are a lender, the key is the exact scope of Google’s support, meaning which obligations it covers and on what conditions. [Read here]

4. Brookfield and NextEra Plan $100 Billion Campus in Paducah, Kentucky

Brookfield Asset Management and NextEra Energy are reported to be developing a $100 billion data center campus in Paducah, Kentucky. That is a project estimate from an unnamed source, and neither company has confirmed it in the dataset, so I treat it as pipeline. The firmer Brookfield move came earlier: it expanded its Bloom Energy partnership fivefold, to $25 billion, to build and finance fast on-site power. [Read here]

5. OpenAI Announces $20 Billion Campus in Effingham County, Georgia

OpenAI announced a $20 billion data center development in Effingham County, Georgia. That is planned investment. About a month later, a 3.2 GW power deal for the site was approved. Acting as its own developer gives OpenAI more control over timing and design, along with responsibility for site, power and community relations. [Read here]

6. Meta and BlackRock Form El Paso Data Center Venture

Meta and BlackRock formed a venture to own a 1 GW campus under construction in El Paso, with about $14 billion in total development costs. Funds managed by BlackRock own 80% and put in about $4.9 billion of cash, partly funded by a $12.5 billion debt raise priced at yields of about 7.5%. Meta keeps 20%, leases the whole campus for an initial four years with options out to 20, and manages construction.

Meta also gives residual value guarantees with a threshold of about $13 billion: under certain conditions in the first 16 years, it covers any shortfall between the campus’s value and that level. It is a useful model for you to study: ownership moves to institutional capital, while the user keeps a share of the value risk. [Read here]

7. Hut 8 Fully Leases 1 GW Beacon Point Campus at $19.6 Billion

Hut 8 signed a second 352 MW IT lease, which fully leased its 1 GW Beacon Point campus and lifted campus-level base-term contract value to $19.6 billion. That is lease value over the base term. The source headline does not name the tenant. Bitdeer signed a $4.7 billion, 16-year lease in Norway, and CleanSpark a 20-year lease in Georgia. Mining sites bring power and land, and the payoff comes from converting them to what AI customers need, at a cost and on a timetable the lease supports. [Read here]

8. ByteDance Begins $38.44 Billion Data Center in Brazil

ByteDance began building a $38.44 billion data center in Brazil, the largest single Latin American commitment in the dataset. That is planned investment. Two more signals followed. Brazil signed its Redata tax regime into law, and Alibaba Cloud opened its first South American region there. Brazil is drawing global platforms at a scale that gives it a real say in how the next wave of capacity is built. [Read here]

9. SK Telecom Sells 49% of SK Horizon to KKR and IMM for $2.2 Billion

SK Telecom launched SK Horizon, its AI data center company, and agreed to sell 49% to KKR and a Korean private equity consortium including IMM for 2. 2billion.That is the price of the stake. Earlier in the quarter, SKToutlinedUS91.5 billion of new Korean data center investment and a plan for 15 GW of AI capacity. Bringing in a large minority partner before construction puts outside equity in at the start of the program. [Read here]

10. Google Commits €13 Billion to AI Infrastructure in Finland

Google committed €13 billion over two years to AI infrastructure in Finland. It had company. Cerebras and Compute Nordic announced a 165 MW site in Mikkeli, TikTok expanded its Finnish capacity, and Microsoft started pre-construction in Vaasa. Finland’s power and climate are drawing a cluster of global builders at once.[Read here]

11. Mitsubishi Estate Enters Data Centers With 2.5 GW and $9.3 Billion

Mitsubishi Estate is entering the data center market with 2,500 MW of planned capacity and a $9.3 billion planned investment. That is a first move at gigawatt scale. Goodman also signed the first hyperscale lease at its 1 GW Tokyo campus, and Keppel DC REIT agreed to buy stakes in two Tokyo data centers for $1.05 billion. Japan’s big property owners are moving into data centers as a business line. [Read here]

12. Firmus Passes 900 MW Contracted and Adds OpenAI as Anchor Customer

Firmus passed 900 MW of contracted capacity, signed OpenAI as an anchor customer and expanded into Malaysia. A month earlier it had closed a fully subscribed $2 billion strategic equity raise. By mid-September it was courting investors for an ASX listing. Contracts, equity and a listing process, all inside one quarter. [Read here]

Join Global Data Center Hub

Welcome to Global Data Center Hub. Join investors, operators, and innovators reading to stay ahead of the latest trends in the data center sector in developed and emerging markets globally.

Join Global Data Center Hub

Capital is moving toward AI chips. In one week, three large chip financings surfaced. Apollo is at the table on all three, alongside names like Goldman Sachs, Blackstone and PIMCO.

Chips are moving from company spending into something lenders will fund like infrastructure. And each deal solves the same problem with a different tool.

Oracle would rent its chips from an investor-owned company. Broadcom would stand behind its customer’s payments. SpaceX would borrow on its own name. All three still need the same thing at the end: a building with power.

What Was Reported, and What Was Not

All three deals are in talks. None is signed.

The Wall Street Journal reported that Oracle is in talks with Apollo and Goldman Sachs to fund a large chip purchase. Per that report, the likely structure is for investors to fund a separate company that buys the processors and leases them to Oracle over time. No amount was reported. Oracle is still talking to other partners.

The same report said Broadcom is arranging more than $50 billion so OpenAI can buy the custom AI chips the two companies are building together. Apollo and Blackstone are among the lenders approached. The talks are early. The size could change. The borrower has not been named.

The Financial Times reported that SpaceX is seeking about $40 billion to buy Nvidia processors. The package would combine about $10 billion of bank loans with $30 billion of investment-grade debt. Apollo is expected to lead and sell the debt on to other investors, and PIMCO is among the firms in talks. The FT’s sources said the deal should close in 2027. Bloomberg described the talks as early and said they could still end without a deal.

None of the companies has confirmed these terms. The filings below are the confirmed part.

The Structures Behind These Deals Already Work

These talks build on deals that have already closed.

In June 2026, Apollo led an initial $35 billion financing for Broadcom’s AI XPV Platform, alongside Blackstone and a group of banks. It supports Anthropic’s expansion of more than 1 GW of compute. The wider platform targets over 20 GW for frontier AI labs through 2028.

In the same month, SpaceX priced $25 billion in senior unsecured notes in five tranches, with maturities running from 2031 to 2056.

And NVIDIA has partnered with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR on platforms meant to mobilize over $500 billion of outside capital over time. That figure is a stated target, not money in place. It shows where the chip business is heading. Financing is becoming part of what gets sold with the chip.

Oracle’s Structure Matches Chip Costs to Chip Revenue

The reported logic is timing. Oracle pays for hardware before it collects the cloud revenue that hardware produces. A lease spreads that cost into payments that run alongside the revenue.

Oracle’s quarterly filing for the period ending August 31, 2026 shows why that matters. It lists $288 billion of lease commitments that have not started yet, substantially all for data centers. They are expected to begin between fiscal 2027 and fiscal 2029 and run for 15 to 19 years. The same filing puts the estimated useful life of its servers and networking equipment at six years.

So the building and the chips run on different timelines. One building lease can span two or three generations of servers. A separate chip vehicle lets each hardware refresh be paid for on its own schedule, while the building lease runs on its own.

The terms that will shape this structure are not public yet. They include whether the lease can flex with demand, whether the processors are backed by a minimum resale value, and whether the vehicle can move hardware to another site. Those terms determine how Oracle and its investors share the work.

Broadcom’s Support Is What Lets Lenders Fund Custom Chips

Custom chips are harder to fund than standard ones. A lender has fewer obvious buyers if the hardware ever has to be sold. Broadcom’s answer is to back its customers.

Broadcom’s quarterly filing for the quarter ending August 2, 2026, explains how its existing platform works. A financial partner took over agreements to buy AI racks built on Broadcom-designed chips, along with the customer’s leases. Broadcom backs the customer’s lease payments over five-year terms.

If the customer defaults, Broadcom owes the gap between 85% of the amount still owed and what the racks bring in when sold. Broadcom can also take over the lease or arrange a sale. Its maximum exposure, once every rack is deployed, is about $29 billion. As of the filing, it had paid nothing, and the backstop’s fair value was not material.

Here is how that 85% line shares the load. The financial partner takes losses up to the first 15% of what is owed. Broadcom covers the shortfall below that, after whatever the hardware fetches. The chip designer becomes part of the financing, which gives lenders a strong party behind a new kind of asset.

The filing also describes a separate facility. Under certain conditions, the customer may issue Broadcom up to $42 billion of convertible notes to fund its lease payments. None had been issued as of August 2. The two figures should not be added together, because the filing does not show that both would ever be drawn at once.

OpenAI and Broadcom announced in October 2025 a plan to deploy 10 GW of custom accelerators starting in the second half of 2026. The new $50 billion package would help fund that plan. It has not been reported whether it uses the same 85% support.

SpaceX Brings Its Whole Business to the Table

SpaceX’s reported package looks like a large company borrowing against its whole business, and SpaceX already has a record in the bond market. Its June notes carry coupons from 5.35% to 6.65%. Its second-quarter results reported $100 billion in cash, cash equivalents, and marketable securities as of June 30, along with 1.4 GW of compute capacity.

If the new debt is unsecured, lenders are relying on SpaceX’s launch, connectivity and AI businesses together. Repayment comes from the whole company, so the debt term and the chips' lifespan do not need to match. No one has reported whether any part of the $30 billion would be secured by the chips.

The Bankability Gap Now Sits at the Building

The Bankability Gap is the distance between announced capacity and capacity that can secure power, financing, permits, equipment, customers, and returns. These deals close much of that gap on the financing side. What remains sits at the building.

A financed chip earns its return once it runs. It needs firm utility power, finished substations, liquid cooling sized to the rack, fiber, and a team to commission it. The demand is there. The International Energy Agency projects data center electricity use rising from about 415 TWh in 2024 to about 945 TWh in 2030, with accelerated servers driving nearly half the increase.

So chip financing runs on four clocks. When the money is drawn. When the chips arrive. When the utility energizes the site. When the customer starts paying. When the four line up, the chip earns from the day the payments start.

That is good news for operators. A tenant with funded chips is a strong tenant. If you host one, tie the financing draw to energization in writing, and agree up front who handles late delivery, late power, cooling changes, or lower rack density than planned.

Regulators Are Building a Clearer Picture

The Federal Reserve Bank of Kansas City published research on October 5. It found that special purpose vehicles are the most highly leveraged entities in the build-out, with broad sponsor guarantees letting some finance 90 percent of their assets with debt. Sponsors support that debt through rent payments and residual value guarantees.

In March, the Financial Stability Oversight Council’s systemic risk committee recommended monitoring hyperscaler debt issuance and “debt-like or contingent obligations that might not be fully visible on balance sheets.” Treasury staff at the same meeting described bank exposure to private credit as contained, while noting the exact amount is unknown. The minutes of the Fed’s September meeting cited market commentary linking higher term premiums partly to heavy private debt issuance for AI infrastructure.

For policymakers, clear disclosure of leases and guarantees is what lets this kind of financing grow with confidence.

What to Look for in the Next Chip Deal

When the next deal is announced, read past the headline amount. You want five answers.

Who is the legal borrower?

Who stands behind the payments, and up to what cap?

What resale value the deal assumes for the hardware.

Whether the money is drawn on chip delivery or on power delivery.

And whether the debt term matches the chip's life.

If you invest, look at leases, guarantees and purchase commitments together, since that is how these deals are built. If you build, line up the financing draw with energization. If you regulate, ask for the backstops to be disclosed clearly.

The next filings will answer some of these. Watch whether the OpenAI package carries the same supplier support as the Anthropic platform. That single term will show how custom chips become a financeable asset.

Join Global Data Center Hub

Welcome to Global Data Center Hub. Join investors, operators, and innovators reading to stay ahead of the latest trends in the data center sector in developed and emerging markets globally.

Join Global Data Center Hub

The fastest new power plant on the largest U.S. grid may be one that is already built.

That is the lesson in the agreement Google and Constellation Energy announced on October 6.

The deal adds 890 MW of new nuclear capacity to PJM by upgrading existing reactors. It also gives 2,700 MW of existing nuclear supply a 15-year home. Together, the two contracts cover 3,590 MW.

So the deal does two jobs. It adds firm power on a timeline measured in years. It also keeps a large block of existing clean power on the grid as demand climbs.

What Google Actually Signed

The package has three parts, according to Constellation’s announcement and Google’s own post.

First, a 20-year power purchase agreement backs 890 MW of added output at 11 reactors in Illinois, Pennsylvania and New Jersey. Constellation says it will invest more than $4.3 billion in the work. The first uprate is expected by 2028. All 890 MW is due before the end of 2032.

Second, a separate 15-year supply agreement covers 2,700 MW from Constellation’s existing PJM fleet. Constellation says this keeps those plants on the grid.

Third, an expanded five-year technology alliance brings Google Cloud and Gemini Enterprise into Constellation’s operations, from uprate planning to plant output.

The deal also includes load shaping and demand response. Google can cut non-critical power use during high-stress grid events. The two companies also set up a framework to look at new generation, storage and demand response, with PJM first.

When you count new capacity, use 890 MW. That is about 25% of the 3,590 MW total. The other 2,700 MW is power that already flows, now under a long contract.

How Existing Reactors Add Power Faster

An uprate is an increase in the maximum licensed power of a reactor that is already operating. The Nuclear Regulatory Commission approves every increase. Utilities have used uprates since the 1970s.

Here, the upgrades cover turbines, steam generators and digital control systems, according to the American Nuclear Society. The plants already have land, operating licenses and grid connections. That saves much of the work that slows a new power plant.

Some steps remain. Each unit needs NRC approval. Each upgrade fits into planned outages at a working plant. And 11 units means 11 sets of engineering, permits and work crews, spread across four years.

Think of it like adding a floor to a house you already own. You still need a permit and a builder. But you skip buying land and running a new driveway.

Why PJM Needs the Megawatts

PJM needs this power. Its independent market monitor reports that the 2028/2029 capacity auction cleared 6,831.3 MW short of the reliability target. The shortfall was 208.7 MW in the two auctions before that.

The same monitor names data center load growth as “the primary reason” for those conditions. It estimates that adding existing and forecast data center load raised capacity market revenues by $29.37 billion across four auctions.

The monitor also cautions that data center forecasts carry real uncertainty. So grid planners face a hard task. They must add supply for demand whose timing is still moving.

Deals like this one help with that task. A buyer with a long contract funds new supply, and the new supply serves the whole grid.

The Rule That Decides Whether 890 MW Counts

PJM has proposed new rules for large loads, filed with FERC as Docket ER26-3515-000. Under the proposal, a large load that brings its own new capacity can avoid an interim service that may curtail uncovered loads during grid emergencies. PJM calls this Bring Your Own New Capacity.

As of October 9, PJM’s large load page marked the registry and the interim service as pending FERC approval.

The Google and Constellation deal aligns with that proposal's direction.

According to POWER magazine’s review of the filing, an uprate qualifies when it raises both installed capacity and capacity interconnection rights. Capacity that has already cleared a PJM auction generally does not qualify, which shapes how the 2,700 MW agreement fits.

The registration path for the 890 MW will become clear as the rules and the unit filings move forward.

