OpenAI’s Sam Altman teases more massive infrastructure deals as Nvidia CEO reacts to AMD partnership surprise

At the same moment Nvidia CEO Jensen Huang was expressing surprise over OpenAI’s multibillion-dollar deal with AMD, Sam Altman was signalling that more partnerships of similar scale are on the horizon.

Appearing on CNBC’s Squawk Box on Wednesday, Huang admitted that he had not been aware of the AMD deal before it was announced. “Not really,” he said when asked if Nvidia had prior knowledge.

The OpenAI-AMD deal stands out as a rare move in the AI hardware race. Under the agreement, AMD will grant OpenAI stock worth up to 10 percent of the chipmaker’s value, depending on milestones such as stock performance and development progress. In return, OpenAI will use and help refine AMD’s next-generation AI GPUs. This effectively makes OpenAI a shareholder in AMD.

Meanwhile, Nvidia’s agreement with OpenAI works the other way around. Nvidia has invested directly in OpenAI, making the AI company one of its portfolio partners. Huang confirmed that the collaboration includes direct sales of AI systems and networking equipment, marking a shift from the earlier model in which OpenAI accessed Nvidia GPUs through cloud providers such as Microsoft Azure, Oracle OCI, and CoreWeave.

“This is the first time we’re going to sell directly to them,” Huang explained. He added that Nvidia would continue to support major cloud platforms while preparing OpenAI “for the day when it is its own self-hosted hyperscaler.”

However, Huang acknowledged that OpenAI still lacks the financial capacity to fund its full infrastructure expansion. “They don’t have the money yet,” he said, estimating that each gigawatt of AI data centre capacity could cost between $50 and $60 billion, covering everything from land to servers.

In 2025 alone, OpenAI has commissioned 10 gigawatts of U.S. data centre facilities through its $500 billion Stargate partnership with Oracle and SoftBank, alongside a $300 billion cloud deal with Oracle. It has also signed major partnerships with Nvidia and AMD for at least 16 gigawatts of AI compute capacity, and additional projects under “Stargate UK” and European collaborations. Analysts estimate the combined commitments could total over $1 trillion in infrastructure deals this year.

While these agreements have drawn criticism for being “circular,” as reported by Bloomberg, with each company essentially funding OpenAI’s purchases in exchange for equity, Altman remains unapologetic about the aggressive strategy.

In a new episode of Andreessen Horowitz’s a16z Podcast, Altman told Ben Horowitz that OpenAI’s dealmaking spree is only just beginning. “You should expect much more from us in the coming months,” he said.

Altman explained that OpenAI’s decision to double down on large-scale infrastructure stems from the anticipated demand for its future AI models. “We have decided that it is time to go make a very aggressive infrastructure bet,” he said, citing confidence in both the company’s research and the long-term economic potential of its upcoming technologies.

OpenAI reportedly generated $4.5 billion in revenue in the first half of 2025, still a fraction of the investment scale it is pursuing. But Altman believes the payoff will come. “I’ve never been more confident in the research road map in front of us and also the economic value that will come from using those [future] models,” he said.

He also acknowledged that OpenAI cannot execute its ambitions alone. “To make the bet at this scale, we kind of need the whole industry, or a big chunk of the industry, to support it… from electrons to model distribution and everything in between. So we’re going to partner with a lot of people.”