At nearly the same moment Nvidia CEO Jensen Huang was expressing surprise over OpenAI’s multibillion-dollar deal with competitor AMD, OpenAI’s Sam Altman was saying that more such deals are in the works. This comes shortly after Nvidia itself agreed to invest up to one hundred billion dollars into the AI model maker.
Huang appeared on CNBC’s Squawk Box on Wednesday. When asked if he knew about the AMD deal before it was announced, he answered, “Not really.” As previously reported, OpenAI’s deal with AMD is unusual. AMD has agreed to grant OpenAI large tranches of AMD stock, up to ten percent of the company over a period of years, contingent on factors like increases in stock price. In exchange, OpenAI will use and help develop the chipmaker’s next-generation AI GPU chips. This arrangement makes OpenAI a shareholder in AMD.
Nvidia’s deal is the reverse. Nvidia has invested in the AI model-making startup, making it a shareholder in OpenAI. While OpenAI has used Nvidia gear for years through cloud providers like Microsoft Azure, Oracle OCI, and CoreWeave, Huang explained that this is the first time Nvidia will sell directly to them. He added that his company would still continue to supply gear to the cloud makers.
These direct sales, which include AI gear beyond GPUs like systems and networking, are intended to prepare OpenAI for the day when it becomes its own self-hosted hyperscaler. In other words, when it is using its own data centers. Huang admits that OpenAI does not have the money yet to pay for all of this gear. He estimated that each gigawatt of AI data center will cost OpenAI fifty to sixty billion dollars to cover everything from the land and power to the servers and equipment.
So far in 2025, OpenAI has commissioned ten gigawatts worth of U.S. facilities through its five hundred billion dollar Stargate deal with partners Oracle and SoftBank. It also penned a three hundred billion dollar cloud deal with Oracle. Its partnership with Nvidia was for at least ten gigawatts of AI data centers, and its partnership with AMD was for six gigawatts. It also has a Stargate UK partnership involving expanding data centers in the U.K., along with other European commitments. By some estimates, OpenAI has this year inked one trillion dollars worth of such deals.
Similar to the AMD deal, Nvidia’s deal has been criticized for being circular. Critics say Nvidia is essentially underwriting OpenAI’s purchases, getting the AI startup’s stock for its efforts.
As Huang was dissecting OpenAI’s infrastructure needs on CNBC, OpenAI CEO Sam Altman’s interview with the Andreessen Horowitz podcast dropped. During the podcast, a16z co-founder Ben Horowitz told Altman that he is very impressed by deal structure improvement, referring to these most recent deals. Andreessen Horowitz is an OpenAI investor.
When asked about these recent deals, Altman said, “You should expect much more from us in the coming months.” Altman sees OpenAI’s future models and upcoming products as so much more capable that they will fuel much more demand. He explained that the company has decided it is time to go make a very aggressive infrastructure bet.
The problem is that OpenAI’s revenue today is currently nowhere near one trillion dollars, though it is, by all accounts, growing rapidly, reportedly hitting four and a half billion dollars in the first half of 2025. Yet Altman obviously believes that eventually all of this investment will pay for itself. He stated that he has never been more confident in the research road map and the economic value that will come from using future models.
But he said OpenAI cannot get to all of that economic lushness on its own. To make the bet at this scale, the company needs the whole industry, or a big chunk of the industry, to support it. This spans from the level of electrons to model distribution and all the stuff in between. Altman said they are going to partner with a lot of people, with more deals expected in the coming months. So stand by, tech industry. OpenAI is still wheeling and dealing.