Accreditation is the second step to watch. PJM rates each unit to decide how much accredited capacity it provides. That rating lines up against the 6,831 MW shortfall.

The Terms That Will Shape the Economics

The commercial terms are private. The parties set the power price, the escalation formula, the treatment of clean energy credits, how capacity value is shared, and the timing provisions.

Those terms will decide the return on Constellation’s $4.3 billion and Google’s delivered cost of power. The 20-year term already tells you something useful. It aligns the revenue period with a long-lived capital program, making the work easier to finance.

The asset list will fill in over time. A Constellation spokesperson named six sites to the American Nuclear Society: Braidwood, Byron, LaSalle, and Quad Cities in Illinois; Limerick in Pennsylvania; and Salem in New Jersey. Those sites hold 12 reactors, and the deal covers 11. The unit list and the split of the 890 MW across units are the next disclosures to look for.

Salem has shared ownership. PSEG owns 57%, and Constellation owns 43%. How the co-owners share costs and added output is another detail to watch.

On cost to other customers, Google says the deal comes at no cost to other ratepayers. Constellation ties it to the federal Ratepayer Protection Pledge. Transmission work and capacity market results will show how that plays out across the system.

The demand response terms matter too. The number of megawatts Google can cut, for how long and on what notice, will show how much flexibility the deal adds in a grid emergency.

Uprate Deals Are Becoming A Pattern

I would read this deal as part of a series. One week earlier, on September 30, Amazon and Constellation signed a 20-year agreement at Calvert Cliffs in Maryland. It covers 690 MW, including a 190 MW uprate, backed by more than $3 billion of investment.

POWER also lists a 176 MW Walmart agreement at Dresden that includes 30 MW of uprates, and a proposed Google arrangement with Georgia Power for about 96 MW of uprates at Vogtle and Hatch, subject to state approval.

Google says its uprate and restart deals have enabled more than 1.5 GW of new nuclear capacity in the U.S. That sits beside a planned restart of the Duane Arnold plant in Iowa with NextEra and an advanced reactor program with Kairos Power. Each runs on its own timeline, and uprates come first.

The shape is clear. Large buyers sign long contracts, and plant owners use that revenue to add output at reactors already on the grid.

Leave a comment

The Lesson Travels Further Than the Reactors

The Bankability Gap is the distance between capacity that is announced and capacity that can secure power, financing, permits, equipment and customers. This deal closes much of that gap in a specific way. The 20-year contract answers the financing question. The existing sites answer most of the land and grid questions. Licensing and the PJM rule remain.

That is why the model travels in two forms.

In markets with nuclear fleets, like Canada, South Korea, Japan and parts of Europe, the asset itself may repeat. There may be output to add inside plants that already run, paid for by a long contract with a large buyer.

In most of the Gulf, Africa, Latin America and Southeast Asia, the contract is the part that travels. A creditworthy buyer signs a long offtake. That offtake pays for new supply on the shared grid. The buyer also agrees to cut load when the grid is under stress.

That second form matters most for emerging markets. The International Energy Agency reports that emerging and developing economies outside China hold about half the world’s internet users but less than 10% of global data center capacity. Reliable power is the condition it names first.

What to Watch Next

Watch FERC. Its ruling on PJM’s large load rules will decide how bilateral deals like this one count as new capacity.

Watch the NRC docket. Unit-level applications will show which reactors, how much added power each one gets, and when.

Watch 2028. The first uprate is the first real test of the schedule.

And when you read “new capacity” in the next power deal, check two things before you count it. Is the output added, or already flowing? And how will the grid operator accredit it?

Join Global Data Center Hub

Welcome to Global Data Center Hub. Join investors, operators, and innovators reading to stay ahead of the latest trends in the data center sector in developed and emerging markets globally.

Join Global Data Center Hub

Edition No. 72

THE WEEK’S THESIS: THE BUYER BRINGS THE POWER

The signed customer is now the asset that gets financed. Power, land and buildings matter. But this week, money moved fastest toward projects with a named buyer already locked in.

DayOne took its customer book to Nasdaq. CoreWeave committed to India as the only tenant on a new campus. Lenders are reportedly building vehicles that lend against chip buyers’ promises to pay. The tenant came first, and the capital followed.

The projects still waiting have everything except a buyer. TeraWulf has the power contract and no tenant. Applied Digital and atNorth have Finnish power and are still marketing it. Then there are the projects with a buyer and no permit, which is the constraint further down.

The structure behind DayOne’s filing is in The Mechanism below.

MARKET SIGNALS

Google pays for nuclear output it doesn’t own. Google and Constellation signed a 20-year deal to add 890MW through upgrades at 11 existing reactors. What counts as new supply on PJM’s books is still open.

Amazon builds beside a dedicated gas plant. AWS filed a preliminary plan for 36 data centers on 1,100 acres at the Homer City Energy Campus, about 50 miles east of Pittsburgh. A 4.5GW gas plant being built by Homer City Redevelopment will power the site. More than 1GW of that output goes back to the grid. The campus is sized to its power plant, and the grid gets the surplus.

Washington and Seoul pick a Texas power builder. The US Commerce Department and South Korea selected Related Companies, NextEra, and Lewis Energy Group to develop Project Star in Encinal, Texas. The $22.3 billion campus will provide 6.47GW of gas-fired power for an adjacent 5GW data center site. Korea and the US will jointly own the power campus under their trade deal. Sovereign money now funds the power block, and the data center follows it.

Onsite engines become a repeat purchase. Wärtsilä booked an order for 282MW of onsite power at a US data center built by an unnamed independent power producer. Fifteen gas engines arrive in 2028, with full operation in 2029. It is Wärtsilä’s seventh US data center order. Its US data center total now tops 3GW. Seven orders are a product line.

Finland’s old paper mills turn into AI campuses. Sesterce plans an AI campus worth more than €10 billion at the former Kaipola paper mill in Jämsä. The first phase is 200MW, rising to 600MW, and the company says an unnamed anchor customer is already committed. It also pledges 1MW of new renewable supply on the Finnish grid for every 1MW it uses. Old mills come with grid connections, and the connection is the scarce part.

CAPITAL WATCH

Broadcom, SpaceX, Oracle and the lenders. Broadcom is arranging more than $50 billion in debt to fund OpenAI’s custom chips, the Wall Street Journal reported. SpaceX is reportedly in talks to borrow ~$40 billion for Nvidia chips, split between ~$10 billion of bank loans and ~$30 billion of investment-grade debt. Oracle is reportedly weighing a vehicle that buys chips with investor money and leases them back to Oracle. Chips are leaving company balance sheets and moving into lease vehicles.

Firmus pulls its IPO. Nvidia-backed Firmus shelved its planned ~$5 billion Australian listing amid growing scrutiny of AI spending, Reuters reported. One Asia-Pacific operator stepped back from public markets in the same week another filed for New York. The next test is whether public buyers pay up for the one with signed tenants.

GLOBAL LENS: THIS WEEK BY REGION

North America: power contracts set the build order

TeraWulf signed an amended electric service agreement with Kentucky Power that doubles contracted capacity at its Muskie campus to 1GW. The second 500MW phase moves up from 2030 to 2029, pending state commission approval. The first 500MW starts ramping in 2028. No tenant has signed, so TeraWulf is selling a power date. In this market, the date is the product.

Brookwood’s city council voted to back the county development authority’s agreement with Applied Digital for its $3.2 billion Delta Forge 2 campus in Alabama. The vote followed weeks of local pushback over environmental impact and transparency. The Alabama Public Service Commission still has to approve the power agreement with Alabama Power. Local consent is done. The power contract is the last gate.

Europe: Finland’s power draws the developers

Applied Digital signed an agreement for access to up to 1GW of potential power in Finland, its first project outside the US. Initial power is expected from 2028. The company is already marketing the site to hyperscalers and says it structured the deal to limit early capital at risk. A US developer bought its first foreign option, and it bought Finnish power.

atNorth plans a new campus in Salo, west of Helsinki, at about €2 billion. atNorth has secured power for the first 75MW phase, with a path to 230MW through a Fingrid substation. Customer servers could add up to €10 billion more, by the company’s estimate. Securing first-phase power is what lets atNorth announce before naming a customer.

Asia-Pacific: tenants arrive before the buildings

CoreWeave plans to enter India with 240MW across three 80MW buildings at AdaniConneX’s Taloja campus in Navi Mumbai. It will be the sole tenant, with an option to double capacity. The first phase is expected in mid-2028 and will run NVIDIA Vera Rubin. A US AI cloud is committing to India at a single campus, at hyperscale size.

The New South Wales planning department approved Stockland’s A$630.4 million data center at Macquarie Park in Sydney. The 10-story building has 49MW of total power and 34.3MW of IT load. The consent bars its backup generators from selling power into the grid. Those generator and emissions caps are now the template other Sydney projects will face.

Share

CONSTRAINT OF THE WEEK: PERMISSION IS THE BOTTLENECK

Permission is now the scarcest input in this market. Finland’s Supervisory Agency ordered Google to suspend site work at planned data centers in Muhos and Kajaani. Regulators cite tree clearing, new roads, and drainage changes made before environmental assessments were done. The company said it fell short of its own standards. The sites stay frozen until those assessments are complete.

Raleigh’s city council voted 7-1 for a six-month moratorium on new data centers. Residents had asked for 32 months. Staff must write rules on noise, water and pollution by April 6, under a 2024 state law that bans downzoning. Six months sounds short. It still stops every new filing in the city.

In Melbourne, Monash council rejected Centuria’s 38MW, A$137 million expansion at Clayton. Its own planners had recommended approval. Councilors pointed to Victoria’s new plan, which proposes 150-meter buffers from homes and self-supplied renewable power. Those rules don’t start until 2027. The council applied them early anyway.

In India, the Andhra Pradesh High Court let Adani’s Vizag campus keep building for Google’s AI hub, at the developer’s own risk. A public interest case challenging state approvals is still open. The court said money spent now can’t be used to argue for relief later. The next bottleneck is legal risk on capital already in the ground, and lenders will price it.

THE MECHANISM

Two customers carry the book, and the listing funds the build

DayOne is a data center platform headquartered in Singapore. It was the international business of China’s GDS until it spun out in 2025, DCD reports. It runs campuses across Malaysia, Singapore, Indonesia, Thailand and Japan, with Finland and Spain in development. On Oct. 6 it filed a Form F-1 for a Nasdaq listing under the ticker DODC.

Revenue reached $484.3 million in 2025, up from $178.1 million a year earlier. The first half of 2026 alone brought $512.0 million. The company lost $367.1 million in 2025. It has 962MW in service and 1,328MW under construction.

Seven customers hold 2.3GW of bookings due by the end of 2028. Finishing them will cost about $11.4 billion, by the filing’s estimate. One customer paid 69.2% of first-half revenue. The filing describes it only as a global technology company with a leading short-form video platform. The second-largest paid 15.1%, and Malaysia supplied 87.0% of all revenue.

So two customers and one country carry a build already backed by $4.35 billion of long-term borrowings. Lenders funded the first gigawatt against those contracts. The listing asks public buyers to fund the next one on the same promise. The filing lists access to power as a core risk, and almost all revenue sits on Johor’s grid.

The share count, price, and proceeds are still blank. Reuters reported in February that DayOne aimed to raise ~$5 billion at a ~$20 billion valuation, per DCD. Customer contracts allow early termination, with penalties applying only after “a certain number of years.” The filing doesn’t say how many.

Leave a comment

NOTABLE TRANSACTIONS

Macquarie Capital: investment in Zerra, amount undisclosed

Macquarie Capital’s Infrastructure Principal team invested in Zerra, a hyperscale developer headquartered in Singapore. AGP, which founded Zerra, remains the majority owner. The amount wasn’t disclosed.

What makes it notable is the order of the steps. Macquarie first joined Zerra’s Western Downs Digital Park in Queensland as a consortium partner and adviser. Then it bought into the developer itself.

Western Downs plans up to 1.44GW across four stages and holds a long-term lease commitment from Anthropic. Zerra’s wider pipeline tops 2GW across Australia, Japan and India. This is Macquarie Capital Principal Investment’s fifth investment since April.

Watch the Western Downs development application, which is still under assessment. The tenant has committed, so the approval is the gate that binds.

Sabey Data Center Properties: Ares lifts its stake above $500 million

Ares Secondaries funds expanded their minority equity stake in Sabey Data Center Properties. Their total commitment now tops $500 million. Sabey is a joint venture between Sabey Corporation and National Real Estate Advisors. The follow-on builds on a stake first announced in July 2026. The tranche size wasn’t disclosed.

The buyer is what makes it notable. A secondaries fund is adding to a minority stake in a private platform, and the founding partners keep control.

Sabey owns and runs 275MW across US markets and expects its pipeline to reach about 737MW by 2033. National Real Estate Advisors manages money for the National Electrical Benefit Fund, a union pension trust. Pension money and secondaries money now share the same build risk.

Watch how the next 460MW gets funded. More partial stakes would show minority capital can carry a platform this size without a full sale.

AirTrunk: ~$1 billion more for its Tokyo TOK1 campus

AirTrunk, owned by Blackstone and CPP Investments, plans to invest another ~$1 billion in its TOK1 campus in Inzai, east of Tokyo. Reuters described it as an added investment, while DCD, citing Nikkei Asia, reported a ~$1 billion green loan led by SMBC, MUFG, and UOB.

What makes it notable is the repeat. AirTrunk closed a ¥191.6 billion green loan for the same campus in March, then the largest data center financing in Japan.

The new money goes to added capacity and liquid cooling for dense AI racks. AirTrunk puts its Japan capacity at 530MW today and targets more than 1GW. Its CEO said customers are looking to Japan for geopolitical reasons.

Watch for AirTrunk’s own release naming the lenders and terms. Two bank-led financings on one campus in seven months would mean Japan’s lenders treat AI capacity as core business.

THE FRAMEWORK: WHO PAYS FOR THE POWER

Each of this week’s power deals answers one question differently: who builds the supply, and who pays for it. The chart below lines them up.

THE QUESTION FOR NEXT WEEK

Will Google’s Finnish subsidiary satisfy regulators by Oct. 14? Tuike Finland must explain its site work and set out next steps by that date, or face enforcement. A clean answer puts Muhos and Kajaani back on the assessment timeline. A weak one tells every developer in Finland that early site work now carries a real cost.

If you are new to Global Data Center Hub, visit our Start Here page for more information about the newsletter.

If you found this newsletter useful, share it with a colleague.

Have a great week.

— Obinna

Share Global Data Center Hub

Welcome to Global Data Center Hub. Join investors, operators, and innovators reading to stay ahead of the latest trends in the data center sector in developed and emerging markets globally.

This article is the 17th article in the series: Ground to Grid: A Free 21-Lesson Guide to Mastering Data Center Development

Join Global Data Center Hub

Why Commissioning Is Where the Building Becomes a Facility

A completed building is not yet an operational data center. Commissioning proves it works, under load, under failure, and to a standard the tenant will approve. Until that happens, you have a building full of untested systems.

Commissioning sits between construction and live load. Construction ends when the equipment is installed. Live load begins when the tenant’s servers go in and revenue starts.

Commissioning tests the systems together, and it tests them when things break. A generator can pass its own test and still fail on the night the utility drops and the cooling plant has to restart at the same moment.

