AI Data Centers
Build-to-suit campuses, powered shells and hyperscale blocks of 10–50 MW, designed around your GPU roadmap and energized by power we secure first.
Best for: Hyperscalers, AI labs and enterprises that need a dedicated hall or campus.
Solutions
PacinfraX delivers AI data center capacity four ways: campuses built to your specification, colocation for your GPU clusters, on-site power infrastructure, and hydrogen and ammonia beside our campuses. Every solution starts with power, so your capacity energizes on the construction schedule.
10–50MW
Hyperscale block size
1,200MW
Planned in Vietnam
4
Site agreements in Vietnam
3
Markets: Texas, Minnesota, Vietnam
Choose by need
Pick the line that sounds like your team. Each one leads to a page with the steps, design basis and questions for that model.

“I have servers today and need a hall to put them in”
Ship your racks into a liquid-cooling-ready hall we operate. The onboarding guide covers every step from first call to go-live.
Colocation & Hosting →“I run GPU clusters for training or inference”
High-density AI and GPU colocation with direct liquid cooling, storage and network space planned around the cluster.
GPU hosting →“I need 10–50 MW built to my design”
A campus or hall designed to your specification, energized by power we secure first, and delivered as a finished facility.
Build-to-suit →“My team fits out its own halls”
An energized building and power plant, handed to your team for fit-out on your schedule.
Powered shell →“Power is the bottleneck on my own site”
On-site natural gas generation sized to your load, so you energize without waiting on the interconnection queue.
Power Infrastructure →“I need ride-through and load smoothing for AI”
Battery storage that rides through events and smooths the sharp load steps of GPU clusters.
Battery storage →“I want to start on site power and join the grid later”
Energize on on-site power first, then add the utility connection when it arrives, with no rebuild.
Bridge-to-grid →“I buy hydrogen or ammonia, or build the technology”
A development program co-located with campus power, for industrial and marine buyers and technology partners.
Hydrogen & Ammonia →Our solutions
Build-to-suit campuses, powered shells and hyperscale blocks of 10–50 MW, designed around your GPU roadmap and energized by power we secure first.
Best for: Hyperscalers, AI labs and enterprises that need a dedicated hall or campus.
AI and GPU colocation in liquid-cooling-ready halls. You bring the hardware; we run power, cooling, security and remote hands.
Best for: Neoclouds, AI labs and enterprises with servers ready to deploy.
On-site natural gas generation, battery storage and bridge-to-grid design that energize your load on your schedule.
Best for: Campus owners and operators for whom power decides the date.
A development program co-located with campus power, for industrial and marine buyers. We are seeking technology partners and offtakers.
Best for: Industrial and marine buyers, and process technology companies.
Delivery models
The power path, block size and design standards stay the same in every model. What changes is how much of the stack you own.
| Model | PacinfraX delivers | You bring | Fits |
|---|---|---|---|
| Build-to-suit | Land, power, building, hall fit-out | Your specification and your racks | A dedicated campus or hall |
| Powered shell | Land, power, building shell | Hall fit-out and racks | Teams with their own fit-out standard |
| Hall lease | Land, power, building, hall, operations | Racks and IT | A full hall with our operations |
| Rack colocation | Everything up to the rack | Servers and network gear | Fast deployment of existing hardware |
Weighing them up? Read colocation vs build-to-suit vs powered shell.
How we deliver
Next
Share megawatts, timing, market and cooling. We match you to a campus and a delivery model.
Next
Our engineers and yours agree density, cooling, redundancy and power for each hall.
Next
Halls and on-site power rise in parallel, so capacity energizes block by block.
Next
You move in with full documentation and add the next block as your workloads grow.
Where available

An AI campus in development with on-site gas generation and storage, built for fast energization.
Texas campus →
Minnesota, powered industrial site, for AI capacity in the upper Midwest.
Minnesota site →
1,200 MW planned across four sites. Nam Thang Binh, Da Nang leads with 200 MW and a 20 MW phase 1 targeted for 2026–2027.
Vietnam sites →New to AI capacity planning? Start with what 50 MW of AI capacity looks like and liquid cooling for AI data centers, or browse all resources.
FAQ
Build-to-suit AI data centers and powered shells, AI and GPU colocation, on-site power infrastructure, and a hydrogen and ammonia program co-located with campus power.
Choose colocation if your servers are ready today, build-to-suit for a dedicated campus or hall, a powered shell if your team fits out its own halls, and power infrastructure when energization speed decides your project.
Every campus is planned with its power path first: on-site natural gas generation and battery storage, or an already-powered site. Halls energize on the construction schedule, without waiting on the interconnection queue.
Yes. A common path is colocation racks now and a dedicated build-to-suit block later on the same campus, with power infrastructure shared across both.
South ERCOT, Texas, where an AI campus is in development; Minnesota, powered industrial site; and Vietnam, with 1,200 MW planned across four sites.
Request capacity
Share your megawatts, market and timeline. We reply within two business days with a site, a power plan and a delivery model.
Power-first AI capacity
Request capacityTalk to PacinfraX
We reply within two business days. WhatsApp and email work at any hour.