Aerial view of STN's Odessa campus at sunrise: rows of Elemental Data Centers, cooling units and power infrastructure on open West Texas land

STN BlogField report

Ector County, Texas

From powered land to running compute: inside STN’s first fully integrated AI campus.

AI’s bottleneck has moved from chips to power and buildings. In Odessa, one team controls the land, the power, the buildings, the compute and the intelligence that runs it all.

Sabur MianFounder & CEO, STN · 9 min read

Photo: STN, Odessa campus

The build

Unused industrial land. Ready for compute by July 31.

Stop Before

Before

The land

A former industrial site in Ector County that had sat largely unused for years.

May

Construction begins

Crews prepare the site while the buildings are manufactured in parallel at the factory.

First buildingJun 11

First EDC placed

The first Elemental Data Center arrives complete, with essential systems installed and tested.

CommissionedJul 31

Ready for compute

Set, connected and fully commissioned: designed for 2 MW and 1,024 NVIDIA B300 GPUs.

Today

Building with Ector County

The project currently supports 375 jobs, with training partners at Odessa College and UT Permian Basin.

As operations expand

200 permanent, highly skilled Dark AI Factory positions expected in Odessa.

Then

The same model, again

Projects in development in Marshfield, Missouri and Orangeburg, South Carolina.

Keep scrolling to follow the buildAll dates 2026

01AI’s bottleneck has moved from chips to power and buildings

The chips are racing ahead. Everything around them isn’t.

AI is moving at a ridiculous pace. The models keep getting smarter, NVIDIA keeps pushing the boundaries of what’s possible and every few months we’re talking about hardware that makes the previous generation look old. I love watching it happen. But there’s a problem we need to solve, and it has very little to do with how fast the chips are getting.

It’s everything around them.

We’re building technology capable of changing entire industries, but getting enough power, cooling and physical infrastructure to support it is becoming one of our biggest challenges. And unlike software, you can’t simply push an update and have a new data center appear overnight.

In the traditional model, the pieces that determine when AI capacity goes live sit with different companies. One holds the land, a utility supplies the power, a developer builds the facility and someone else operates the hardware. Each company can be on schedule, but one delayed handoff can hold up the entire project. Not exactly ideal when the technology you’re building for is evolving this quickly.

That’s why we made the decision to vertically integrate STN. We control the land, the power, the buildings, the compute and the intelligence that operates it all through STN ONE. Five layers working together, with one team accountable for delivering AI capacity.

Our first fully integrated campus is in Odessa, Texas. And for me, this is where a pretty ambitious idea starts becoming something real.

Aerial view of the Odessa campus: Elemental Data Centers at center, with rows of backup power units in the foreground — **Fig. 1**STN’s Odessa campus in Ector County, Texas, with Elemental Data Centers at center and N+1 redundant backup power in the foreground.

02What vertical integration means in AI infrastructure

Five companies, five schedules. Or one team.

Vertical integration means one company takes responsibility for everything required to turn a piece of land into running AI compute.

Traditionally, that process involves a landlord, a utility, a developer, a colocation provider and an infrastructure operator. Each has its own schedule, priorities and dependencies. You can have five companies doing everything right individually and still end up with a delayed project because the pieces aren’t moving together. That’s the problem we wanted to eliminate.

The traditional model

LandlordHolds the land

UtilitySupplies the power

DeveloperBuilds the facilityWaiting

Colocation providerLeases the space

OperatorRuns the hardware

With STN

One team, one coordinated schedule

When we own the entire lifecycle, site preparation, power infrastructure, building production and compute deployment can operate on one coordinated schedule. Our engineers design each layer around the others rather than trying to make separate systems fit together after the fact.

And when something needs attention, there’s one accountable team that understands the entire stack, from the electrical systems to the GPUs. We aren’t spending time figuring out whose problem it is. We’re solving it.

We aren’t spending time figuring out whose problem it is. We’re solving it.One accountable team, from the electrical systems to the GPUs.

03The STN stack, layer by layer

Five layers. All of them in Odessa today.

We’ve brought five layers together under one operating model. Each one exists in Odessa today, and each plays a specific role in getting AI capacity online.

Land

Every campus starts with a site that can support large-scale power.

In Odessa, we chose a former industrial site in Ector County that had sat largely unused for years. West Texas has spent decades building the energy infrastructure that supports some of the country’s most demanding industries. Now we’re putting that existing industrial footprint back to work for the next generation of computing.

There’s something exciting about that. A region known for powering the energy industry is helping build the infrastructure that will power AI.

The Odessa site among transmission structures and industrial buildings, with Elemental Data Centers in the distance — **Fig. 2**A former industrial property put back to work: the Odessa site sits among the energy infrastructure West Texas has built over decades.

Power

Power is arguably the biggest constraint in AI infrastructure today. You can have the GPUs, customers and capital lined up, but without electricity, you’ve got a very expensive collection of hardware sitting idle.