Certification adds outside confirmation. No tenant or lender can witness every test. A recognized certificate, and the tenant’s own sign-off, stand in for that.

The Stages of Commissioning

ASHRAE, the U.S. engineering society behind the main commissioning guidance, numbers this work as five levels, from factory acceptance to integrated testing of the whole facility. Each contract defines its own numbering.

1. Equipment and Component Checks

The first stage checks each piece of equipment against its specification. Switchgear, uninterruptible power supply (UPS) units, generators, and chillers are tested at the factory before they ship, in what is called factory acceptance testing.

Then everything is checked again on site, because a truck ride and a field crew can undo a clean factory result. Crews confirm that what arrived matches what was approved, then start each unit.

Everything after this assumes the equipment underneath is sound. If it is not, every later test is measuring the wrong thing.

2. System-Level Testing

The second stage tests whole systems. The power train runs as one system, from switchgear through UPS to the distribution that feeds the racks. So does the cooling plant, from chillers through pumps to the air handlers.

Each system is run through normal operation, maintenance, and the failure of its own parts, using load banks (temporary electric loads that stand in for servers).

Engineers are watching for one thing. Do capacity, controls, and alarms behave the way the design says they should?

3. Integrated Testing Under Load

The third stage, integrated systems testing, runs the whole facility at once, with power, cooling, controls, fire systems, and operators all in play.

The team stages the events the site was built to survive. They cut utility power and watch the generators start while the UPS systems carry the load through the gap.

Then they fail a redundant component on purpose and check that its backup takes over.

If you remember one stage, make it this one. A facility that only works when nothing fails has not been proven.

4. Client Validation and Sign-Off

The last stage belongs to the customer. Wholesale and hyperscale tenants often audit the facility against their own standards and run their own acceptance tests. Their sign-off is the final approval that hands the facility to operations.

The Certifications That Matter

1. Uptime Institute Tier Certification

Uptime Institute’s Tier Classification System rates how a site’s power and cooling are arranged, from Tier I to Tier IV. Tier III means every component and distribution path can be taken out for planned maintenance without affecting IT. Tier IV goes further. A single unplanned failure does not affect IT either.

Uptime certifies in stages. Tier Certification of Design Documents (TCDD) reviews the drawings, and design awards expire after two years.

Tier Certification of Constructed Facility (TCCF) inspects the built site and witnesses live demonstrations. So when someone says “Tier III design,” ask which one they mean.

Tier is the benchmark most tenants and lenders recognize. It does not promise a percentage of uptime.

2. ISO Standards

ISO writes standards. Independent certification bodies issue the certificates. ISO 9001 covers quality management, ISO/IEC 27001 information security, ISO 14001 environmental management, and ISO 50001 energy management.

These certify how an organization is run. None of them rates a building’s physical resilience. The scope statement on the certificate tells you what is actually covered.

3. Tenant-Specific Standards

Hyperscalers set their own commissioning requirements, and these often go further than general certification. They can cover liquid-cooling interfaces and block-by-block handover. It is often the hardest bar to clear before approval.

Why Commissioning Failures Are Expensive

It comes down to timing. By the time integrated testing runs, the building is finished and the tenant is waiting.

A failure at this stage can trigger months of rework and retesting, and every week of delay pushes back the day revenue begins.

Many start earlier. A defect that quality control should have caught during construction shows up here, behind finished walls, where it costs more to reach.

The riskiest failures are transitions between power sources. In Uptime’s 2025 survey, power caused 45% of impactful outages, with UPS, transfer switch, and generator failures prominent. Integrated testing is where those transitions get proven.

A Real-World Example

Hyperscalers often bring their own commissioning teams and will not sign off until the facility meets their standards. A failure they catch can mean months of rework before the building earns a dollar.

The failures nobody catches cost more. On August 30, 2023, a thunderstorm caused a utility power sag near a Microsoft Azure data center in Sydney. All five running chillers in one hall tripped. Of the two standby chillers, one restarted and held.

Microsoft’s preliminary review said three staff were on site that night, too few to restart the chillers in time. Microsoft powered down two data halls to protect hardware. Some services were affected for up to 46 hours.

Restarting a cooling plant after a power sag, with the staff actually on shift, is what integrated testing exists to prove.

Strategic Implications

For developers: Give commissioning its own schedule and its own budget. Bring the commissioning provider in at design, so the tests are planned before anything is built.

Protect the integrated test window when construction runs long. The failures you find there usually trace back to shortcuts taken earlier.

For investors: Certification and tenant sign-off are the proof the asset performs.

Financing that assumes them before they happen is financing a risk. Ask for the integrated test report, the list of tests that failed, and proof that each one was retested.

For operators: The commissioning results set your operating baseline and show you the failure modes you will have to manage.

Take part in integrated testing with the people who will run the site, and run your own procedures during the tests.

Sydney showed what happens when the night shift is too small to recover.

For policymakers: Recognized certification standards narrow the information gap between builders and the people who finance and lease their sites.

Without a common reference, every lender and tenant checks everything from scratch, and both financing and leasing slow down.

When public rules or procurement cite a standard, name the exact certificate type and scope.

Key Takeaway

Construction produces a building. Commissioning produces a data center.

The difference is proof. The facility has to be tested under failure, certified to a standard, and approved by the tenant who will pay to use it. Until then, the tenant and the lender are taking the site on trust.

Proof also sets the calendar. Every week lost at integrated testing is a week of revenue that never arrives.

So the date that matters is the day the tenant signs off. Plan for it from the first design review. If you treat commissioning as the last box to tick, you will find your failures at the project’s most expensive moment.

Do you think tenant-specific commissioning standards or third-party certification give more real assurance about a facility? Why?

Join Global Data Center Hub

Welcome to Global Data Center Hub. Join investors, operators, and innovators reading to stay ahead of the latest trends in the data center sector in developed and emerging markets globally.

Join Global Data Center Hub

Most people meet data centers through headlines.

A tech company announces a giant new campus, a warehouse full of the computers that run AI tools like ChatGPT and puts a number on it in megawatts.

That number is a promise. The building behind it tends to show up years later, and sometimes later than even its builders planned.

So how long does it take to build a data center? The short answer is years. The longer answer starts in Denton, Texas.

Core Scientific spent years mining bitcoin before agreeing to convert its Denton site into an AI campus for CoreWeave, which rents computing power to AI firms. Core Scientific puts the full campus at 262 megawatts.

The Wall Street Journal reported on December 15, 2025, that CoreWeave planned to use it to serve OpenAI, the maker of ChatGPT.

The trouble showed up early. Core Scientific’s quarterly report for the period ending June 30, 2025, warned that weather and construction delays had pushed its delivery dates later in the year.

Then, on November 10, 2025, CoreWeave lowered its forecast for the year. It blamed a third-party data center developer that was behind schedule, and said its customer had agreed to move the contract dates.

A month later the Journal filled in the details. Summer rain and wind had kept crews from pouring concrete at Denton for about 60 days. Completion slid back by several months. For a site that size, one wet summer was enough.

The site did come online, in stages. On March 2, 2026, Core Scientific told investors that Denton was running about 130 megawatts that customers were paying for, roughly half the campus.

The cost was real. Capacity meant for the company behind ChatGPT arrived months late, and CoreWeave had to cut its outlook for the year.

Denton was unlucky with the weather, but the shape of its problem is common. AI companies sign up for computing space long before they need it.

Every one of those contracts then has to pass through the same physical steps: power reaching the site, heavy equipment arriving, crews building, engineers testing. If one step slips, the whole building waits.

The Long Road From Signed Contract to Live Building

A data center comes to life in stages, and some stages take far longer than others.

Power comes first. The builder asks the local utility for electricity.

The utility then studies whether its high-voltage lines, the tall steel towers that carry power across a region, can deliver that much to one spot. If those lines are already full, someone has to build new ones.

Next comes the gear that links the building to the grid.

The key piece is the transformer, a large machine that turns long-distance electricity into a level a building can use. After that, crews build the halls and install the cooling and backup power, and engineers test the whole system.

Only at the very end do customers roll in their servers, the flat computer trays that do the actual work, and start paying.

Think of a restaurant that takes bookings before it opens. The bookings are real. But nobody eats until the kitchen is built and the gas is hooked up.

NEXTDC, a large Australian data center company, has put this gap in its own numbers. In a stock exchange release on July 21, 2026, it reported customer contracts for 740 megawatts as of June 30, 2026.

Customers were paying for 175 of them. Put another way, for every four megawatts signed, about one was running.

NEXTDC expects the rest to switch on in stages through its 2030 financial year.

Headlines Count Contracts. The Grid Counts Buildings.

Most readers would assume an announced data center is running, or close to it. It is a fair guess.

Announcements and working buildings use the same unit, the megawatt, so the numbers look like they mean the same thing.

Companies report what they have signed because a signed contract is a real commitment with real money behind it.

News stories repeat that figure because it is the biggest, newest number on the page.

Look closer and each megawatt has a life story. A contracted megawatt is a reservation. Later, once power reaches it, it becomes energized.

Only at the last stage does it run any AI, because a billing megawatt has servers inside doing work for a paying customer.

NEXTDC’s 740 and 175 are both true. They describe different points in that life story.

Core Scientific drew the same line on March 2, 2026, when it said it would report megawatts only once they start billing.

Power Lines and Transformers Run on a Slower Clock

The slowest steps usually happen outside the building, in the power grid, and they take years to fix.

Northern Virginia has lived through this. Loudoun County, around the town of Ashburn, holds one of the densest clusters of data centers anywhere.

In July 2022, Dominion Energy, the main utility there, told data center builders that its high-voltage lines in the eastern part of the county could no longer carry power to new buildings on the schedule they expected.

Builders who had already bought land and drawn up plans were left waiting on the utility, with no firm date for when the power would come.

Dominion started connecting new buildings again that fall, in small amounts. Even so, DatacenterDynamics reported in October 2022 that some planned projects faced waits of up to four years for their full power.

The fix is a loop of new high-voltage lines around the area people call Data Center Alley, and it is arriving one piece at a time.

Dominion’s project page shows the first equipment on its new Mars to Wishing Star line switched on in January and February 2026.

The last piece, the Golden to Mars line, won final route approval from Virginia’s State Corporation Commission on June 29, 2026.

Dominion has said that line must be in service by June 1, 2028. Count from the first warning to the last line, and the repair takes about six years.

Equipment brings its own wait. Wood Mackenzie, an energy research firm, found in its survey for the second quarter of 2025 that large power transformers took an average of 128 weeks to arrive after ordering.

That is about two and a half years for one machine, and a large campus needs more than one.

A storm can stop a concrete pour for a couple of months. A new power line needs an approved route, settled through public hearings, before the first tower goes up. A transformer needs a slot in a factory’s order book. Those waits run to years.

Who Waits When a Data Center Runs Late

Start with the companies counting on the space. When Denton slipped, CoreWeave had to tell its shareholders, and OpenAI had to wait longer for computing it had planned around.

An AI tool can only serve as many people as the buildings behind it can handle, so a late building leaves less room to grow for the app on your phone.

Then there are the neighbors. The new lines that end these long waits have to run somewhere.

In Loudoun County, homeowners fought the Golden to Mars route, which will bring tall new towers close to houses, and one neighborhood group asked regulators to reconsider.

If you live near a planned data center, the slowest part of the project may be a power line through your neighborhood, and the public hearing that decides where it goes.

And if you simply follow AI news, the gap changes how every big announcement reads. A headline marks the start of a long build.

The AI features you use next will run in buildings that have already made it all the way through.

Four Clues Hidden in Every Data Center Announcement

Start with the verb next to the number. Signed and contracted describe reservations, while energized and billing describe buildings doing work.

Company announcements and quarterly reports often use both in the same paragraph, and NEXTDC’s July 2026 release shows how far apart the two numbers can sit.

Watch for dates that drift. When quarterly filings keep pushing delivery dates later, as Core Scientific’s did through 2025, the schedule is under strain well before a headline says so.

Those filings sit on company investor pages and in the public database of the US Securities and Exchange Commission.

Look for the power line. A data center that needs new high-voltage lines will show up in a state utility commission case or on a county planning agenda, the way the Golden to Mars line went through Virginia’s State Corporation Commission.

Those cases show how many years the power side expects to take, often in more detail than the developer’s announcement.

Finally, set the opening date against the equipment wait. Large transformers took about 128 weeks to arrive in Wood Mackenzie’s 2025 survey.

If a campus promises to open much sooner than that, either the equipment was ordered long before the announcement, or the date is likely to move.

Announcements move at the speed of a press release. The buildings that run AI move at the speed of the grid and the factories that supply it.

Can utilities and equipment makers speed up enough to close that gap, as the announcements grow larger? Every new campus headline puts that question back on the table.

Join Global Data Center Hub

Welcome to Global Data Center Hub. Join investors, operators, and innovators reading to stay ahead of the latest trends in the data center sector in developed and emerging markets globally.

Join Global Data Center Hub

Every month, I track the top data center and AI infrastructure transactions globally. In August 2026, I tracked 125 deals.

Ten things stood out to me about this month’s top transactions:

1. AI labs became the anchor tenants.

Anthropic stands behind four of August’s transactions, OpenAI behind three. The marginal buyer of capacity isn’t a hyperscale cloud provider anymore. The labs are signing for it themselves.

2. NVIDIA appeared in seven transactions in five different roles.

Guarantor of a $105 billion exposure cap. Reported acquirer of a stake in a power developer. Strategic investor in two more. Participant in a $2 billion equity round. Architect of a blueprint to mobilize $500 billion of third-party capital. And GPU counterparty in a two-million-chip agreement. I cannot recall one company taking up this much of the sector’s structure in a single month.

3. Capacity is being sold like infrastructure concessions.

Five deals carried tenors of 15 to 20 years. That’s build-to-suit territory, not colocation.

4. Power now picks the site.

A Department of Energy brownfield, a crypto mine’s interconnection, Marcellus gas rights, Nordic hydro, a Texas wind farm. Nothing in the top tier was sited for connectivity.

5. Behind-the-meter gas is America’s speed play.

At least six U.S. deals paired data centers with dedicated on-site generation rather than waiting in a grid queue.

6. Utility regulators became dealmakers in the US.

Two state public service commission approvals and a federal energy authorization were among the month’s most consequential events. Power contracts are the new permits.

7. Sovereigns showed up as principals, not regulators.

Two Saudi builds, Armenia’s government-funded AI factory, Mongolia’s 863MW MoU program, Germany’s sovereign-cloud campus, Singapore’s capacity tender, and a UK state guarantee sitting under private debt.

8. The financing toolkit stretched to fit the asset class.

August produced a residual-value guarantee capped at $105 billion, $5.75 billion of convertible notes, an Islamic Murabaha facility from a Qatari bank, a debut Australian dollar bond from a hyperscaler, a private equity continuation vehicle, and a telecom carve-out selling minority stakes to global funds. No single instrument fits AI infrastructure yet, so the market is reaching for all of them.

9. Capital moved upstream into the companies that build the buildout.

Partners Group took majority control of AVK, the on-site power and energy-as-a-service platform, with an initial investment above $1 billion. H&MV Engineering recapitalized at a €1.4 billion headline valuation. Kingspan paid a firm €850 million for BMC Manufacturing, and H.I.G. Capital acquired Phoenix ME. Four transactions in one month bought the supply chain rather than the data centers themselves.

