Are AI Factories the Future of Hyperscale? G42’s Vietnam Bet Could Be the Tipping Point

Hyperscale once meant power and racks. Now it means inference, interconnect, and integrated AI stacks.

Jul 18, 2025

Welcome to Global Data Center Hub. Join 1600+ investors, operators, and innovators reading to stay ahead of the latest trends in the data center sector in developed and emerging markets globally.

Not All Hyperscale Is Built Equal Anymore

The rise of AI workloads is redrawing the blueprint. What used to be massive warehouses for cloud and content delivery are now evolving into vertically integrated compute platforms, designed not just to store and serve, but to train, tune, and infer.

G42’s $2B proposal in Ho Chi Minh City isn’t just another data center.

It’s an AI Factory, a purpose-built facility that could redefine what hyperscale means in the age of artificial intelligence.

The Old Hyperscale Model Is Breaking Under AI Demands

Traditional hyperscale emerged to support consumer cloud, enterprise SaaS, and video.

The design was simple: optimize for density, cost per megawatt, and availability. Success was measured in PUE, uptime, and square footage.

But AI workloads break that model.

Large language models require tight coupling of GPUs, massive memory bandwidth, low-latency fabrics, and the ability to shift from training to inference rapidly.

Traditional facilities simply weren’t designed for that.

The AI Factory: Purpose-Built for Intelligent Workloads

G42’s “AI Factory” concept marks a shift from generalized cloud toward vertically integrated AI infrastructure.

Unlike conventional data centers, AI Factories are optimized for:

AI lifecycle orchestration (from ingestion to inference)
High-density, liquid-cooled racks
On-prem or sovereign AI models
Regional training + low-latency inference zones
Stack control across hardware, models, and data

In short, they’re not just neutral space for any tenant. They’re strategic platforms that serve specific compute missions with embedded intelligence at every layer.

Why Vietnam Could Be the Perfect Test Case

Vietnam offers a unique testbed:

High demand, low supply
Cost-effective construction (~$6.9M/MW)
Strong government support for AI and digital transformation
Central location in Southeast Asia with geopolitical neutrality

The G42-led consortium, paired with Microsoft, FPT Software, VinaCapital, believes this facility could become the region’s first true AI Factory.

But it’s not just about Vietnam.

Hyperscale Design May Be About to Flip

If G42’s model works, the definition of hyperscale will shift:

From:

"Who can deliver 100MW the fastest?"

To:

"Who can deliver full-stack AI performance with data sovereignty, energy resilience, and model fine-tuning baked in?"

This has implications for:

Data center design and engineering firms
Cloud operators looking to retain AI clients
Emerging markets competing for AI investments
Governments trying to balance control and competitiveness

This Isn’t Just a Facility. It’s a Prototype for the Next Decade

This isn’t just a one-off project. It’s a prototype.

If Vietnam approves G42’s proposal, and the AI Factory model proves viable, it won’t just change how facilities are built.

It could redefine what data centers are for.

The hyperscale era isn’t ending.
It’s evolving.

Global Data Center Hub