A major new financing package is moving OpenAI's Stargate infrastructure plan from announcement toward physical scale. The Abilene, Texas data center under construction has secured $11.6 billion in new funding, raising the total committed to the project to $15 billion.
The site is expected to become the largest data center used by OpenAI once it is complete next year. It also stands out because it is described as the first concrete piece of the broader Stargate vision, a planned network of AI data centers associated with OpenAI, SoftBank and Oracle.
What the new funding covers
The $11.6 billion in new financing comes from Crusoe and Blue Owl Capital. Crusoe will operate the Abilene facility, while Blue Owl Capital is the investment firm involved in the round.
With this financing in place, the project now has $15 billion committed. That figure matters because large AI training facilities require more than ordinary data center capacity. They need dense computing infrastructure built around specialized chips, long-term operational planning and customers prepared to lease large amounts of server capacity.
For OpenAI, the Abilene site is not simply another computing contract. It is linked to a larger effort to expand the infrastructure available for training and running advanced AI systems. The source article frames the project as part of OpenAI's move to reduce dependence on outside providers, even though the Abilene arrangement still includes major partners.
Inside the Abilene data center plan
The facility will consist of eight buildings. Each building will be capable of housing up to 50,000 Nvidia Blackwell chips.
Those chips are specifically designed for training large language models. That detail explains why the site is closely tied to OpenAI's future capacity needs: large language model training depends on specialized hardware, and the Abilene design concentrates that hardware at enormous scale.
The structure of the deal also shows how many companies are involved in making the site useful to OpenAI. Crusoe will operate the facility. Oracle will lease the site for 15 years and supply the hardware. OpenAI will lease server capacity at the center through Oracle.
That chain is important. OpenAI is the expected user of the computing capacity, but the site is being financed, operated, leased and supplied through a set of partners. The result is an infrastructure project built around OpenAI's needs without OpenAI being described as the direct operator of the facility.
Why Stargate matters for OpenAI
Until now, OpenAI relied exclusively on Microsoft's infrastructure. According to the Wall Street Journal, the company was dissatisfied with the available capacity.
The Abilene project should be understood against that backdrop. If a company building large language models cannot get enough compute capacity from existing arrangements, it has a strong reason to support dedicated infrastructure with long-term access.
That does not mean OpenAI is moving alone. The Abilene plan still involves Oracle, Crusoe and Blue Owl Capital. But the project gives OpenAI a clearer path to large-scale server capacity than relying only on the infrastructure it previously used.
OpenAI CEO Sam Altman announced the Stargate project back in January alongside SoftBank and Oracle. Stargate was presented as a planned network of AI data centers backed by a proposed $500 billion investment.
The announcement attracted criticism, including from Elon Musk, who questioned the project's funding. At the same time, Musk's own company xAI is also scaling up to hundreds of thousands of GPUs.
Abilene as the first visible piece
So far, Abilene is the clearest physical example of what Stargate is meant to become. Altman described it on X as the world's largest AI training facility.
The source article says other potential Stargate sites have not been made public. That leaves Abilene carrying much of the public evidence for the project for now. It is under construction, has named financing partners, has a long lease arrangement through Oracle and has a defined hardware direction built around Nvidia Blackwell chips.
Those details make the site more concrete than a general infrastructure pledge. The project has a location, an operator, a financing total and an expected completion window. It also has a defined role: OpenAI will lease server capacity there through Oracle.
For the AI industry, the broader signal is straightforward. Training large language models is pushing companies toward purpose-built computing facilities at a scale that requires many firms to coordinate. Abilene shows how that coordination can look in practice: financing from one set of players, operations from another, hardware supplied through a long-term lease, and server capacity reserved for the AI company that needs it.
What remains unknown
The Abilene facility answers some questions about Stargate, but not all of them. The source article does not identify other sites in the planned network. It also does not provide a full map of how the proposed $500 billion investment would be allocated.
What is clear is that the Abilene project has moved beyond broad ambition. A total of $15 billion is now committed, including $11.6 billion in new funding, and the site is expected to be completed next year.
For OpenAI, the practical issue is capacity. For its partners, the project is a long-term infrastructure bet. For Stargate, Abilene is the first concrete test of whether a proposed AI data center network can translate into operating facilities built for the demands of large language model training.