10. Crypto infrastructure found its second life.

Conversions and grid-position acquisitions turned mining assets into AI capacity in three states, because an approved interconnection is worth more than the business running on it.

Leave a comment

Top 10 Global Data Center Deals

The ten highest-ranked transactions of the month span all five regions: four in North America, three in Europe, one in Asia-Pacific, one in the Middle East, and one in South America.

1. OpenAI–SB Energy: 8GW in Ohio for Twenty Years (North America)

OpenAI, the AI lab behind ChatGPT, agreed to secure about 8GW of IT capacity at SB Energy's PORTS-Pike campus, a developer of power and AI data-center infrastructure, on the former Department of Energy site in Pike County, Ohio, under a twenty-year lease. SB Energy builds, owns, and operates; delivery runs in phases from 2028, with the first roughly 800MW riding on existing AEP infrastructure and the remainder requiring new generation and transmission.

Development still depends on permits, environmental reviews, and financing, so treat the 8GW as a plan, not a certainty. What nobody can walk back is who signed: a model developer, not a cloud provider, took the industry’s largest tenancy, on a site chosen for its power.

2. NVIDIA’s $105 Billion PORTS-Pike Guarantee (North America)

This is the deal that makes the first one financeable. NVIDIA, the chipmaker whose processors power the AI buildout, entered definitive residual-value guarantee arrangements covering the initial 4.25GW of the same Ohio campus, with a disclosed aggregate maximum payment obligation of $105 billion. No cash moves at signing; the figure is a ceiling on exposure, not spending.

Each guarantee phases in as the corresponding lease commences. An option to support roughly another 3.75GW is excluded. Why does a guarantee like this exist at all? Because no lender syndicate will underwrite a single campus at this scale on a young tenant’s covenant. So the chip supplier stepped into the credit stack behind its own demand chain.

3. Volta and an Unnamed AI Lab: $10 Billion at Tydal (Europe)

Europe’s largest deal is a $10 billion strategic partnership between Volta, an AI-factory developer, and an unnamed leading AI lab, anchoring the 121MW Tydal AI factory in Norway that Volta is developing alongside Bitdeer. The same campus carries a fully disclosed sixteen-year, $4.7 billion lease between Bitdeer and Volta Tydal.

The two are legs of one transaction story covering the same 121MW of critical IT load, running entirely on renewable power. The values must not be stacked and the megawatts count once. It reads like Ohio at European scale: a long tenor, lab demand, and a site picked for cheap, clean electricity.

Share

4. HUMAIN–MIS: Saudi Arabia’s Sovereign Buildout Quadruples (Middle East)

HUMAIN, Saudi Arabia’s state-backed AI company, expanded its data center construction award with Al Moammar Information Systems, the Saudi-listed IT company, by 200MW, taking the contracted program from 50MW to 250MW. The amended contract value exceeds $2.3 billion, according to a Saudi Exchange disclosure that the post-expansion contract exceeds 689 percent of MIS’s 2025 revenue.

What strikes me is the speed. A sovereign AI program quadrupled its physical footprint through a mid-contract scope change, at a pace you would normally associate with hyperscale procurement.

5. Riot–Anthropic: 191MW and a Second Life for Crypto Infrastructure (North America)

Riot Platforms, the U.S. Bitcoin mining company, signed an executed twenty-year lease with Anthropic, the frontier AI lab, covering 191MW of critical IT capacity at its Rockdale, Texas campus, with roughly $9.1 billion in contracted revenue over the initial term.

That figure is lease revenue over two decades, not construction cost; optional extensions could take it toward $16.1 billion. The money is striking, but the conversion matters more: a Bitcoin mining site became long-duration AI capacity because it already held an approved grid interconnection. In Texas, the queue is the moat.

6. Nscale–Anthropic: $45 Billion in West Virginia (North America)

The month’s largest contract by value: a six-year agreement under which Nscale, a provider of dedicated AI compute, gives Anthropic 460MW at its Monarch campus in West Virginia, valued at $45 billion. The figure is compute-rental revenue rather than project cost.

Note where the electricity comes from: on-site natural gas generation, because six-year AI demand can’t wait in a grid queue. Labs are paying for speed, and speed currently burns gas.

7. CoreWeave Enters Asia with 360MW in Indonesia (Asia-Pacific)

CoreWeave, the AI cloud provider, announced its first Asia-Pacific expansion: three Indonesian facilities totaling exactly 360MW of contracted IT power, expected online in 2028, all owned and operated by CoreWeave.

An AI cloud, not an enterprise cloud, is what finally pulled major western compute investment into Indonesia. New markets are now opened by training and inference demand rather than corporate IT.

8. TERRANOVA Breaks Ground on Latin America’s Largest AI Campus (South America)

TERRANOVA, an Actis-backed developer of purpose-built AI campuses in Brazil, began construction in Campinas on a campus with 216MW of initial verified IT load, expandable to 386MW, served by dedicated substation and transmission infrastructure and targeting global hyperscale and AI customers.

Initial-phase investment is stated only as exceeding $500 million. It is the region’s lone entry on this list, and the pairing matters: purpose-built AI capacity at three-digit megawatt scale is under construction in South America in the same month the continent recorded zero closed financings.

Share

9. Firebird Lights Up Armenia (Europe)

Firebird, an AI infrastructure developer, brought its Hrazdan AI factory in Armenia operational in August. The NVIDIA DSX facility, described as the largest AI factory in the CIS region, is funded by a disclosed $500 million investment from the Armenian government.

A country of three million people bought itself frontier-grade compute as a matter of national capability. Small states have concluded that AI infrastructure is closer to an airport than an office park: something a nation builds because sovereignty requires it.

10. Schwarz Group’s €5.6 Billion Answer to Digital Sovereignty (Europe)

Schwarz Group, the German retail group that owns Lidl, disclosed plans for a data center campus at Dummerstorf in northern Germany, with an estimated investment of approximately €5.6 billion (about $6.5 billion), running exclusively on renewable power during regular operations.

The capacity serves the group’s own ecosystem and its STACKIT sovereign-cloud platform, and the project remains proposed, subject to approvals. Europe’s sovereignty debate has mostly produced policy papers; here it produced a retailer committing billions to build the alternative itself.

Step back from the top 10 global data center transactions and the list reads as two markets running on different fuel.

In developed markets, the story is financial engineering: a $105 billion guarantee standing behind an Ohio campus, six of the world’s largest financiers designing platforms to mobilize $500 billion, twenty-year leases priced like infrastructure concessions.

In emerging markets, the story is the state: Armenia’s government wrote the $500 million check for its own AI factory, Saudi Arabia quadrupled one sovereign build mid-contract and committed $5 billion to another at NEOM, and Mongolia signed 863MW of memoranda.

And in Brazil, the builders simply went first: TERRANOVA broke ground on a 216MW campus with more than $500 million committed to its first phase, ahead of a continental financing market that is still to open.

Developed markets are inventing new instruments to fund the machine. Emerging markets are reaching for the oldest one: the sovereign balance sheet. Both are building. The question August left open is which system gets to price the next gigawatt.

Share Global Data Center Hub

Welcome to Global Data Center Hub. Join investors, operators, and innovators reading to stay ahead of the latest trends in the data center sector in developed and emerging markets globally.

Join Global Data Center Hub

Google Cloud plans to double its technical infrastructure in Brazil by 2030. It said so at Google Cloud Summit Brasil 2026 in São Paulo on September 24.

Eduardo López, Google Cloud’s president for Latin America, put a time frame on it.

The company built its Brazil capacity over nine years, from 2017 to 2026. It now plans to double that capacity in four.

If you read only the headline, you see twice the buildings and twice the power. López answered that reading at the press conference. He said Google measures the plan in processing power and service capacity.

He said floor space is a separate number, because newer hardware does more in less room.

The one idea I want you to take away is simple. A capacity pledge counts work. You have to translate it into megawatts, buildings, leases, and cable access before anyone can lend or build against it. I use the Bankability Gap to show you how.

The Bankability Gap is the distance between capacity someone announces and capacity that can line up power, money, permits, equipment, customers, and steady returns.

What a Capacity Pledge Measures

A capacity pledge is a public promise to grow how much work a company can do in a place. The work here is cloud computing. A hyperscaler is a company that runs cloud computing at very large scale. Google Cloud is one.

Think of a bakery that plans to double the bread it sells. It could open a second shop.

It could buy faster ovens for the shop it has. It could rent oven time from the baker down the street.

Every path doubles the bread. Each one means something very different for the landlord and the power company, so you want to know which path the bakery takes.

Google’s own words name what it will double. Its Brazil blog post says the plan covers computing, storage, and high-performance processing. Those are measures of output.

You can read them as how much work the machines will do. The count of buildings and the draw of electricity are separate numbers you have to find.

That is where the Bankability Gap starts. A pledge sits at the first step on the path from announced to built. To move along that path, the pledge has to become things a bank, a landlord, or a power company can count.

For you, if you build around Google in Brazil, the pledge cuts both ways. On one side, it is a clear signal of demand.

Google is telling the market that Brazilian companies will use far more cloud computing by 2030. On the other side, the signal arrives in a unit that differs from what a builder sells. A builder sells space and power.

The pledge counts processing. Until the two are matched, you know demand is rising, and you still need to learn what share of it lands on your site.

Reading Compute Growth in Megawatts

A megawatt is a unit of electric power. It is the number a data center developer, a power company, and a lender all watch. A rack is a tall metal shelf that holds computer servers. A data center is a building full of racks.

Here is the reading skill I want you to practice. A newcomer sees “double the capacity” and doubles everything: racks, halls, megawatts. I ask one question first. What does each new rack do compared with each old one?

López gave the reason this question matters. He said advances in hardware let Google do more with less floor space. He also said the number of racks or buildings is no longer the main measure.

The hardware news from the Summit shows why. Google’s blog post says G4 virtual machines with Nvidia RTX PRO 6000 Blackwell GPUs are now available in the São Paulo cloud region.

A GPU is a chip built for heavy math; the kind AI uses. At the 2025 Summit, Google brought its sixth-generation TPUs to Brazil, according to Flávio Franco, a Google Cloud director for Latin America. A TPU is Google’s own AI chip.

Newer AI chips pack far more work into each rack. So compute can double while floor space grows much less. The same building can hold far more output after an upgrade.

Now read the power line, because it moves on its own. I call the first reading the density read. Denser racks usually draw more electricity each, and they need stronger cooling. So a plan that needs little new floor space can still need a lot of new power.

The second reading is the efficiency read, and it is just as real. Newer chips also do more work for each unit of electricity they use. If those gains are large, megawatts could grow more slowly than compute. Franco said compute will double in both standard cloud and AI.

The public record leaves open how much new power each part will need.

So you hold both reads at once, and the pledge alone settles neither. Doubled compute could mean modest new megawatts in upgraded halls. It could also mean a large jump in power to feed denser racks.

Keep the pledge in one hand and ask for megawatts with the other. A doubling in compute is a fact about output. The megawatt figure is the fact about the grid.

Share

How the Build or Lease Choice Moves Megawatts

The structural driver is where the new compute will live. A hyperscaler has three options: upgrade existing buildings, build new ones, or lease space from a developer.

A lease is a long-term rental contract, and the developer who owns and rents out the building is the lessor.

López named two of these paths. He said doubling could mean expanding current facilities or building new ones, depending on demand. He said investment planning is always tied to demand. He gave the same answer when asked about a large new data center in Brazil by 2030.

Google did not publish an investment figure. López said the company does not usually disclose one.

The path chosen decides who waits on the decision. Picture a family that needs more room. If they add a floor to their own house, the builder they hire gets one job. If they rent the apartment next door, the landlord gets a tenant for years. If they buy furniture that folds away, nobody new gets paid for space at all.

The family has more room in every case. You can see the money flow to different people.

The same thing happens here. If Google upgrades its own halls, most new spending goes to chips and cooling gear. If it builds new sites, land sellers and builders take part. If it leases, a São Paulo lessor signs a long contract and the power company must deliver more electricity to that building.

Equipment is the second driver. López pointed to ReData, Brazil’s special tax regime for data centers. He said ReData helps lower the cost of running these facilities, especially when buying components.

Cheaper components make upgrades more attractive. When upgrades cost less, more of the doubling can happen inside buildings that already exist.

The counter-factor sits in the same place. Upgrades still need power and cooling that older halls may lack. If an older building cannot take the extra load, the cheaper route stops working there, and new space comes back into play.

The two drivers can also work together. A tax break that lowers equipment costs frees money for new power and cooling in the halls that can take it.

You should expect a mix of paths, and the public record has yet to say how the mix will split.

Hyperscaler Cable Landings as the Site Signal

For new capacity to land on your site, one thing has to go right. The site has to sit near the hyperscaler’s own subsea cable landing.

A subsea cable is a bundle of fiber that runs along the ocean floor and carries internet traffic between countries. The landing point is the beach where the cable comes ashore. Latency is the delay data feels as it travels. Long or indirect routes add delay.

López explained Google’s rule. He said Google aims to keep its data centers close to the landing points of its own subsea cables. He said relying on other connections to reach its facilities would add latency. The record shows where that rule points today.

Google’s own announcement says its Firmina cable, which Google built, lands at Praia Grande, Brazil, on the coast of São Paulo state.

Fortaleza shows you how to read this signal with care. Fortaleza is Brazil’s main hub for subsea cables. López said Google already has a cable landing there, Monet, at Praia do Futuro. He also said most of Google’s infrastructure sits in São Paulo, and that all regions are under study.

He said cable decisions are made globally and that the job of the regional team is to show demand. He added that Google is exploring alternatives for the future.

The tension axis gives you two honest reads of a city like Fortaleza. The opportunity read says a Google landing already exists there, the city is a major cable hub, and Google says every region is under study.

The concentration read says today’s infrastructure sits near the larger cluster of landings in São Paulo, so new capacity tends to follow that cluster first. Both reads come from the same public facts, and both apply to any deal of this kind.

Both reads lead you to the same early question. Where do this hyperscaler’s own cables land, and how much of its network comes ashore there? The answer sits in public view.

Hyperscalers announce their cables, with landing points, on their own sites, and their leaders talk about where data centers go. Ask the question at site screening, before you sign an option on land.

A land option is a paid right to buy a plot later. The answer costs you nothing before the option and a great deal after it.

Leave a comment

Every Capacity Pledge Needs a Megawatt Translation

The lesson carries well past Brazil. A capacity pledge is a promise about output. Every party that builds, lends, or invests around it needs a translation into its own unit.

Those units are megawatts, buildings, leases, and network links.

If you work in private capital, such as an infrastructure fund or lender, your question is what backs the loan or stake.

A developer in São Paulo pointing to Google’s doubling has shown demand. Ask for the lease tying that demand to the building and the megawatts the power company has agreed to deliver.

You can finance a signed lease and a power agreement. A pledge in compute units is where that path begins.

If you follow public markets, your question is what shows up in reported numbers.

Listed data center operators, power companies, and equipment makers report revenue and load in their own units. A doubling in compute may show up mostly as equipment sales if Google upgrades.

It may show up as new leases and new grid load if Google builds or rents. Read their guidance for which path they expect.