Our Odessa campus uses on-site power generation and battery storage alongside a utility connection. It’s designed to meet its own energy needs rather than depend entirely on the grid’s ability to deliver additional capacity.

That also means taking responsibility for the costs our infrastructure creates, consistent with the direction Texas is taking for large power users.

I believe if we’re going to build infrastructure at this scale, we need to be responsible for powering it. Generating our own power gives us greater control over our delivery schedule while reducing our dependence on additional grid capacity.

The side of an STN-branded Elemental Data Center at dusk, with power lines and a substation behind it — **Fig. 3**An Elemental Data Center at dusk, with the campus’s utility interconnection behind it.

Building

Our Elemental Data Centers, or EDCs, take a different approach to construction. They’re complete buildings, factory-built to STN’s specifications, with essential operating systems already installed and tested. While crews prepare the site in Odessa, the buildings are being manufactured in parallel.

It’s a simple idea, but it changes the timeline considerably. Why wait until the site is finished to start building the data center?

Once delivered, the EDCs are set, connected and commissioned. We’ve designed the process so that building production and site preparation happen together rather than one after the other.

One after the other, or together

Traditional

Site prepBuild on siteCommission

STN EDC

Site prepFactory buildSet + commission

Schematic, not to scale.

Compute

Inside those buildings, STN operates AI compute through GPU One. Because we’re responsible for both the facility and the hardware inside it, our engineers know exactly what the environment around every rack is designed to deliver. Power, cooling and compute requirements are planned together from the beginning.

That becomes increasingly important when every new generation of GPUs introduces new demands on the infrastructure supporting them.

Inside an Elemental Data Center: GPU racks below overhead cooling lines, with a red fire suppression cylinder at the end of the aisle — **Fig. 4**Racks connected to the sealed cooling loop, with fire suppression built in.

Looking up a rack of servers with yellow and blue cabling — **Fig. 5**Cabled, tested, ready.

Operations

STN ONE is our enterprise AI agent platform and the operating layer across the stack. We’re already using it across STN’s support, finance, security and engineering operations, with agents working alongside our people to monitor systems, identify potential problems and automate routine work.

We’ve made two principles central to that approach.

We run it on ourselves first.STN ONE already works across our own support, finance, security and engineering operations.

Humans stay responsible.People remain responsible for sensitive decisions and actions. AI should make our people more capable, not take accountability away from them.

Ultimately, we’re building toward infrastructure that can increasingly monitor, maintain and optimize itself. And owning the entire stack gives us the visibility needed to make that possible.

04Odessa: from construction start to commissioned capacity

Placed June 11. Commissioned July 31.

Construction began in Odessa in May 2026. We placed our first Elemental Data Center on June 11 and fully commissioned it by July 31, ready to accept compute.

50 daysfrom first building placed to fully commissioned

Odessa, 2026

Construction beginsMay

First EDC placedJune 11

CommissionedJuly 31

MayJuneJulyAug

That’s a timeline I’m particularly proud of, because it demonstrates what happens when you control the dependencies instead of waiting on them. The building was manufactured while the site was being prepared. We controlled the land, planned the power and coordinated the infrastructure around one delivery schedule. The building became the primary date governing when capacity could come online.

We also designed the campus specifically for West Texas. Our sealed, closed-loop cooling system is engineered for Odessa’s arid climate, with no process water discharge. Water matters in this region, and we wanted an approach that takes those local conditions seriously rather than treating them as an afterthought.

Rows of cooling units and insulated piping beside the Elemental Data Centers

Sealed, closed-loop cooling. No process water discharge.Fig. 6 · The closed-loop cooling system: rows of cooling units and insulated piping serve the Elemental Data Centers, designed for West Texas’ arid climate.

05Built for the next generation of AI hardware

Hardware changes in years. Buildings last decades.

AI hardware changes in years. Data centers are built to last decades. If we design around one generation of GPUs, we’re potentially limiting the building before the next generation even arrives.

Our Elemental Data Centers are designed to adapt. The first EDC is designed for 2 megawatts and 1,024 NVIDIA Blackwell Ultra B300 GPUs. By changing a single element, that same footprint can support 1,008 NVIDIA Grace Blackwell GB300 GPUs.

1,024NVIDIA Blackwell Ultra B300 GPUs

BuildingThe first Odessa EDC

Designed for2 megawatts

FootprintThe same

What changesNothing yet

That’s the advantage of owning the entire stack. Our engineers can adapt the power, cooling, building and compute together rather than starting over with every new generation of hardware. We can’t predict every advancement in AI, but we can build infrastructure that’s ready for change.

06What it means for customers

Anyone can host a workload. We operate it with you.

I can talk about the engineering behind all of this all day. But ultimately, the question is what it changes for the customers who depend on us.

One accountable partner.

From the site to the application, one team is responsible. No chasing down landlords, utilities, builders and operators to figure out who owns an issue.

Delivery dates tied to what we control.

