Oracle is preparing one of the largest AI infrastructure moves now tied to OpenAI: an investment of around $40 billion in Nvidia chips for a new data center in Abilene, Texas, according to the Financial Times.
The project is described as the first Stargate project, but the financing and operating structure is more specific than the name suggests. Oracle will provide the computing power to OpenAI, while the facility itself is owned by investment firm Blue Owl Capital and data center startup Crusoe.
What Oracle Is Buying
The reported investment centers on Nvidia hardware. The Abilene, Texas facility will use about 400,000 Nvidia GB200 chips, placing the chip order at the center of the project’s scale.
For OpenAI, the practical result is access to a very large amount of compute. For Oracle, the arrangement is a major infrastructure play: the company will rent the computing power to OpenAI rather than simply supplying equipment or owning the site outright.
The numbers in the plan are unusually large even before the facility is operational. Around $40 billion in Nvidia chips, about 400,000 Nvidia GB200 chips, and 1.2 gigawatts of expected power together explain why the project is being ranked among the world’s largest AI data centers.
How The Abilene Data Center Is Structured
The ownership structure separates the facility from the compute rental relationship. The data center is owned by Blue Owl Capital and Crusoe. Crusoe has raised $15 billion in funding, according to the source article.
That means the project is not simply an Oracle-owned data center being handed to OpenAI. Instead, several roles sit side by side:
- Oracle is investing around $40 billion in Nvidia chips and will rent computing power to OpenAI.
- OpenAI is the customer for that computing power.
- Blue Owl Capital and Crusoe own the Abilene, Texas facility.
- Stargate is the project banner, but Stargate itself is not financially involved.
This distinction matters because the name Stargate could imply a single financial entity behind the buildout. The source says the opposite: Stargate itself is not financially involved in the Abilene project.
Why The Power Figure Matters
The facility is expected to deliver 1.2 gigawatts of power. In the context of AI data centers, the power figure is one of the clearest indicators of scale because chips, servers, cooling systems, and supporting infrastructure all depend on available energy.
The source article links that 1.2 gigawatts figure directly to the project’s status among the world’s largest AI data centers. It also helps explain why the chip count and the investment level are being discussed together. A data center built around about 400,000 Nvidia GB200 chips requires an infrastructure plan large enough to support that hardware.
The Abilene site is therefore not only a real estate project or a hardware purchase. It is a combined compute, energy, and financing arrangement designed to supply OpenAI with large-scale AI capacity.
Timeline And Next Steps
Operations are set to begin in mid-2026. Until then, the reported project remains a buildout in progress, with the core commercial arrangement already outlined: Oracle invests in Nvidia chips, the facility owners provide the site, and OpenAI rents the resulting computing power.
The source also says OpenAI plans a similar project in Abu Dhabi with other partners. No further details are given in the source article, but the mention shows that Abilene is not the only large infrastructure project connected to OpenAI’s future compute needs.
For now, the Texas facility is the clearest example in the source of how the Stargate label is being applied: a massive AI data center project, tied to OpenAI, powered by Nvidia GB200 chips, and arranged through partners rather than through Stargate’s own financing.
The Bigger Takeaway
The Abilene project shows how AI infrastructure is being assembled through overlapping responsibilities. One company can supply chips and rent computing power, another set of partners can own the data center, and OpenAI can use the resulting capacity without owning every part of the stack.
That structure also makes the reported $40 billion Nvidia chip investment more than a procurement headline. It is the central input in a facility expected to deliver 1.2 gigawatts of power and become one of the world’s largest AI data centers when operations begin in mid-2026.
Based on the details available, the project’s importance comes from its scale and its structure. Oracle, Nvidia, OpenAI, Blue Owl Capital, and Crusoe each occupy a different role, while Stargate serves as the project name without being financially involved.