If you run an operator, such as a developer or a landowner, your translation comes first in the deal. Convert the pledge before you price the site. Ask how many megawatts it implies. Ask whether the hyperscaler plans to expand, build, or lease.

Ask how close the site sits to that hyperscaler’s own cable landing. Ask what the local tax regime does to the cost of equipment.

If you work for a government or a development lender trying to attract data centers, the cable question is yours too. Landing points shape where buildings go, and so does the tax regime on equipment. Both sit within your reach.

One number in the Brazil pledge is still open, and it matters to every reader above.

How many megawatts does each unit of new compute need? Google has told the market how far it plans to grow its work. When you see the first power agreement or lease tied to that doubling, you will have your answer.

Join Global Data Center Hub

Welcome to Global Data Center Hub. Join investors, operators, and innovators reading to stay ahead of the latest trends in the data center sector in developed and emerging markets globally.

Join Global Data Center Hub

On October 1, 2026, Lambda closed a $1.0 billion loan to buy GPU servers.

Moody’s rated it Baa1. Morningstar DBRS rated it A (low). Both grades sit in investment grade; the band rating agencies give to debt they judge most likely to be repaid.

If you read only the headline, you see a big loan and big demand for AI computing.

That reading holds. Lambda says the money will fund three committed customer deployments for two investment-grade customers. But the loan's structure tells you more than its size.

I want you to take away one idea. A GPU loan can borrow like safe, long-term debt when the whole loan is built around customer contracts. I use the Bankability Gap to show you why.

The Bankability Gap is the distance between capacity someone announces and capacity that gets built.

The Bankability Gap and the Contract Behind Every GPU

To get built, a project needs power, money, permits, equipment, customers, and returns. Any one of them can hold a project back. For GPU projects, money is often the hardest piece to line up.

Here is why. A GPU is the chip that trains and runs AI models. Faster chips arrive every year or two, so an older chip loses value quickly. If you lend against the chip alone, you lend against something that shrinks.

So, lenders look past the chip to the contract. When a big customer signs up to rent computing for years, it promises a stream of payments. You can lend against that promise. The chip becomes the backup.

Banks have done this for decades with power plants. A lender funds the plant because a utility has signed a contract to buy its power. This is called project finance. The contract pays the loan, and the plant stands behind it.

Lambda’s October loan follows that pattern. Lambda says it is secured by the GPU servers it pays for and by the contracted cash flows from those servers.

Contracted cash flows are the payments customers have agreed to make. Lambda says those payments come from two investment-grade customers across three deployments in multiple data centers.

Think of someone buying an apartment building with every unit already rented. A bank lends to that owner more easily, because signed leases come with the building.

The building matters. The leases pay the bank.

That is the trade I follow through this article. Tying debt to contracts opens up cheaper, longer-term money for you as a borrower. It also makes the loan exactly as strong as the contracts behind it.

How to Read a Delayed Draw on a GPU Loan

If you are new to this, you see $1.0 billion and picture Lambda with $1.0 billion in the bank. That picture is off. The October loan is a delayed-draw term loan.

A term loan has a set end date and a set repayment plan. A delayed draw means Lambda doesn't take all the money on day one. It takes the money in pieces.

Lambda says each piece is tied to a cluster going into service. A cluster is a large group of GPU servers wired together to work as one machine.

Picture building a house with a construction loan. The bank does not hand over the full amount when you sign. It pays the builder in stages. The foundation passes inspection, and the bank pays. The walls go up, and the bank pays again.

The October loan works the same way. Lenders put money in only as each cluster switches on and starts serving a customer. Until then, the money stays promised but unspent.

This changes who carries what. The lenders skip the building stage. They never fund a cluster still waiting on parts, power or wiring. That stage and its risk stay with Lambda. So, the lenders’ money sits only behind working machines with paying customers.

This structure has a track record. On March 31, 2026, CoreWeave closed an $8.5 billion delayed draw facility secured by GPU assets and a customer contract.

Moody’s rated it A3, and Morningstar DBRS rated it A (low). Lambda’s August loan used the same idea, and the October loan continues it.

This reading has limits. A delayed draw also ties the loan to the build pace. If a cluster goes live late, its draw comes late too.

Lambda’s release does not say how long it has to draw the full amount. So, treat the headline number as a promise that turns into real money only as machines go live.

Share

Fixed Rates and the Investors Who Buy GPU Debt

The second thing to look at is the interest rate. Here the two Lambda loans part ways.

The August loan was a $926 million term loan B. A term loan B is a loan sold to a wide group of investors, often funds that buy and trade loans.

It paid a floating rate of SOFR plus 3.00%. SOFR is a benchmark interest rate that moves with the market, so the August loan’s rate moves too.

Lenders paid 99.5 cents on every dollar of the loan, a small discount that boosts their return. Moody’s rated it Baa2.

The October loan pays a fixed rate of 6.78%, with interest paid twice a year. Lambda marketed it to insurance companies and fixed income investors. Lambda describes it as its first U.S. fixed-rate financing. Moody’s rated it Baa1, one step above the August loan on its scale.

Fixed and floating work like the two common kinds of home loans. An adjustable mortgage moves with market rates. A fixed mortgage locks the payment for the life of the loan. You know your bill. Your lender knows its income.

That certainty is why insurers like fixed-rate debt. An insurer has promised to pay claims and pensions many years from now. It wants steady, known payments it can match against those promises. A long, fixed-rate loan backed by investment-grade contracts fits that need.

So, the fixed rate opens the door to a new group of buyers. Michel Combes, Chief Executive Officer of Lambda, called this the third new credit market Lambda has opened in the last 18 months.

A fixed rate spares the borrower the surprise of rising rates. The insurer, in turn, gets the steady income it needs. The two sides want the same thing.

The counterfactor sits on that same rate. A fixed-rate buyer locks in for the full term.

It cares most that every payment arrives on time, right up to the last one. That makes loan length the next thing you should study.

Contract Length Against Loan Length in GPU Debt

For any GPU loan like this one to work, one thing has to go right. The customer must keep paying for as long as the loan needs paying.

Here is how to see it. The October loan is fully amortizing. Each payment covers some interest and pays down some of the loan, so the balance reaches zero at the end. The final payment is due May 30, 2033.

The August loan ends December 31, 2030. Lambda said that loan’s schedule matched the contracted cash flows and the GPUs' useful life. The October loan runs 29 months longer.

So, the October loan needs cash flowing until mid-2033. In a deal like this, those payments come from customer contracts first. If the contracts run as long as the loan, they cover it the whole way. If a contract ends sooner, the last stretch has to come from somewhere else.

The servers would need a new customer, or their own value would have to cover what is left. By then, those GPUs would be several chip generations old.

The trade gives you two readings. The opportunity: a fully amortizing loan shrinks with every payment, so only a small part remains in the final years.

Two investment-grade customers also spread the contract base across more than one name. The condition to manage: if a contract ends before the loan does, the last payments rest on servers that lose value every year.

Both readings lead you to the same early question. Do the customer contracts run at least as long as the loan takes to pay down? Lambda’s release doesn't include contract length.

The answer sits in the loan agreement and in the rating reports from Moody’s and Morningstar DBRS, where the agencies explain each grade. Ask before money goes in, while you review the loan papers.

Leave a comment

Match the Debt to the Contracts That Repay It

The lesson travels well past Lambda. GPU debt lasts only as long as the contracts behind it. When money arrives in step with working machines, and payments run as long as the contracts, fast-aging chips can carry slow, safe debt. Use the Bankability Gap to see where contracts close the money side of the gap, and where they stop.

If you invest in private credit or work at an insurer, set each contract’s end date beside the loan’s final payment. Where the dates line up, the contracts carry the loan. Where they part, the servers carry the gap.

If you follow public markets, read a GPU cloud company’s debt by its rate, its length, and its buyers. Those three facts show you how much of the company’s borrowing rests on contracts and how much rests on chips.

If you build or run data centers, know what lenders will pay for. A signed contract with a strong customer is what turns your GPU build into investment-grade debt. The longer and firmer your contract, the cheaper and longer your money.

If you work for a government or a development lender, the lesson is narrower. When you want local AI computing built, ask whether the planned sites have signed customers, because lenders fund signed customers.

Match the debt to the contracts that repay it. Then confirm the match, date by date.

Next, watch the rating reports on the October loan. They can show you whether the contracts and the loan end on the same calendar.

Join Global Data Center Hub

Welcome to Global Data Center Hub. Join investors, operators, and innovators reading to stay ahead of the latest trends in the data center sector in developed and emerging markets globally.

Join Global Data Center Hub

Six Samsung companies committed $1 billion to Helix Digital Infrastructure in late September 2026.

KKR formed Helix Digital Infrastructure in June to build data centers, power plants, power lines and fiber for the biggest cloud companies.

It started with more than $10 billion from KKR, the Kuwait Investment Authority, NVIDIA and Vistra.

If you read only the headline, you see a bet on growth. That reading holds as far as it goes. But four of the six Samsung companies also make things data centers need.

So the money does two jobs. It funds the company, and it brings a supplier into the partnership.

Here is the one idea I want you to take away. A strategic investor can get paid in two ways, and the strongest deals plan for both from the start. I use the Ownership Ladder to show you where Samsung stood before this deal and where it stands now. It is a six-step map of how much of an AI build each participant owns.

The Ownership Ladder and the Supplier Who Invests

The Ownership Ladder sorts everyone in AI infrastructure by how much of the build they own. Each step up brings more control. Each step up also raises the stakes. Every rung is a valid choice with its own logic.

The bottom rung is the consumer. A consumer buys AI services and owns nothing in the building. The next rung is the demand host, the land and the country where the build lands.

Then comes the infrastructure participant. This is the company that sells into the build: chips, cooling gear, batteries, construction.

Above that is the operating partner, who runs the site day to day.

Then the capital owner, who puts in money and shares in the returns. At the top sits the platform or capability owner, who owns the system or skill other people build on.

Before this deal, Samsung Electronics, Samsung C&T, Samsung SDS and Samsung SDI sat mostly on the infrastructure participant rung. They sold to builders.

Samsung SDS also designs, builds and operates data centers, so it already knows the operating partner rung.

After this deal, all six Samsung companies share in Helix Digital Infrastructure's returns. KKR’s release says so plainly. That puts them on the capital owner rung. The four that build still build. Samsung now stands on two rungs at once.

Picture a builder who also helps pay for your new house. The builder now cares how the whole house turns out, beyond the walls it puts up. It knows the materials. It can get crews on site sooner. And because it owns part of the result, a bad job hurts its own investment.

That is the trade I want you to follow. Supplier capital brings parts, skills, and a partner who wants the project built. It also needs clear rules, so both ways of getting paid point in the same direction.

The rest of this piece shows you how to read both sides.

Share

How to Read a First-Look Right in Fine Print

Most readers stop at the headline. The headline here says $1 billion. If you keep reading, you find a key part of the design at the bottom of KKR’s release, under a heading called Notice to Readers.

That notice says the Kuwait Investment Authority, NVIDIA, Samsung, and Vistra are investors in Helix Digital Infrastructure and will share in its returns. It says these and other investors will act as strategic partners. And it says they may have certain rights, such as priority or first-look rights, to supply goods or services to Helix Digital Infrastructure projects.

A first-look right means you get the first chance. Picture selling your car. Before you post an ad, you let one friend make an offer. Your friend sees the deal first. A first look in general is a chance to bid early. Your friend can still lose the car to a better offer.

A new reader skips this notice. It looks like legal filler. I go to it early, and I suggest you do too, because it shows how the company is built. Each strategic investor brings something hard to get.

KKR’s June launch release named Vistra as Helix Digital Infrastructure's preferred power provider. It named NVIDIA as a strategic partner for its AI factory designs. Samsung is the third supplier to invest, adding cooling, construction, batteries, and chips.

Other platforms use the same design. In March 2025, NVIDIA joined the AI Infrastructure Partnership as a member and technical advisor.

BlackRock, Global Infrastructure Partners, Microsoft and MGX formed that partnership. GE Vernova agreed to help it plan its supply chain. So the pattern is spreading across AI infrastructure platforms.

KKR put the investor rights in writing from the start, for every strategic investor. So you can see the design before you judge it. There is a limit to what you can learn here.

The notice says “may.” It does not say what rights Samsung holds or what they cover. Your job is to read the notice, see the design, and then ask how the rights work in practice.

Two Ways a Supplier Investor Gets Paid

Here is the driver behind the whole structure. When you look at any investor, ask how it gets paid. There are two ways.

The first way is the investor return. That is the investor’s share of the profits, or of the company’s value when it grows. Every investor in Helix Digital Infrastructure earns it if the company does well.

The second way is the supplier margin. That is the profit a supplier makes on what it sells. Only an investor that also supplies can earn it.

KKR’s release lists what each Samsung company brings. Samsung Electronics makes chips, including memory.

Through FläktGroup, which it bought in 2025, Samsung Electronics also sells data center cooling, from air systems to coolant distribution units for liquid cooling.

Samsung C&T builds data centers and power plants as an engineering, procurement and construction contractor. Samsung SDS designs, builds and runs data centers and rents out GPU computing. Samsung SDI makes uninterruptible power supplies and battery backup units.

KKR says Helix Digital Infrastructure expects to leverage Samsung’s expertise in advanced technology, construction, energy storage, and cooling.

Two of the six play a different part. Samsung Life Insurance and Samsung Fire & Marine Insurance are insurers. The releases list nothing they supply. So within one $1 billion commitment, two Samsung companies earn only the investor return. Four can earn both.

The two often help each other. A supplier that owns part of the company wants its gear to work, arrive on time, and keep the sites running. If the projects do well, its investor return grows too. That shared interest is a big part of why strategic capital exists.

To know which matters more, you need two numbers. One is what share of returns the supplier holds. The other is how much it sells. Neither number is public. Samsung Electronics put in $500 million. The other five shared the rest, and the releases don't disclose the split.

The Price Test Behind Every First-Look Right

The one signal you should watch is a set of terms the public will not see. The releases mention first-look rights. They do not give the terms, and fund terms usually stay private. Those terms decide which goods the rights cover, how a price gets tested, and who decides when an investor bids on a job.

Every platform with supplier investors has to get one thing right. The price a supplier investor charges should hold up against what an outside seller would charge. When it does, the investor return and the supplier margin both work for the project. Good terms are what make that hold.

The trade gives two reads. The opportunity read: cooling gear, backup power, and builders can be hard to get. A supplier that is already committed can help a project start sooner, and a late project earns nothing while it waits. The read to manage: a first look works best when the price is still being tested, so speed comes at a fair cost.

Both reads lead to the same early question for you. When an investor bids on a job, how is the price compared, and who signs off? You will find the answer in the fund papers and side letters that set out the rights. You will also find it in any rule that has an investor step out of the vote on its own contract. Ask before money goes in, while every party can still shape the rules.

Leave a comment

Count Every Way Your Investor Gets Paid

You can use this lesson well beyond Samsung. Before you judge an investor, count every way it gets paid. Investor returns and supplier margins can pull together or apart, and the terms decide which. Use the Ownership Ladder to see when one company stands on two rungs.

For private capital, the lesson is about who sits next to you. A fund that invests beside supplier investors gains partners who bring parts and know-how. Ask how prices get compared and how votes work, so you can count that value with confidence.

