Solutions · AI Data Centers

Build-to-suit AI data centers and hyperscale capacity

PacinfraX designs, builds and powers AI data centers to your specification. Choose a build-to-suit campus, a powered shell or hyperscale blocks designed in 10–50 MW increments. On-site generation is planned with the halls, so your capacity energizes ahead of the interconnection queue.

Updated

  • 1,200MW

    Planned in Vietnam

  • 4

    Site agreements in Vietnam

  • 10–50MW

    Hyperscale block size

  • 3

    Markets: Texas, Minnesota, Vietnam

Who it's for

Built for teams that need dedicated AI capacity

If your roadmap depends on megawatts arriving on schedule, this is where to start.

Photo: GPU racks with liquid cooling
High-density GPU cluster racks with direct liquid cooling manifolds
  • Hyperscalers and AI labs

    Dedicated campuses for training clusters, built to your design standard with the power plan settled before steel goes up.

  • Neoclouds and GPU providers

    Powered shells you fit out on your own timeline, so new GPU capacity starts earning sooner.

  • Enterprises

    We handle land, power and construction while your team focuses on models, products and customers.

Need fewer than 10 MW or a faster start in a leased hall? See AI colocation and hosting.

What you get

Everything an AI data center build needs, from one team

Photo: data hall steel frame
Steel frame of a data hall rising on a construction site at sunset

A PacinfraX engagement covers the whole path from land to a commissioned hall. You deal with one schedule and one accountable team, and you keep control of the decisions that shape your cluster.

Six things every engagement covers
  • Power first

    Every campus starts with its power plan. On-site generation and storage energize halls ahead of the interconnection queue.

  • Build-to-suit or powered shell

    Take a finished facility built to your specification, or a powered shell your own team fits out.

  • Scale in 10–50 MW blocks

    Hyperscale blocks are designed in 10–50 MW increments. Add capacity as clusters grow without re-planning the campus.

  • Liquid-cooling-ready halls

    Halls are designed for direct-to-chip liquid cooling from the first drawing, with air-cooled zones where you need them.

  • Redundancy you choose

    N+X is set per hall to match each workload, so you pay for the resilience each cluster actually needs.

  • One accountable team

    One team owns land, power, design and construction through handover.

Know your load and timing? Request capacity.

Delivery models

Build-to-suit, powered shell or hall lease: who does what

The delivery model decides who designs, who builds and who fits out. PacinfraX offers three ways to take AI data center capacity on a campus, plus rack-level colocation for smaller deployments.

Diagram: who does what per model
Diagram: who does what in build-to-suit, powered shell, hall lease and rack colocation
Responsibilities across build-to-suit, powered shell, hall lease and rack colocation.
Delivery models compared
Delivery models compared
ModelYou controlPacinfraX deliversBest for
Build-to-suitSpecification, density, cooling, redundancyLand, power, design, construction, commissioningLarge, specific clusters
Powered shellInterior fit-out, systems, timelineLand, power, building envelope, utilities to the shellOperators with their own fit-out team
Hall leaseRacks, workloadsFitted hall with power and coolingMid-size deployments
Rack colocationServersRacks, power, cooling, securityDeployments from a few racks

Phased delivery

Delivery in phases, energized block by block

Every project moves through the same phases: requirement, design, power secured, construction, energization and operations. Power and building work run in parallel, which is why a first block can energize while later blocks are still being built.

The four phases in brief
  • Requirement: your load, timing and delivery model.
  • Design: basis of design agreed with your engineers.
  • Power secured: generation, storage and fuel supply planned and lined up for the campus.
  • Construction and energization: power plant and halls built together, commissioned block by block.
Diagram: delivery phases
Diagram: delivery phases from requirement to design, power secured, construction, energization and operations
Delivery phases from requirement to operations.

How we deliver

From first call to energized hall in five steps

Design basis

Reference design for an AI data center campus

A campus is organized as hall blocks of 10–50 MW arranged around a power island. Each block is a repeatable unit, so the second block is built to the same pattern as the first.

Campus reference design table
Reference design for an AI data center campus
ElementDesign intent
Power pathOn-site natural gas generation and battery storage feed the campus, with a bridge to grid supply.
CoolingLiquid-cooling-ready halls with direct-to-chip loops; air-cooled zones for storage and network gear.
RedundancyN+X set per hall for power and cooling, agreed with you at basis of design.
DensityHigh-density AI racks planned per hall and sized to the GPU platform you deploy.
SecurityLayered perimeter, access-controlled halls and monitored entry to every customer zone.
HandoverPowered shell or fully fitted hall, commissioned and tested before your team moves in.
Diagram: campus layout around power
Diagram: campus layout with 10–50 MW data hall blocks around a power island
Campus layout with 10–50 MW data hall blocks around a power island.

