Altman, Huang negotiations sealed $100 billion Openai-Nvidia deal

Abilene, Texas – Sam Altman had a deadline. Openai’s CEO heads to Texas to announce the company’s next big infrastructure push, nvidia CEO Jensen Huang wanted this action.

Through rushed negotiations, late-night calls and last-minute contract adjustments, the two artificial intelligence giants attacked the $100 billion partnership on Monday hours before Altman boarded a flight to Abilene, a city of about 130,000, about 180 miles west of Dallas.

Huang and Altman were part of President Donald Trump’s state visit to the state a week ago, allowing the president to get used to the agreement in advance.

The deal, which Huang described to CNBC as “a monumental size,” marks a fork moment in the tech industry.

Huang currently hosts the world’s most valuable public company, worth around $4.5 trillion after winning $170 billion after its announcement on Monday.

Openai’s rise to the forefront of generation AI relies on Nvidia’s powerful graphics processing unit (GPU). Now, businesses are more closely linked than ever as they plan to pave the path to collaborate on the next wave of AI supercomputing facilities.

“We should expect a lot from us in the coming months,” Altman told CNBC’s John Fort in an interview at Nvidia’s Silicon Valley headquarters on Monday. “There are three things Openai needs to do well. We have to do great AI research. We have to make these products that people want to use, and we have to find ways to do this unprecedented infrastructure challenge.”

Altman and Huang negotiated the agreement primarily through a mix of virtual discussions and one-on-one meetings in London, San Francisco and Washington, DC.

The arrangement calls for Nvidia to invest $10 billion at a time in Openai, the company behind ChatGpt. Once BuildOut is deployed, Nvidia will also supply state-of-the-art processors powered by many new data centers.

Openai will become more intimate with Nvidia, but will need to be manipulated through many high-stakes relationships with other important partners.

Openai only notified Microsoftits major shareholders and major cloud providers, the day before the contract was signed, said those familiar with the issue. Earlier this year, Microsoft lost its status as Openai’s computing power provider.

The agreement will arrive within two weeks of disclosure from Oracle Openai has indicated it has agreed to spend $300 billion in computing power on the company over nearly five years starting in 2027. Earlier this year, Openai joined Stargate, a multi-billion dollar project announced by President Trump and supported by Oracle and Bucked. Softbank, Build a next-generation AI infrastructure.

Going forward, all Openai infrastructure projects will be under Stargate’s umbrella.

Representatives from Microsoft, Oracle and SoftBank did not immediately respond to requests for comment.

Nvidia and Openai provided scarce details on when and where the buildout will take place, except that the first site of the 10 gigawatt sites will be online later next year.

Executives said they have reviewed between 700 and 800 potential locations since they announced Stargate in January. In the months that followed, they announced a flood of proposals from North American developers providing land, electricity and facilities. The list is narrowing as a funding condition, as Openai has considered energy availability, allowed timelines, the company said.

In its announcement Monday, Openai described Nvidia as a “preferred” partner. However, executives told CNBC that it is not an exclusive relationship, and the company continues to work with large cloud companies and other ship manufacturers to prevent it from being locked up in a single vendor.

For Nvidia, investment in Openai is historic size, but it is just a big part of its rapidly expanding portfolio.

Last week, Nvidia invested $5 billion Intel As part of a joint venture with a troubled chip manufacturer with a data center and PC chips. Nvidia also said it has invested nearly $700 million in UK data centre startup Nscale. coreweaveWe held an IPO in March.

The financing structure of Openai transactions is designed to avoid heavy dilutions. The first $10 billion tranche is locked in a $500 billion valuation and is expected to close for about a month once the transaction is closed. Nine consecutive $10 billion rounds are planned, they said, each will be priced at the company’s then valuation, as new capabilities will arise online.

Nvidia and Openai’s relationship was before the 2022 release of ChatGPT.

When Openai was still a small non-profit research lab and Nvidia was best known for building graphics chips for video games, Huang personally delivered his company’s first DGX supercomputer to Openai’s office in 2016.

After almost a decade and trillion dollars worth, Huang and Altman are probably the most important power players in the tech industry.

Last October, Nvidia officially made Openai’s financial interests, taking part in a $6.6 billion funding round, valued at $157 billion. A month later, in Tokyo, Openai executives met up with Softbank CEO Masayoshi Son and brainstormed what they called the next expansion phase. Of that session, the codename “Stargate,” short for Openai’s most ambitious build-out plan, came out.

Stargate currently covers all major transactions in terms of computing power, including this week’s partnership with Nvidia. Securing the rights to the name required careful manipulation, but Openai accepted it as a banner for its long-term infrastructure strategy.

The $100 billion commitment from Nvidia is just part of what it takes to build out the planned 10-Gigawatt. Openai leasing Nvidia tips for deployment, but other paths will be needed to fund the broader effort. Executives call equity the most expensive way to fund data centers, and say startups are preparing to undertake debts to cover the remaining expansions.

With the rise in Openai’s computing essentials, the big problem is where the company hosts its workloads, which has previously been housed primarily at Microsoft Azure. Working in-house will bring you closer to operating as a first-party cloud provider, the leading market for Openai Amazon Web services that continue with azure, Google and Oracle.

Executives are openly floating the idea, suggesting that it may not be far away. We have shown CNBC that if Openai has enough calculations to cover its own needs, the commercial cloud offering could emerge within a year or two. For now, there is little demand for training frontier models, but Openai hasn’t completed it looking for new opportunities.

As Altman and Huang unveiled details of the arrangement announced this week, Openai’s infrastructure team met Softbank’s son and Tokyo to discuss broader funding and manufacturing support.

The parallel talks highlight the scale of Altman’s ambitions, and the web of global players is involved in making it happen.

Watch: Openai Reorganization Clears the Hurdles