If you follow public markets, the lesson is about where the payoff shows up. When a listed company like Samsung Electronics invests in a platform it may supply, part of its payoff may arrive later as sales. If you read only the investment line, you see half of that story.

For operators, the lesson is about what supplier money brings. It can bring gear and a partner who wants the project built. It also brings a partner with a second reason to be in the room. Plan for both before you take the money.

If you work for a government or a development lender, the lesson is about what arrives with the money. A platform like this can bring its equipment makers and builders along with its capital. You can see both sides of the partnership before you choose a site.

So count the investor return and the supplier margin, and plan for both. The first named Helix Digital Infrastructure project will be the first public view of how this partnership works on the ground.

Share Global Data Center Hub

Welcome to Global Data Center Hub. Join investors, operators, and innovators reading to stay ahead of the latest trends in the data center sector in developed and emerging markets globally.

Join Global Data Center Hub

Edition No. 71

In this week’s issue:

First, Akamai’s warrant. Akamai signed an $11.6 billion, seven-year compute deal with Anthropic and gave it a path to ~5% of its stock. Why would a supplier hand equity to a customer?

Then, a wind-fed campus finds its tenant. Crusoe’s 1GW-plus Texas campus now has its hyperscale tenant, and the power comes straight from a wind farm.

Finally, Brazil doubles down. It took nine years to build what Google Cloud now plans to double in four, by 2030.

Let’s get started.

The $11.6 billion contract and the ~5% warrant

Akamai gave a customer equity because the price was small next to the prize. Chief executive Tom Leighton put the warrant’s value at a little over $150 million, against an $11.6 billion contract.

Akamai announced the seven-year agreement with Anthropic on 24 September. It supplies compute for the AI lab’s CPU workloads and could grow by up to $9 billion. Akamai shares rose as much as 20% in after-hours trading, according to Investing.com.

Akamai keeps the hardware and the operating risk. The customer gets compute and a growing claim on its supplier. A services contract now carries an ownership tie.

The harder question is timing, because the cash goes out long before it comes back. The structure behind the deal is in The Mechanism below.

THIS WEEK BY REGION

North America: power and local consent now carry a price

Amazon will put $1 billion over five years into US communities that host its data centers. The program is called Built Together. It funds free community college and workforce training, energy and water upgrades for schools and homes, and flexible local grants.

Amazon says communities will decide what is most useful. That’s the cost of keeping permission to build. Local consent has become a budget line, and Amazon is paying it up front.

Bain Capital is reportedly weighing a bid for Koch’s Edged, in a sale Bloomberg reported at over $15 billion. Edged runs seven US facilities. A privately held industrial group is testing what private equity will pay for a full operating platform. The price will set a marker for every mid-sized US operator that comes to market next.

In Texas and Oklahoma, AB Energy secured about 2GW of on-site power projects for five data centers. The units run on INNIO Jenbacher gas engines. AB Energy says it already holds the equipment for the full 2GW. That’s power moving next to the load.

Developers in those states are building their own generation to avoid waiting on the grid. AB Energy is already arranging more units, so engine supply is the next bottleneck.

Asia-Pacific: approvals arrive, and power sets the terms

In Sydney, New South Wales planners approved Microsoft’s AU$1.263 billion, 96MW data center at Huntingwood. The consent requires its operating demand to be matched at all times by additional, firmed renewable supply.

The power condition now sits inside the planning consent itself. In New South Wales, contracting firmed renewables is becoming a cost of getting approved.

In Canberra, territory planners approved CDC’s AU$203.5 million Beard 2 data center. It adds 84MW across 12 data halls, next to the existing 39MW Beard 1. Building beside a running site lets CDC reuse power and fiber it already has. Together, the two Beard buildings will pass 120MW.

Microsoft also plans to double its Taiwan data centers from two to four, its Taiwan general manager Sean Pien said. He named electricity and rack space as the shortage. No investment figure was given. Taiwan’s compute demand is outpacing local capacity, and power is what’s holding it back.

Europe: grid works set the schedule

Caverion signed a €15 million substation contract for Nebius’s 70MW second site at Mäntsälä, Finland, due in autumn 2027. It’s Caverion’s third agreement with Nebius in Finland. The substation, more than the building, sets the date that 70MW goes live.

The same contractor keeps winning the grid work, which makes delivery capacity a scarce asset in the Nordics.

In Norway, 3i-backed Lefdal Mine bought a 40-acre development site in Sunnfjord for NOK 74 million. It plans a NOK 9 billion build there. The land cost under 1% of the planned build, so the money at risk sits almost entirely in construction and power.

In Dublin, Google won approval on appeal for a 72,400 sqm extension at Grange Castle. South Dublin County Council had refused it in 2024, citing grid capacity. The approval requires evidence of engagement with power purchase agreements.

For Ireland, the grid is still the gate, and approvals now come with power conditions attached. The decision can still be subject to judicial review.

South America: cheap hydropower meets hyperscale demand

Google Cloud will double its computing capacity in Brazil by 2030, matching in four years what took nine to build. It has already added Nvidia Blackwell GPUs to its São Paulo region. Brazil remains the region’s hyperscale anchor, and the second doubling is coming more than twice as fast as the first.

Araico unveiled Iguazú AI City, a reported ~$12 billion, 280MW campus at Yguazú, Paraguay, with Founders Fund as anchor investor. The power contract with state utility ANDE, reportedly at $45 per MWh, isn’t signed yet. The whole plan rests on that contract. Paraguay is trying to sell its hydropower as compute.

Share

THE MECHANISM

Capex now, revenue later, equity if it grows

Akamai built its name delivering web content from servers spread across thousands of locations. It has been turning that footprint into a cloud business, and this deal is its largest test.

The work is CPU compute, the general-purpose processing that runs around AI models. On Akamai’s investor call, finance chief Ed McGowan said CPUs generally earn more revenue per megawatt of power than GPUs.

The cash goes out first. Akamai plans ~$5.5 billion of capex: ~$1.7 billion in the fourth quarter of 2026, ~$3.1 billion in 2027 and ~$0.7 billion in 2028. The first tranche pre-buys supply chain components, including memory. Revenue comes later.

McGowan guided ~$150 million to ~$300 million for 2027, starting in the second half. He expects a ~$1.7 billion annual run rate by the end of 2028.

Akamai will pay for it from its own balance sheet. McGowan cited $4.6 billion of cash and a $1 billion revolver, and said no new financing is needed. The power sits in leased colocation space, about 95MW to 105MW, mostly in the US. Akamai signs multi-year leases of 10MW to 30MW per site. It owns the servers and rents the buildings and the power.

Then there’s the warrant. Anthropic holds a seven-year warrant for non-voting convertible preferred stock, 7.7 million shares on an as-converted basis, at $111.33 a share. About 2% of Akamai vests with today’s commitment.

Roughly 1% more vests for each extra $3 billion the customer buys. That’s the cheapest part of the deal and the most telling. The remaining ~3% only vests if the contract grows. Akamai is paying for future demand with future equity.

Akamai will commit ~$4.8 billion of capex by the end of 2027, before most of the revenue arrives, to one customer. Only Anthropic can trigger the $9 billion expansion. I’d want one number from Akamai: contracted revenue through 2028, set against the ~$4.8 billion it spends before then. I’m watching its next quarterly report for the first capex draw and the colocation leases behind it.

Leave a comment

NOTABLE TRANSACTIONS

Crusoe and Google: 1GW-plus Goodnight campus beside a Texas wind farm

Crusoe named Google as the tenant at its Goodnight campus in Armstrong County, near Amarillo. DCD reports more than 1GW across at least six buildings. What makes it notable is the power.

The campus connects directly to Serena’s Goodnight wind farms. One is running at 265.5MW, and a second of the same size is under construction. Surplus wind flows back to the grid.

Construction began in June 2025, so the news is the tenant. Naming the tenant turns a speculative build into a committed one, changing how the campus can be financed. Aterio Research puts total investment at ~$29 billion. Watch whether Crusoe raises debt against that commitment, and whether Goodnight 2 finishes on time.

Zone Frontier: 30-year ground lease with a purchase option on 4,077 Texas acres

Zone Frontier signed a ground lease on 4,077 acres in Potter County, Texas. It plans an initial 200MW campus, with room to pass 500MW. What makes this unique is the structure. Base rent starts only when construction starts in each area. The lease runs 30 years, with two 10-year extensions and an option to buy the data center site.

The land surrounds Southwestern Public Service’s 345kV Potter County substation, near two gas pipelines. At least 1,800 acres are set aside for solar and batteries, and the land includes a long-term water agreement.

The developer has locked up land, water and grid access without paying full rent up front. The first tenant signature decides whether this option becomes a campus. Until then, Zone Frontier says it will weigh co-investment equity and project finance.

If you are new to Global Data Center Hub, visit our Start Here page for more information about the newsletter.

If you found this newsletter useful, share it with a colleague.

Have a great week.

— Obinna

Share Global Data Center Hub

Welcome to Global Data Center Hub. Join investors, operators, and innovators reading to stay ahead of the latest trends in the data center sector in developed and emerging markets globally.

This article is the 16th article in the series: Ground to Grid: A Free 21-Lesson Guide to Mastering Data Center Development

Join Global Data Center Hub

Why Sequencing Is the Real Construction Skill

Building a data center is a sequencing problem. Speed in any single-phase matters far less than the order in which dozens of connected tasks happen.

Get the order right, and a delay in one task stays contained. Get it wrong, and that delay spreads into all the others.

The discipline of the sequence is what keeps the schedule honest.

Construction is where the design and procurement plans become a building that can carry load.

Construction risk is cumulative. A two-week slip in one phase is manageable. Two-week slips in four phases are a different project, because each slip pushes the start of the next phase.

Quality control during the build stops problems from surfacing at commissioning.

Commissioning is the formal testing that proves the facility works.

It is also the most expensive place to find a defect, because by then the problem sits behind finished work.

The Phases of Construction

Four phases do most of the work. They overlap in practice. Each one still depends on the one before it.

1. Civil and Site Work

Civil work prepares the ground. Crews grade the site, build drainage, lay duct banks (buried concrete channels that carry power and fiber cables), and pour foundations.

None of this is visible once the building is up. A duct bank sized too small cannot be fixed cheaply later.

2. Shell and Structure

The shell is the building itself: the steel or concrete frame, the roof, and the walls. A data center shell is built for heavy mechanical and electrical loads.

The shell also protects everything that comes after it. The National Institute of Building Sciences notes that enclosure performance cannot be fully verified until the building is enclosed.

Equipment should not be set, and systems cannot be tested reliably, until the building is closed to weather.

3. Mechanical, Electrical, and Plumbing (MEP)

MEP is the core of the facility. It includes transformers, switchgear, uninterruptible power supplies (UPS), generators, cooling plant, and the distribution that carries power and cooling to the racks.

MEP is the phase most exposed to equipment lead times. Switchgear and large transformers are built to order in a factory.

If they arrive late, installation waits, no matter how far along the rest of the site is.

A purchase order date says little. The useful signals come from the factory: materials in hand, assembly under way, factory test passed.

4. Integration and QA/QC

Integration is where separate pieces become working systems. QA/QC is how the team proves each piece is right before the next one depends on it.

The U.S. Army Corps of Engineers uses a three-phase model. Before work starts, the team confirms drawings, approved materials, test methods, and safety. When work starts, the first section installed is inspected and becomes the standard for the rest.

After that, daily checks continue until the work is complete, and defects get fixed before the next trade covers them.

The key tool is the hold point. A hold point is a step the work cannot pass without sign-off.

It belongs in the project schedule, so a missed inspection moves every date that depends on it. A hold point that lives only in a spreadsheet can be walked past.

Installed and done are different states. A pulled cable still needs terminations, test results, labels, and inspection before it can be energized. A schedule that counts it as complete is reporting progress the next team cannot use.

Why Delays Compound

Delays stack. A slip early in the job carries into every phase that follows.

A permitting delay pushes back civil work and everything behind it. A late transformer delays MEP regardless of site progress.

Crews can finish every other room and still wait on the one piece of equipment that energizes them.

A commissioning defect traced to weak QA/QC forces rework. Finished work comes out, and tests run again.

Each slip may start on its own. On the schedule, they line up end to end.

The cost is well measured. A Construction Industry Institute (CII) study of 359 projects found direct rework often reaches about 5% of construction cost.

The Get It Right Initiative in the U.K. puts total error costs at up to 21% once delays and disputes are counted.

How the Best Builders Stay Predictable

The first practice is overlapping design and construction, carefully. Starting construction before design is finished can save months.

It works when each package of work is released with approved design behind it. It fails when crews build from drawings that are still changing and the field ends up making design decisions.

The second is standardized layouts and prefabricated components. A repeated layout means crews build the same thing many times, with less variance each time.

Prefabricated modules move work into factories, where it runs alongside site work.

The risk moves with it. It concentrates at design freeze, transport, and the field connections between modules. So, factory work needs the same inspections and test records as field work.

The third is QA/QC at the point of installation. This is the three-phase model and hold points, applied every day.

The turnover records get built while the work is done. Records assembled at the end tend to have gaps, and missing evidence is hard to recreate.

The fourth is early procurement, so on-site work never waits on equipment. Uptime Institute adds a rule for the end of the job: put all commissioning time in the schedule and reserve time for integrated testing, so construction delays cannot consume it.

A Real-World Example

The operators who outperform do not rely on speed alone. They create schedules they can actually meet.

They procure equipment early, reduce design variance, and treat quality control as a discipline. Predictable delivery is the real advantage.

Berlin Brandenburg Airport shows what the opposite looks like.

It was set to open in October 2011 on a budget of about €2 billion. It opened in late 2020 at a cost above €7 billion.

The failure centered on the fire protection and smoke extraction system.

Finishes went up while life-safety systems were incomplete and noncompliant.

Inspectors then found defects behind finished ceilings and walls. Fixing them meant tearing out completed work.

Berlin is an airport. Data centers depend on the same kind of concealed MEP and life-safety systems.

Strategic Implications

For developers: Sequence for resilience as well as speed. The schedule that survives is the one where a single slip does not cascade.

In practice, that means finishing complete systems.

A full power train in one data hall can be tested. Rough-in spread across three floors cannot.

For investors: A developer’s track record on predictable delivery is worth more than its fastest-ever build.

Ask how much reported progress is installed but not yet inspected, tested, and accepted.

For operators: The QA/QC discipline during construction decides how many defects you inherit at handover.

If the integrated test window is cut to save a date, construction risk moves into your first year of operations.

For policymakers: Predictable permitting lets developers build predictable schedules.

Unpredictable review is where construction risk starts, because it pushes civil work and every phase behind it.

Key Takeaway

The best builders are the most predictable ones. Their average project tells you more than their fastest one.

Speed on a single project can come from luck. Predictability across many projects comes from discipline: a sequence that absorbs a slip, quality checked at installation, and equipment ordered before it is needed.

That matters because people are waiting on the date. A tenant plans its deployment around handover.

A lender sizes its loan around the day revenue starts. When the date moves, both plans move with it.

In a business where a missed date loses a tenant, a schedule you can meet beats a schedule that looks fast on paper.

Would you rather work with a developer who is fastest on their best project or most consistent across all of them? Why?