Cooling and density

Halls designed around liquid cooling

AI racks concentrate heat, so cooling shapes the building. We design halls with direct-to-chip liquid loops, coolant distribution units and manifolds in the plan from the start, and we leave air-cooled zones for the equipment that still needs them.

Rack density, loop temperatures and piping

Your engineers set the rack density and the loop temperatures for your GPU platform. We size piping, plant space and floor loading to match, so your first deployment goes in without rework and the next generation of hardware has headroom.

Photo: liquid cooling manifold
Direct-to-chip liquid cooling manifold at the back of a GPU rack

Power

Power planned alongside the building

A powered AI campus depends on three things arriving together: generation, storage and a path to the grid. PacinfraX plans all three with the halls, so energization follows construction rather than waiting for a utility study to finish.

Generation first, storage for ride-through, grid as a bridge

The campus runs first on its own natural gas generation, with battery storage for ride-through and peak support. The grid connection follows later as a bridge, and your halls see the same supply the whole time. Read the detail on our power infrastructure page.

Photo: data hall exterior at dusk
Exterior of a large data hall at dusk with substation lighting

After handover

Operations that keep clusters running

A campus control room watches power and cooling across every block. Operators see generation output, battery state and cooling loops on one set of screens, with alarm feeds that can route to your team as well as ours.

What you receive at handover

At handover you receive the commissioning record for your hall, the as-built drawings and a runbook for the systems you share with the campus. You can run your hall yourselves or agree a services scope with us.

Photo: campus control room
Campus control room with monitoring screens for power and cooling

Where available

Build where the power is planned

Pick the market that fits your latency, customers and timeline.

Where we build: Texas, Minnesota, Vietnam
Data center campus site in South ERCOT, Texas

South ERCOT, Texas

An AI campus in development with on-site gas generation and storage, planned so halls energize without waiting on the interconnection queue.

Explore the Texas campus →
Data center campus site in Vietnam

Vietnam

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.

Explore Vietnam sites →

Before you call

What to send us for a fast, useful reply

You do not need a finished specification to start. The more of this you share, the faster we can match you to a campus and a power plan:

Six things to send us before you call
  • Megawatts needed at first energization, and the size of later phases.
  • Target date for the first hall and the date you expect to reach full load.
  • Preferred market: Texas, Minnesota, Vietnam or flexible.
  • Delivery model: build-to-suit, powered shell, hall lease or undecided.
  • Rack density and cooling approach, if known.
  • Redundancy target for power and cooling.

Ready? Request capacity, or read why power-first delivery changes the schedule and meet the leadership team.

Photo: engineers reviewing drawings
Engineers reviewing data hall drawings and rack layouts on a table

FAQ

AI data center build-to-suit questions

What is an AI data center build-to-suit?

A build-to-suit is a data center designed and built to one customer's specification. PacinfraX secures land and power, designs to your standard and hands over a commissioned facility.

What is the difference between build-to-suit and a powered shell?

In a build-to-suit we deliver the finished hall. In a powered shell we deliver the building and the power, and your team fits out the interior. Both start with the same power plan.

How large are your hyperscale capacity blocks?

We design hyperscale blocks in 10–50 MW increments. Start with one block and add more on the same campus as demand grows.

How do you deliver power faster?

We pair on-site natural gas generation with battery storage and a bridge to the grid. Halls energize ahead of the interconnection queue.

4 more questions
Where can PacinfraX build?

Our markets are South ERCOT, Texas, where an AI campus is in development; Minnesota, powered industrial site; and Vietnam, with 1,200 MW planned across four sites.

Do your halls support liquid cooling?

Yes. Every hall is liquid-cooling-ready, with direct-to-chip loops planned from the first drawing. Density and redundancy are set per hall with your engineers.

What do I need to start a conversation?

Your megawatts, target dates and preferred market. A finished specification is optional; we help you shape it at basis of design.

Can I start smaller and grow?

Yes. Begin with one block, or with rack colocation, and add capacity on the same campus as your clusters grow.

Request capacity

Request capacity for your next AI campus

Tell us how many megawatts you need, where and when. We come back with a site, a power plan and a delivery schedule.

  • A matched site and power plan
  • A delivery schedule for your first block
  • One team from first call to handover

Request capacity or write to [email protected].

One form, one team. Requests go through our contact page and reach the capacity team directly.

Request capacity