OpenAI’s next phase is not just about better models. It is about the physical infrastructure needed to run them. The company, together with Oracle and SoftBank, announced plans for five new US AI data center sites for Stargate, its joint AI infrastructure project.
The announcement brings Stargate to nearly 7 gigawatts of planned capacity and over $400 billion in investment over the next three years. The companies say the expansion keeps them on track toward the full $500 billion, 10-gigawatt commitment they announced in January by the end of 2025.
What Stargate Is Adding
The five new sites expand a project that already includes the flagship Stargate campus in Abilene, Texas. Oracle began delivering Nvidia hardware to that site in June, and OpenAI has already started using the data center for training and inference.
Training means building new models. Inference means running services such as ChatGPT. Both require large amounts of specialized computing power, and OpenAI says current demand already strains what it can provide.
Three of the newly announced locations are being developed through an OpenAI and Oracle partnership:
- Shackelford County, Texas
- Doña Ana County, New Mexico
- An unspecified Midwest location
Those sites, combined with a 600-megawatt expansion near Abilene, Texas, can deliver over 5.5 gigawatts of capacity. In practical terms, that means the computers at those sites could draw up to 5.5 billion watts of electricity when operating at full load.
The companies expect those Oracle-linked sites and the Abilene expansion to create over 25,000 onsite jobs. They also said the locations were selected after reviewing over 300 proposals from more than 30 states in a nationwide process launched in January.
SoftBank’s Role In The Buildout
Two of the sites will be developed through a partnership between SoftBank and OpenAI. One is in Lordstown, Ohio, where SoftBank has broken ground and where the site is on track to be operational next year.
The second is in Milam County, Texas. That site will be developed with SB Energy, a SoftBank Group company. Together, the two SoftBank-linked sites may scale to 1.5 gigawatts over the next 18 months.
The announced Stargate footprint so far includes Abilene, Texas; Shackelford County, Texas; Doña Ana County, New Mexico; an undisclosed Midwest location; Lordstown, Ohio; and Milam County, Texas. The Abilene campus is already operational with Nvidia GB200 racks, while the other sites form the next wave of planned capacity.
Why OpenAI Wants So Much Compute
The simplest explanation is demand. ChatGPT serves 700 million weekly active users, more than double the US population. Those users regularly turn to the AI assistant to develop software, provide personal advice, and compose or edit correspondence and reports.
The outputs may be imperfect at times, but the source article makes clear that people still want them. OpenAI regularly faces severe capacity constraints, which can limit how often users are able to query the chatbot.
Those limits are not only a free-user problem. ChatGPT Plus subscribers frequently encounter usage limits too, especially when using more compute-intensive features such as image generation or simulated reasoning models. Free users face even stricter limitations, while also serving as a gateway for future subscriptions.
OpenAI’s infrastructure challenge is therefore twofold. It needs enough capacity to run today’s products, and it also needs room to train future AI models. Training next-generation systems requires thousands of specialized chips running continuously for months.
“AI can only fulfill its promise if we build the compute to power it,” OpenAI CEO Sam Altman said in the announcement. “That compute is the key to ensuring everyone can benefit from AI and to unlocking future breakthroughs.”
That statement captures the company’s argument for Stargate: without more compute, AI services cannot scale to the number of people OpenAI hopes to reach. The question is whether the scale and financing of the buildout can hold up.
The Financing Questions Around AI Infrastructure
The Stargate announcement is also drawing scrutiny because of how money moves between AI companies and infrastructure providers. The July agreement between OpenAI and Oracle to develop up to 4.5 gigawatts of additional Stargate capacity represents a partnership worth over $300 billion between the two companies over five years.
Earlier this week, Nvidia announced it would invest up to $100 billion as OpenAI deploys Nvidia systems. Bryn Talkington of Requisite Capital Management told CNBC: “Nvidia invests $100 billion in OpenAI, which then OpenAI turns back and gives it back to Nvidia.”
Oracle’s arrangement follows a similar pattern, with a reported $30 billion-per-year deal where Oracle builds facilities that OpenAI pays to use. The concern is that infrastructure providers are investing in AI companies that then become major customers for those same providers.
The Information reported this week that Nvidia is discussing leasing its chips to OpenAI rather than selling them outright. Under that structure, Nvidia would create a separate entity to purchase its own GPUs, then lease them to OpenAI.
Tech critic Ed Zitron described a similar pattern on Bluesky last week, writing: “NVIDIA seeds companies and gives them the guaranteed contracts necessary to raise debt to buy GPUs from NVIDIA, even though these companies are horribly unprofitable and will eventually die from a lack of any real demand.” Zitron was referring to companies like CoreWeave and Lambda Labs, which have raised billions in debt to buy Nvidia GPUs based partly on contracts from Nvidia itself.
What Happens If Demand Falls Short
The risk is not that data centers disappear if AI demand disappoints. These facilities are physical assets built on real sites. The issue is whether investors paid AI-boom prices for infrastructure that may need to find other uses later.
Altman himself warned last month that “someone will lose a phenomenal amount of money” in what he called an AI bubble. The source article compares the situation to the dot-com bubble, when fiber optic cable laid during the boom years eventually found use as Internet demand caught up.
Something similar could happen with Stargate-style facilities. They could potentially pivot to cloud services, scientific computing, or other workloads. But that does not erase the possibility of massive losses for investors if AI demand fails to match the projections behind the buildout.
For now, the case for OpenAI’s data centers rests on two facts that exist at the same time: ChatGPT demand is already large enough to create capacity constraints, and the financial structure behind the new infrastructure is complicated enough to deserve scrutiny.