Join Global Data Center Hub

Welcome to Global Data Center Hub. Join investors, operators, and innovators reading to stay ahead of the latest trends in the data center sector in developed and emerging markets globally.

Join Global Data Center Hub

Every month, a power bill arrives with a total most people pay without reading the lines.

Part of that total now covers power set aside for data centers that exist only on paper.

On December 17, 2025, PJM Interconnection published the results of its yearly power auction.

PJM runs the electric grid for more than 67 million people across 13 states and Washington, D.C.

Each year it pays power plants to stay ready to run three years later. That promise is called capacity.

The price hit the highest level the rules allowed, $333.44 per megawatt-day, according to PJM.

A megawatt-day is one megawatt held ready for one day, and a megawatt can run several hundred homes. Even at that price, the grid came up 6,623 megawatts short of its safety target, the first shortfall across the whole PJM grid.

The auction cost $16.4 billion. On January 5, 2026, Monitoring Analytics, PJM’s independent market watchdog, traced $6.5 billion to data centers, including $6.2 billion for facilities not yet built.

Those buildings existed only as forecasts. Homes and businesses across the region still help pay for them through their utilities.

This keeps happening because asking for power is fast and building it is slow.

A company can request a large block of electricity before it pours a foundation. A power plant takes years to build.

Grids must plan for the demand they are told to expect, because running short risks blackouts. So, every inflated forecast becomes a real purchase, paid by everyone connected to the wire.

How a Forecast Turns Into a Price Tag

Picture a caterer who cooks one meal per RSVP. If one guest RSVPs to five dinners the same night, five meals are prepared, but only one is eaten. Someone pays for the other four.

A capacity auction works the same way. Each utility tells PJM how much power its customers will need in three years. Those forecasts include every data center that has asked to connect.

PJM adds them up and buys enough standby power to cover the total, plus a safety margin. When the forecast grows faster than the supply of power plants, the price climbs.

The climb was steep. PJM’s auction in July 2024 set a price of $269.92 per megawatt-day for June 2025 through May 2026.

For the delivery year before, the price had been $28.92, according to PJM’s auction records. That is more than nine times higher in a single year.

The July 2025 auction then reached $329.17, the top of a temporary price cap. The December 2025 auction hit the cap again.

The RSVP problem is real. In testimony to Ohio regulators, AEP Ohio, the utility serving central Ohio, reported inquiries from more than 50 customers at more than 90 sites, totaling more than 30,000 megawatts.

AEP Ohio reported on February 13, 2026, that peak demand across all its customers runs between about 8,000 and 10,500 megawatts.

The requests added up to roughly three times everything the utility delivers on its busiest day.

Fifty customers at ninety sites means many companies asked about more than one location.

A company hunting for the fastest connection can ask about several sites and plan to build on one of them. Each of those requests still shows up in the forecast.

What Ohio Found When Requests Came With a Price

Most people assume their power bill pays for power someone actually used. That assumption made sense for decades.

Demand grew slowly, forecasts were close to reality, and the power a grid reserved was power it soon needed.

The picture has changed. A bill in a capacity market also pays for power reserved against a forecast, and a forecast can include projects that will never be built.

The auction prices were widely reported. The split between built and forecast data centers was reported far less often.

Ohio tested how much of a forecast was real. On July 9, 2025, the Public Utilities Commission of Ohio approved a new rule for AEP Ohio’s largest data center customers.

Data centers above 25 megawatts must pay for at least 85% of the power they reserve each month, whether they use it or not, for up to 12 years. They also face exit fees if they walk away.

The requests shrank fast. Data Center Dynamics reported on October 1, 2025, that AEP Ohio’s pipeline fell from more than 30 gigawatts to about 13 gigawatts after the rule took effect.

A gigawatt is 1,000 megawatts. AEP Ohio said the commitments pushed data center companies to estimate their power needs accurately.

Data center growth in Ohio continued under the rule. AEP Ohio reported on February 13, 2026, that data centers had signed contracts for 17,861 megawatts, including 5,642 megawatts signed under the new terms.

More than half of the requests disappeared once asking for power came with a price.

The rest of the pipeline is what a forecast looks like when someone has to pay for it.

Why Fixing the Numbers Takes Years

The fix is slow because every part of the system runs on long clocks.

Forecasts get corrected after the money is spent. On January 14, 2026, PJM published a new forecast that cut its expected summer peak for 2027 by about 4 gigawatts and for 2028 by 4.4 gigawatts, according to Utility Dive.

PJM named stricter vetting of large new customers as one reason. The December 2025 auction had run four weeks earlier, on the older numbers.

New supply is slower still. PJM reported that the December 2025 auction cleared 774 megawatts of new power plants and plant upgrades.

The grid needed 6,623 megawatts more than it bought.

A gas plant, a battery site, or a transmission line each needs permits, equipment, and construction crews, and that work takes years.

Rules change one state at a time. Ohio’s rule came out of a case before state regulators and applies to one utility.

PJM serves 13 states and Washington, D.C., and each state regulator decides for its own utilities how data centers should pay.

The Households Paying for Buildings on Paper

The cost lands first on anyone who pays a power bill in the PJM region.

Utilities buy capacity for their customers and pass the cost through to monthly bills.

If you live anywhere from New Jersey to Illinois, part of your bill funds standby power for the grid’s forecast, including the data centers in that forecast.

The same bill also buys something real. Standby power is what keeps the lights on during a heat wave, and a grid that runs short risks rolling blackouts.

It also lands on people who follow AI and wonder where the costs of the boom show up.

The costs appear in chip orders and company earnings. They also appear in a household line item most people never read.

And it reaches anyone watching a data center proposal in their own county. A proposal that never gets built can still shape what the whole region pays for power three years later.

Four Clues to Spot Paper Demand in the News

The first clue is a utility whose data center requests exceed its entire system.

When requests add up to more than the utility delivers on its busiest day, many of them are likely duplicates or early bets.

These numbers appear in utility filings and testimony before state regulators, as they did in AEP Ohio’s case.

The second clue is a forecast cut after new vetting rules.

When a grid operator tightens how it screens large customers and the forecast drops, the old forecast carried projects that could not pass the screen.

PJM publishes its forecast every January, and trade outlets report the changes.

The third clue is a shrinking queue after a new payment rule.

When a state requires data centers to pay for power they reserve and requests fall sharply, the drop shows how much of the queue was never going to be built.

These rules appear in state utility commission decisions.

The fourth clue is a market watchdog splitting costs between built and forecast demand.

PJM’s market monitor publishes this split after each auction. Headlines usually carry the price. The monitor’s report carries who the price was for.

The Delivery Years Will Settle the Question

The forecast years are coming. The power bought in December 2025 covers June 2027 through May 2028, and by then the region will see how many forecast data centers turned into buildings.

If they arrive, households paid for power the grid truly needed, and the lights stayed on. If many never arrive, households paid for the caterer’s extra meals.

The open question is who should carry the cost when a forecast turns out wrong.

Ohio gave one answer: the company that asked for the power.

Every other state that shares a grid with data centers will face the same question, and the answer will show up on the bill.

Join Global Data Center Hub

Welcome to Global Data Center Hub. Join investors, operators, and innovators reading to stay ahead of the latest trends in the data center sector in developed and emerging markets globally.

Join Global Data Center Hub

BDx Data Centers announced on September 22, 2026 that it had broken ground on AI Campus 2 (CGK4), a 640MW AI data center campus in Jatiluhur, Purwakarta, West Java. BDx is backed by I Squared Capital. The campus has six planned buildings, and construction has started on the first.

BDx’s 640MW Plan at Jatiluhur

Building 1 will deliver 120MW of IT capacity in phases. BDx said the first contracted phase is expected to enter service in early 2027. The full campus is expected to take about three years, with buildings commissioned one after another.

The site covers 111,000 square meters of land and about 180,000 square meters of floor area, according to BDx. It sits about five kilometers from the Jatiluhur Dam. BDx said CGK4 uses direct-to-chip liquid cooling built to support up to 500kW per rack. The campus received NVIDIA DGX-Ready Colocation certification in April 2025.

In July 2024, BDx announced the completion of Phase 1 at CGK4. It then described the site as a campus “scalable up to 500MW,” with racks of up to 120kW. DCD reported that first phase as 70MW.

West Java Governor Dedi Mulyadi and Purwakarta Regent Saepul Bahri Binzein attended the groundbreaking with BDx and BDx Indonesia leadership. BDx Indonesia is the joint venture BDx formed in 2022 with Indosat Ooredoo Hutchison and Lintasarta.

PLN Power and the US$320 Million Loan

BDx said CGK4 is backed by 845MVA of grid capacity from PLN, Indonesia’s state electricity company. BDx described this as part of a secured power position of more than 1.2GVA across its campuses in Indonesia.

Antara reported that BDx and PLN signed the power supply agreement on May 19, 2026. According to Antara, the agreement includes a new 150kV substation at Jatiluhur. BDx said it is pursuing a hydroelectric power arrangement with Perum Jasa Tirta II, which runs the dam’s 187.5MW plant.

BDx linked CGK4 to a US$320 million loan facility that closed on April 1, 2026. Bank Permata, BCA and KB Bank led it. BDx listed three uses: developing its CGK3 campus in Jakarta, refinancing existing debt, and raising grid capacity at CGK4 and at CGK5 in Suryacipta, Karawang.

Hyperscaler Interest and the First Contracted Phase

BDx said CGK4 is attracting strong interest from multiple hyperscalers, global technology companies, AI-native cloud providers and enterprises. The first phase of Building 1 is already contracted. BDx did not name the customer or disclose how many megawatts that phase holds.

The first contracted phase is scheduled to enter service in early 2027.

The First 120MW Draws on a Grid Deal Already Signed

BDx is starting with 120MW at Jatiluhur, 19 percent of the 640MW campus, on grid power it has already contracted with PLN. Coverage reported the 845MVA as a supporting detail and led on the 640MW headline. Set against PLN’s own figures, the power is the bigger story, because it lets BDx grow the campus building by building as demand arrives.

BDx Holds a Large Share of Indonesia’s Data Center Power

The strongest fact in this deal is the power. PLN says it has 2,038MVA connected to all 159 of its data center customers today. BDx’s contracted 845MVA for one campus equals about 41 percent of that. PLN says about 99 percent of that load sits in the Java-Madura-Bali grid.

That is a large power position in the part of Indonesia where demand already is. BDx describes CGK4 as part of more than 1.2GVA it has secured across its Indonesian campuses.

The power arrives in steps. Antara reported the supply rising through pairs of 150MVA, 265MVA and 380MVA connections, behind a new substation at Jatiluhur. That staging fits a campus designed to open one building at a time. Over the next 12 to 24 months, each substation step will set how fast the next building can follow Building 1.

Indonesian Demand Is Running Ahead of Supply

The market behind this campus is growing fast. In August, Danantara’s chief technology officer spoke about new data center capacity in Indonesia. “If we build 100 megawatts, 200 megawatts, it is immediately sold out,” he said. He put customer waits at six months to a year.

The supply numbers point the same way. Cushman & Wakefield counted 322MW operational in Jakarta at the end of 2025, with 186MW under construction and 901MW planned. The planned pipeline alone is almost three times the capacity running today.

That gap is the opportunity. Capacity that can switch on in 2027 and 2028 will meet buyers who are already waiting. Campuses with contracted power are best placed to capture it.

Share

The build-out around Jakarta is moving to large campuses with their own power. Princeton Digital Group is putting USD 1 billion into a 120MW campus near Jakarta, with its first phase due in the fourth quarter of 2026. BDx is building a 640MW campus at Jatiluhur on the same kind of model.

CGK4 itself shows the scale-up. BDx called it a 500MW campus in 2024 and again in April 2025. After the PLN agreement in May 2026, the plan rose to 640MW. Rack design moved from 120kW to 500kW per rack, which BDx says is built for current and future AI chip generations.

BDx is also building more than one campus. The US$320 million loan from Bank Permata, BCA and KB Bank funds work at CGK3 in Jakarta and grid capacity at CGK4 and CGK5. Lenders are already backing the power build-out across the platform.

The local partners add reach. BDx Indonesia is a joint venture with Indosat Ooredoo Hutchison and Lintasarta, a telecom operator and an IT services group. The campus also holds NVIDIA DGX-Ready certification, a signal to buyers of large AI systems.

The next milestone is construction finance for Building 1. When BDx announces that financing, it will show how quickly the rest of CGK4 can follow.

What Capacity Buyers in West Java Should Check First

This read is for Operators: the hyperscalers, AI cloud companies and developers weighing capacity in West Java. The market now offers several large campuses outside central Jakarta, each on its own grid timeline. Your question for every campus is the same: how much power you can use, and when.

First, check when the power is energized. PLN agreements in West Java arrive in steps, behind substations still being built. CGK4’s reported figure moved from 788MVA in June to 845MVA in September, so confirm the contracted number in the agreement itself. This belongs in site selection, before you reserve space.

Second, line up the delivery dates. Princeton Digital Group targets its first phase for the fourth quarter of 2026. BDx targets its first contracted phase at Jatiluhur for early 2027. The evidence is each developer’s own schedule. This belongs in shortlisting, before term sheets.

Third, match the design to your chips. CGK4 is built for direct-to-chip liquid cooling at up to 500kW per rack and holds NVIDIA DGX-Ready certification. Ask every campus on your list for the same two facts. This belongs in technical due diligence.

Fourth, check how each building is funded. Grid money and construction money are separate. The evidence is the developer’s sources and uses. This belongs in counterparty review, before you sign a lease.

Skip these checks and you risk a queue that Danantara’s chief technology officer puts at six months to a year.

Early 2027 Puts Indonesian AI Demand on the Record

The first half of 2027 will show how fast this market is filling. Princeton Digital Group’s first phase is due in the fourth quarter of 2026. BDx’s first contracted phase at Jatiluhur follows in early 2027.

If those phases fill as quickly as Danantara describes, West Java moves from a promising market to a proven AI hub. CGK4, with its power already signed, is placed to grow with it one building at a time.

The open question is pace. How fast will the rest of Building 1, and the five buildings after it, find their customers?

Share Global Data Center Hub

Welcome to Global Data Center Hub. Join investors, operators, and innovators reading to stay ahead of the latest trends in the data center sector in developed and emerging markets globally.

Join Global Data Center Hub

Anthropic and OpenAI are exploring data center deals of 20 to 30 MW each, CNBC reported on September 18, 2026, citing people familiar with the talks.

No counterparty has been named, and no party has confirmed the terms.

The talks cover deployments in the United Kingdom, the Nordics and the United States, according to CNBC.

CNBC reported no price, capital figure, delivery date or signed contract.

Anthropic and OpenAI Small Site Talks

Four people told CNBC that Anthropic had explored deals of that size in the UK and the Nordics.

Two said OpenAI had looked at similar sites in the Nordics. One described talk in the US involving both labs.

CNBC did not say whether 20 to 30 MW means total facility power or the power available to computing equipment.

OpenAI told CNBC that it is building a diversified compute portfolio. It said it assesses sites on requirements, performance, reliability, timing and cost. It declined to discuss specific talks.

Jabez Tan, head of research at Structure Research, told CNBC that a few megawatts at a site with power already connected can be more practical than waiting for a larger block in one place.

