A potential deal between Nvidia and OpenAI, estimated to be worth as much as 100 billion dollars, remains at the stage of preliminary arrangements. Two months after the high-profile announcement in September, which sparked market euphoria and fueled a rally in AI infrastructure stocks, executives at the chipmaker now admit that a final contract has yet to be signed.
“We’re still working on the final agreement”
Speaking at a technology and AI conference in Scottsdale, Arizona, Nvidia CFO Colette Kress stressed that cooperation with OpenAI is still based solely on a letter of intent. When asked about the status of the commitment to deliver 10 gigawatts of computing power, she stated plainly that “the final agreement has not yet been completed” and that both sides are still working through the details.
This marks a significant adjustment from earlier statements. Nvidia CEO Jensen Huang had described the project as “the largest AI infrastructure undertaking in history,” and markets largely treated it as almost a done deal. The latest comments from management, however, show that there is still a long way to go before anything is finalized.
100 billion dollars not in the order backlog
Kress emphasized that the potential agreement with OpenAI is not included in Nvidia’s current order backlog, which is valued at around 500 billion dollars. That figure reflects already contracted sales of Blackwell and Vera Rubin chips through 2026. Any future deal with OpenAI would therefore be on top of these commitments, not part of them.
At present, OpenAI purchases Nvidia chips mainly indirectly — via major cloud providers such as global technology giants — rather than through a direct, long-term infrastructure agreement. The new deal would fundamentally change that model and formalize a direct supplier–customer relationship.
In its recent regulatory filings, Nvidia explicitly warns that there is “no certainty” the planned investment will be completed on the expected terms — or that it will go ahead at all. The company lists a range of critical dependencies: the availability of suitable data-center infrastructure, securing sufficient electricity supply, and raising the capital needed for expansion on an unprecedented scale.
Energy infrastructure as a bottleneck
Nvidia describes the build-out of the energy backbone required to power successive generations of AI data centers as a “multi-year” process, exposed to regulatory, technical and construction risks. According to the company, these are among the main factors slowing down the conclusion of a record-breaking agreement.
Despite the lack of a final contract, Colette Kress stressed that Nvidia’s relationship with OpenAI remains a “very strong partnership” that has been developing for more than a decade. She noted that OpenAI “wants to work directly” with Nvidia, but crafting the target model and delivery schedule simply takes time. Following her appearance, Nvidia’s share price rose by about 2.6%, a sign that the market still believes in the strategic importance of this relationship.
At the same time, Kress rejected suggestions that Nvidia is losing its edge in the AI race or that the sector has entered a speculative bubble. In her view, demand for AI-driven computing infrastructure is structural and multi-year in nature, and the key projects are in any case being built on Nvidia’s platform. Still, the final shape of the agreement with OpenAI remains one of the most important open questions investors are waiting to see answered.