We manufacture the buildings, control the site and generate the power. That gives us greater control over the schedule and makes capacity something customers can actually plan around.

Infrastructure that keeps pace.

Our facilities are designed to adapt as AI hardware evolves. The infrastructure we build today needs to remain useful for the next generation.

Operated with you, not just hosted.

STN began as a managed service provider for regulated industries. We know what it means to be responsible for critical systems, and we bring that operational discipline to every environment we run.

We’ve always believed there’s a difference between providing infrastructure and taking responsibility for it. Anyone can host a workload. We operate it with our customers, from the facility to the application.

07Building with Ector County

A region that powered energy, now powering AI.

The Odessa project currently supports 375 jobs and is expected to create 200 permanent, highly skilled Dark AI Factory positions as operations expand.

375jobs the Odessa project supports today

200permanent Dark AI Factory positions expected as operations expand

We’re also working with regional education and workforce organizations, including Odessa College and the University of Texas Permian Basin, to develop training for careers in AI infrastructure. That includes advanced computing equipment maintenance, data center operations and the specialized electrical and technical skills these facilities require.

West Texas has an incredible history of building and operating complex energy infrastructure. We’re excited to help bring a new generation of technical careers to the region, building on the expertise that’s already there.

Odessa CollegeUniversity of Texas Permian Basin

Two people in hard hats and safety vests stand beside a crane on the Odessa site, reflected in a puddle — **Fig. 7**On site in Odessa.

From the beginning, STN has proactively looked for ways to support our county. They acknowledged the most common community concerns right away. For instance, they made it clear that the build will rely on its own water supply. They are prepared to be a good partner to Ector County by bringing an emerging industry and opportunities to a community that has, for decades, relied solely on oil and gas. We’re excited for what STN is bringing to our community in terms of jobs, workforce training, and tax revenues.Dustin FawcettEctor County Judge

08What’s next

One campus is a start. The model is the point.

Odessa is our first fully integrated AI campus, but we’re not building all of this just to stop at one location. We have additional projects in development in Marshfield, Missouri and Orangeburg, South Carolina, each using the same integrated model.

CommissionedOdessa, TexasFirst fully integrated AI campus

In developmentMarshfield, MissouriSame integrated model

In developmentOrangeburg, South CarolinaSame integrated model

I’ve talked before about STN becoming a utility company for the AI age. That’s the bigger picture. AI agents are getting more capable, robotics is moving from programmed repetition toward real-world reasoning and we’re only scratching the surface of what these technologies will make possible. Every one of those advancements needs computational power. An extraordinary amount of it. And someone has to build the infrastructure to deliver it.

We’re also working toward autonomous data centers, where STN ONE can take on more of the routine work of monitoring, maintaining and optimizing infrastructure. Not to replace the talented people running these facilities, but to give them the freedom to work on harder problems. That’s where I believe AI gets really interesting. And it’s why owning the entire stack matters. You can’t intelligently operate systems you can’t see or control.

I don’t pretend to know exactly what AI will look like ten years from now. Frankly, I hope it surprises us. But I do know we’re going to need a whole lot more infrastructure to get there.

That’s what we’re building at STN. Odessa is just the beginning.

Newsroom

Odessa in the news

Announcements, coverage and updates from STN’s first fully integrated AI campus in Ector County, Texas.

All STN news ↗

GlobeNewswireNov 11, 2026Press release

STN Expands Into AI Data Centers With Fully Integrated Model

STN now controls land, power, data center buildings and compute, with Odessa as the first campus to bring all four together.

Read the release

Frequently asked questionsThe short answers on Odessa, Elemental Data Centers and how STN builds.

What is vertical integration in AI infrastructure?

It means one company manages every stage of delivering AI compute: securing land, providing power, building the data center, and operating the compute inside it. STN One Incorporated controls all of these layers, which removes handoffs between separate providers and puts delivery on one schedule.

What is an Elemental Data Center?

An Elemental Data Center is STN’s factory-built data center. Each one is a complete building manufactured to STN’s specifications with essential operating systems already installed, then delivered, set, connected, and commissioned on site.

How is the Odessa campus powered?

The campus uses on-site power generation and battery storage alongside a utility connection. It is designed to meet its own energy needs, in line with Texas’ direction that large power users cover their associated costs.

How is the Odessa campus cooled?

It uses sealed, closed-loop cooling engineered for an arid climate, with no process water discharge.

How quickly did STN bring capacity online in Odessa?

Construction began in May 2026. STN placed its first Elemental Data Center on June 11, 2026, and fully commissioned it by July 31, ready to accept compute.

Where else is STN building?

In addition to Odessa, Texas, STN has projects in development in Marshfield, Missouri, and Orangeburg, South Carolina.

AI capacity, delivered by one team

Plan your next cluster around a date we control.

Land, power, buildings, compute and operations, under one accountable team. Tell us what you need to run, and we’ll show you where and when it can go live.

Talk to our team about AI capacity ↗Explore GPU One ↗