Samuel Altman had a flight to catch. But before boarding to Texas, he had to lock in one of the biggest deals in AI history.Just hours before takeoff, Sam and Jensen Huang finalized a $100 billion agreement between OpenAI and Nvidia, a deal that would tie the two most important names in artificial intelligence tighter than ever.Sam was headed to Abilene, a small city 180 miles west of Dallas, to announce OpenAI’s next phase in AI infrastructure. Jensen wanted Nvidia to be at the center of it.The two CEOs had already spent time together during President Donald Trump’s recent visit to the U.K., where the president was quietly briefed about the deal in advance. What followed was a blur of virtual calls, private meetings in London, D.C., and San Francisco, and last-minute changes to documents. No banks. Just two power players doing it themselves.Sam and Jensen skip the banks and go all inJensen called the deal “monumental in size” during an interview with CNBC on Monday. That same day, Nvidia’s market value surged by $170 billion, pushing the company near $4.5 trillion in worth. Sam’s OpenAI, now valued at $500 billion, becomes even more dependent on Nvidia’s chips. The plan is to work together to build massive AI data centers.Sam, speaking at Nvidia’s Silicon Valley HQ, said: “You should expect a lot from us in the coming months. There are three things that OpenAI has to do well: we have to do great AI research, we have to make these products people want to use, and we have to figure out how to do this unprecedented infrastructure challenge.”The structure is simple but huge. Nvidia will put in $10 billion at a time, supplying chips as new compute capacity goes live. Sam and Jensen plan to do this ten times, $100 billion total. First chunk is locked at OpenAI’s current $500 billion valuation. The rest will track the company’s future value.This isn’t their first link-up. Back in 2016, Jensen personally dropped off Nvidia’s DGX supercomputer at OpenAI’s office in San Francisco. The place is now home to Elon Musk’s xAI. In October 2024, Nvidia formally joined OpenAI’s $6.6 billion round that valued it at $157 billion. A month later, OpenAI execs met SoftBank’s Masayoshi Son in Tokyo. They picked the codename “Stargate” to brand the new AI buildout plan. Stargate now includes every major compute deal OpenAI touches.Microsoft finds out late, Oracle gets a massive sliceThis new bond with Nvidia doesn’t mean OpenAI is ditching other partners, but some didn’t get much warning. Microsoft, OpenAI’s biggest investor and cloud provider, was only told one day before the announcement. Earlier this year, Microsoft also lost its position as OpenAI’s exclusive compute partner.Then there’s Oracle. Less than two weeks ago, Oracle revealed that OpenAI had agreed to spend $300 billion on compute power from them starting in 2027. That’s separate from this Nvidia deal. Oracle, SoftBank, and President Trump also back Stargate, which is now the main label for all of OpenAI’s infrastructure plans.So far, OpenAI has looked at 700 to 800 possible locations for new 10-gigawatt data centers. Sites across North America have sent in proposals, offering land, power, and buildings. The first site is expected to go live late next year. Nothing exclusive here, OpenAI says Nvidia is a “preferred” partner, not the only one. They’re still working with other cloud providers and chip companies.Nvidia keeps spending while OpenAI eyes new marketsThis isn’t the only move for Nvidia. Just last week, they dropped $5 billion into Intel to co-develop chips. They also spent $700 million on U.K. startup Nscale, following a similar playbook to what they did with CoreWeave, which went public in March. So while the OpenAI investment is big, it’s just part of Nvidia’s growing stack.For OpenAI, leasing Nvidia’s chips is only part of the financial plan. Executives say they’ll avoid giving up more equity. They see debt as the smarter option to fund these giant facilities.With all this infrastructure being built, OpenAI may not rely on others for cloud services much longer. Their workloads mostly run on Microsoft’s Azure, but that might change. Some execs told CNBC that OpenAI could launch its own commercial cloud offering within two years. Once they’ve got enough compute for themselves, they may flip the switch and start selling it to others too.While Sam was putting the deal together with Jensen, OpenAI’s infrastructure team was in Tokyo again. They were meeting with SoftBank to discuss more financing and help with manufacturing.All of OpenAI’s future infrastructure efforts now live under Stargate. And with Nvidia locked in for at least $100 billion, this might be just the beginning of something even bigger. KEY Difference Wire: the secret tool crypto projects use to get guaranteed media coverage