Anthropic and the Akamai Contract

On September 24, Akamai Technologies announced an $11.6 billion, seven-year contract with Anthropic.

The contract covers work on central processing units, the general-purpose chips that also support AI services, across Akamai’s distributed cloud network.

It can expand by up to $9 billion, to about $20 billion in total.

Akamai issued Anthropic a warrant for up to about 5 percent of its common stock.

About 2 percent is expected to vest with this commitment. Akamai expects to spend about $5.5 billion to build the capacity, TechCrunch reported.

The Large Sites OpenAI and Anthropic Already Hold

In April 2026, Anthropic committed more than $100 billion over ten years to Amazon Web Services, for up to 5 GW of capacity.

In August, Anthropic signed a cloud contract with Nscale valued at about $45 billion, covering about 460 MW at a West Virginia campus, Blockspace reported.

Anthropic also signed a 20-year lease with TeraWulf for about 401 MW of computing load in Kentucky, according to Blockspace.

OpenAI passed the original 10 GW target for its Stargate program in April and has since added 3 GW in Georgia and 8 GW in Ohio, Quartz reported.

The Ohio leases cover about 8 GW of computing load, which would need about 10 GW of total power, according to Blockspace.

The first systems under the Nscale contract are scheduled for late 2027, Blockspace reported.

U.S. Hubs Have No 5 MW Block Left This Year

CBRE says no contiguous block of 5 MW, or more will be available in North America’s primary data center markets during 2026, and the labs are reportedly shopping for 20 to 30 MW.

The September coverage of the CNBC report described small sites as the faster route to capacity.

None of it set that route against CBRE’s count.

Read the talks as a speed move, and you miss where the small blocks have to come from.

Small Sites Will Come From Outside the Big Hubs

The big U.S. hubs are full. Vacancy in North America’s primary markets fell to a record low of 1.4 percent in the first half of 2026, according to CBRE.

The next large contiguous projects in those markets are not expected until late 2027 at the earliest.

So, a US deal at 20 to 30 MW this year has two places to land. It can sit in space a tenant already holds and gives up. Or it can sit outside the primary markets.

CBRE already points to the second. It expects demand in emerging markets to shift toward 25 to 75 MW sites serving inference and enterprise AI.

Inference is the work of answering user requests once a model is trained. NVIDIA is building toward the same size band.

With EPRI, Prologis and InfraPartners, it plans about 25 small data centers of 5 to 20 MW next to utility substations, across five US utilities, with pilot construction by the end of 2026.

Over the next 12 to 24 months, the first signed site will show which path the labs take.

If it lands in a secondary US market, the small-site map moves away from the hubs for good.

Britain’s Queue Grew Where Sites Are Largest

In the UK, the jam sits in the big connections. Contracted demand in Britain’s connection queue rose from 41 GW to 125 GW between November 2024 and June 2025, Ofgem reported.

Transmission offers, the high-voltage connections large sites need, grew from 17 GW to 97 GW. Distribution offers grew from 24 GW to 29 GW.

Data centers account for at least 80 GW of the total, according to Ofgem.

The transmission queue holds 315 data center projects asking for about 73 GW.

That is about 230 MW per project, roughly ten times the size the labs are reportedly seeking. Britain’s peak demand in 2025 was about 45 GW.

Ofgem proposed a fee on July 29 for large data center developments, set between £237,500 and £712,500 per MW.

The developer gets it back at energization and loses it if the project leaves the queue early. The published summaries do not state the size at which the fee starts.

That threshold decides the UK economics of a 20 to 30 MW site. Watch the final rule over the next year.

It will show whether small sites in Britain pay the same entry price as the 230 MW average.

Large Projects Break Before They Energize

The large sites break in the years between signing and switching on.

Local opposition blocked or delayed 45 US data center projects worth about $68 billion between April and June 2026, according to Data Center Watch. It counts 843 opposition groups across 49 states.

Equipment adds its own clock. Power transformers averaged 128 weeks for delivery in the US, according to a Wood Mackenzie survey from the second quarter of 2025.

JLL puts the average wait for a grid connection in primary markets at more than four years.

Anthropic’s Nscale capacity shows the same clock inside the labs’ own book. It was signed in August and delivers its first systems in late 2027.

No report ties the small site talks to a delay at any of the labs’ own campuses. The public record shows the gap. It does not show the cause.

Operators With Live Power Get the First Call

This read is for Operators: developers, owners of sites with power already connected, and landowners holding a grid position. The labs are asking for a size you may already have.

First, state what your megawatts measure. OpenAI’s Ohio leases show the gap: about 8 GW of computing load needs about 10 GW of total power. The evidence is your connection agreement and electrical design.

Settle it before you market the site, because the buyer will size the deal on the smaller number.

Second, prove the power is live. The evidence is an energization record or a signed connection offer. In Britain, add your queue position and any exposure to Ofgem’s proposed fee. This belongs in the first buyer screening.

Third, price what the buyer asks for beyond rent. Akamai gave Anthropic a warrant for up to about 5 percent of its stock alongside an $11.6 billion contract.

The filing is the first public record of that trade in this market. Take it into term sheet talks and decide your answer before the question comes.

Skip these, and the cost is concrete in Britain. Under Ofgem’s proposal, a large project that leaves the queue early loses £237,500 to £712,500 for every megawatt it holds.

The Small-Site Window Runs to Late 2027

The big US blocks return in late 2027, on CBRE’s timeline.

The Nscale capacity arrives in late 2027. Until then, sites with live power set the terms.

Rents already rose across every deployment size CBRE tracks in the first half of 2026, with 3 to 10 MW deployments up 8.3 percent.

That makes the next 15 months the window for anyone holding power. After late 2027, the large blocks come back and compete again.

The open question is what the labs are buying. Nobody has said whether 20 to 30 MW means total power or computing load, or whether any of these talks will sign.

The first signed deal will answer both, and it will set the price for every powered site behind it.

Join Global Data Center Hub

Welcome to Global Data Center Hub. Join investors, operators, and innovators reading to stay ahead of the latest trends in the data center sector in developed and emerging markets globally.

Join Global Data Center Hub

Google plans to send four AI chips into orbit on Thursday, October 1.

The New York Times reported that the test satellite, called MVP, will ride a SpaceX rocket from California.

Google confirmed the launch on its own blog and said the satellite was built with Planet, a company that makes Earth-watching satellites.

Most of the talk about this launch is about rockets. How cheap can a launch get? When will a data center in space cost the same as one on the ground? Those are fair questions. But they skip a simpler one that tells you more.

Every data center is held back by one thing before anything else. On the ground, that thing is usually power. In space, power is easy to get, so something else takes its place.

This piece teaches you how to find that one thing and how to use it to read any big infrastructure claim.

It does not cover the laser links that would join satellites together, and it does not say whether any company in this race is a good investment.

Why Data Centers on the Ground Hit Power First

Every system has one input that runs out first. That input sets the size of everything else.

Think of a restaurant with ten cooks and one oven. The cooks can chop all day. But the kitchen can only serve as many meals as the oven can bake. Hire more cooks and nothing changes. Buy a second oven and the kitchen grows. The oven is what runs out first.

Factory managers have used this idea for decades. The lesson is simple. Find the part that runs out first. Fix that part. Then look again, because a new part will now run out first.

A data center works the same way. It needs land, chips, buildings, water, workers, and electricity. On the ground, electricity is usually the part that runs out first.

Here is why. A new data center has to connect to the power grid, the network of wires and power plants that carries electricity. New power plants and batteries also have to connect before they can send power to anyone. They wait in a line called the interconnection queue.

At the end of 2025, more than 2,060 gigawatts of these projects were waiting in that line in the United States, according to Lawrence Berkeley National Laboratory. A gigawatt is a billion watts, enough to run a very large data center campus. The typical project built in 2025 waited more than five years to get through.

So on the ground, you can have the land, the money, and the chips, and still wait years for power. Power is the oven.

Google’s Project Suncatcher starts from that fact. Its idea is to move the computers to where power is easy to get. In a special orbit that follows the line between day and night, called a dawn-dusk orbit, a satellite’s solar panels face the sun almost all the time.

Google’s research paper says a panel there can collect up to eight times more solar energy per year than the same panel on the ground at a middle latitude, such as the United States or Europe.

That is the opportunity. It is real. But it comes with a risk that follows from the idea itself. When you fix the thing that runs out first, something else takes its place. The question for Suncatcher is what that new limit is.

How to Read the Space Data Center Cost Claim

The most repeated claim about Suncatcher is about cost. Headlines say space data centers could match the cost of ground data centers by the mid-2030s. To see what that claim means, you have to look at what is being compared.

Google’s research paper does the math this way. Today, launching one kilogram to low Earth orbit costs about $3,600 on a reusable Falcon 9 rocket, the figure the paper uses. The paper projects that the price could fall to about $200 per kilogram by the mid-2030s.

At that price, the paper spreads the cost of launching a satellite over the years it works. It turns that into a cost for each kilowatt of power the satellite provides, for each year. A kilowatt is a thousand watts, about what a microwave oven uses.

For a satellite built like SpaceX’s Starlink satellites, the paper puts that cost at about $810 per kilowatt per year.

Then it compares that number to what ground data centers in the United States spend on electricity. The paper puts that range at $570 to $3,000 per kilowatt per year. Since $810 falls inside that range, the paper calls the two costs roughly comparable.

A new reader sees this and thinks space data centers will cost the same as ground data centers. That is the surface reading.

A careful reader asks a different question. What is on each side of the comparison?

On one side is the cost of launching the satellite. On the other side is the electricity bill. The paper itself says the comparison leaves out the cost of the chips and the cost of buildings. It also says plainly that it “does not constitute a full economic analysis.”

So, the claim is narrower than the headlines. It says the launch bill could match the power bill. The full cost of a space data center is a separate question the paper leaves open.

The claim also depends on a lot of rockets. To bring prices down to about $200 per kilogram, the paper estimates the world would need about 1,800 more launches of SpaceX’s Starship, or about 180 a year. The whole cost story rests on one company’s rocket flying that often.

There is a fair counter-read. Electricity is one of the largest costs a data center pays, year after year. Matching it would matter. And Google states the limits of its own math in the paper, which is more than many plans do.

So the skill is this. When anyone tells you two things will cost the same, list what sits on each side. Then ask what was left out. The answer often tells you where the real limit is.

Share

Why Heat Is the First Limit in Orbit

Every watt of power a chip uses turns into heat. A chip that uses a thousand watts makes a thousand watts of heat. That heat has to go somewhere, or the chip gets too hot and stops working.

On the ground, this is a solved problem. Fans blow air across the chips. Water carries heat away. The air and water take the heat out of the building.

In space, there is no air. There is no water to spare. There is only one way to get rid of heat. The satellite has to glow it away as invisible light, the way a stovetop glows heat into a kitchen. Engineers do this with flat panels called radiators. Heat moves from the chips to the panels, and the panels send it out into space.

Radiators work, but they are slow. In a video Google released about the project, one of its engineers explained the gap in plain terms. A square meter of radiator gets rid of a few hundred watts.

A chip small enough to hold in your hand makes thousands of watts. The engineer said that means “meters and meters squared of area per TPU.” A TPU is Google’s own AI chip.

Now look at the test satellite. The New York Times reported that MVP runs on about 1 kilowatt of power and carries four chips. The Times also reported that the chips will run for about 15 minutes and then shut off to cool down. The smallest possible test already has to stop and rest.

Google’s engineers say future satellites will carry dozens of chips. In the same video, they said they are aiming for much more power than has been done before on a spacecraft.

Here is where heat connects back to cost. Rockets charge by the kilogram. More chips means more heat. More heat means more radiator area. More radiator area means more weight. And more weight means a bigger launch bill. So the heat limit feeds straight into the cost question.

There is a real counterfactor. The cooling methods are not new.

Google’s engineer said the thermal technology has plenty of history on past spacecraft. Google’s blog says the team tested its setup in a thermal vacuum chamber, a sealed room that copies the cold and airless conditions of space.

The team is also looking at heat pipes and pumped fluid loops, which carry heat out to the radiators. Better tools may raise the limit over time.

So the honest reading is that heat sets the pace. On the ground, power sets how fast data centers can grow. In orbit, the evidence so far says heat will.

The 15 Minute Clue on Google’s Test Satellite

Of everything disclosed about this launch, one detail deserves more attention than it gets. It is the 15 minute work cycle.

The share of time a machine can run at full power is called its duty cycle. A car engine that can only run for 15 minutes before it must cool down has a low duty cycle. It still works. But it does less work per hour than an engine that can run all day.

This matters because of how the cost math is built. The paper’s figure of about $810 per kilowatt per year spreads the launch cost over the power a satellite provides. If the chips can only use that power part of the time, each kilogram launched does less work.

The cost of each hour of useful computing goes up. This is the point where a plan like this could break: the launch price falls as hoped, but the chips cannot run long enough to make each launch pay.

There are two fair ways to read the 15 minutes.

The opportunity read says MVP is not built for this job. According to the New York Times, a Planet engineer said Google wanted to be in orbit this year, so it put its chips into a satellite Planet had already built.

A satellite designed from the start around cooling, with larger radiators, could run much longer. By this reading, the 15 minutes reflects a rushed test more than the idea itself.

The risk read says the problem grows with size. MVP has four chips. Future satellites will have dozens. Each new chip adds thousands of watts of heat, and each watt needs radiator area. By this reading, the 15 minutes is an early look at the limit that will decide how big these satellites can get.

Both readings lead to the same early check. The question to ask is simple: in orbit, what share of the time can the chips run at full power? The evidence that answers it will come from MVP itself.

The New York Times reported the satellite is planned to operate for about a year. Its results, and the radiator size of later designs, will show whether the duty cycle rises.

The place to ask this question is at the very start, before you compare any space computing cost to a ground data center. A cost per kilowatt means little until you know how much of the time the kilowatt can be used.

Reading Heat Before Cost in Every Data Center Deal

The lesson from Suncatcher is bigger than one satellite. It is a way to read the whole sector. Find what runs out first. When someone says they fixed it, ask what runs out next.

For operators, the people who build and run data centers, what runs out first on the ground has not changed. Power is still the oven. James Manyika, who leads research at Google, told the New York Times not to expect “anything usefully operational in the next few years.”

So the work of getting grid connections, signing power deals, and planning around multi-year waits stays the same. Projects being planned today should assume their power will come from the ground.

For private capital, the funds and investors who back new companies and projects, the lesson is a question to ask of every plan. SpaceX has filed with the Federal Communications Commission for up to 1 million computer satellites.

Blue Origin has filed for up to 51,600. A filing tells you what a company hopes to launch. It does not tell you how the satellites will get rid of their heat. A plan that talks only about launch prices has named one limit and skipped the next.

For public markets, the investors who own listed shares and bonds, the lesson is about scale. Alphabet, Google’s parent company, has guided to $175 billion to $185 billion in spending this year, mostly on ground-based AI infrastructure, according to Fortune.

Planet builds the Suncatcher satellites, but the value of its contract has not been disclosed. Next to the ground spending, the space test is small. That tells you where the real limit still sits for the companies that matter most.

The principle outlasts this launch. Every system is limited by one thing first. Solving it moves the limit somewhere else. The next time a builder tells you they solved the power problem, ask them one question: what did you run out of next?

Share Global Data Center Hub