Nvidia has confirmed that its proposed 100 billion dollar partnership with OpenAI has not reached a final agreement. The disclosure came from Nvidia CFO Colette Kress during the UBS Global Technology and AI Conference. The company stated that the plan remains at the letter of intent stage and that negotiations are ongoing.
Early reports described the potential agreement as a multi year deployment of millions of Nvidia GPUs for OpenAI’s infrastructure. The large scale project would involve an estimated 10 gigawatts of data center capacity. Kress said that discussions continue but added that no binding contracts have been completed. The information was also included in Nvidia’s recent 10 Q filing, where the company noted that there is no assurance a definitive agreement will be reached.
Nvidia outlined several risk factors associated with long term supply arrangements. These include long lead and non cancellable orders that could create excess inventory if customer requirements change. The company also identified challenges related to forecasting demand in a market where new architectures are released annually. Kress stated that current demand projections for Blackwell and Vera Rubin systems in 2025 and 2026 do not include any OpenAI related purchases. Nvidia estimates approximately 500 billion dollars of demand for these systems across that period, independent of any potential agreement with OpenAI.
Nvidia shares increased by roughly 2.6 percent following the remarks, although market concerns regarding the broader AI investment environment remain. Analysts have raised questions about circular investment practices in which Nvidia invests in AI startups that later buy Nvidia hardware. Institutions including the Bank of England have highlighted the possibility of sharp market corrections if investment patterns shift or if long term agreements become difficult to sustain.
When asked about competitive pressure from companies building their own accelerators, Kress said Nvidia is not concerned. She noted that many industry participants continue to rely on Nvidia’s hardware and that any shift to alternative platforms would take considerable time.
