- Oracle signed on to build another 4.5GW of capacity for OpenAI's Stargate project
- The Stargate data centers already under development will house over two million chips
- The announcement comes as President Donald Trump prepares to unveil his AI Action Plan this week
Oracle is emerging as a high-profile player in OpenAI's Stargate project, committing to build nearly half of the project's expected data center capacity.
As part of a new agreement, Oracle has agreed to build out an additional 4.5 gigawatts (GW) of capacity for the project. That means that 5GW of the original 10GW Stargate commitment is now under development. The facilities announced to date will house more than two million chips, OpenAI said in a press release.
"Stargate is an ambitious undertaking designed to meet the historic opportunity in front of us," OpenAI said in a statement. "That opportunity is now coming to life through strong support from partners, governments and investors worldwide."
OpenAI's announcement comes the same week President Donald Trump is expected to sign AI-related executive orders to enable his "AI Action Plan."
Though it wasn’t , the Stargate announcement comes just a few weeks after Oracle revealed in a regulatory filing that it inked a cloud deal “expected to contribute more than $30 billion in annual revenue starting in FY28.”
Fierce has asked Oracle if the Stargate announcement is tied to the aforementioned revenue deal and will update this story when we hear back. It’s worth noting, though, that in addition to Stargate, Oracle is also involved in the Stargate UAE project and has committed to invest $14 billion in Saudi Arabia as well, which is home to the AI-focused HUMAIN project.
Announced by OpenAI in January, the U.S. Stargate project initially included plans to build 10 gigawatts of domestic AI infrastructure over the next four years via a $500 billion investment.
SoftBank, OpenAI, Oracle and MGX were all listed as investment partners. Oracle was also listed among Stargate’s initial technology partners, alongside the likes of Arm, Microsoft and Nvidia.
It’s not clear how much capacity Oracle originally agreed to build, but it has already been involved in the construction process that has been ongoing in Abilene, Texas.
OpenAI indicated that Oracle delivered the first Nvidia GB200 racks in June, which are already being used to run early training and inferencing workloads. It is also working with Softbank on site assessments for locations beyond the Abilene facility.