It was one of the biggest deals in the history of the technology industry. In September 2025, Nvidia and OpenAI announced a partnership worth up to $100 billion. Together, they planned to build data centers with a capacity of at least 10 gigawatts. Analysts described it as an unprecedented deal. Then March 2026 arrived, and at a conference in San Francisco, Jensen Huang calmly said that the $100 billion investment was no longer going to happen. In fact, that the $30 billion investment was probably the last one. So what is really behind it?
Conference in San Francisco: Huang speaks, the world listens
The Morgan Stanley Technology, Media & Telecom Conference in downtown San Francisco. Jensen Huang, Nvidia's CEO, stands before a room full of investors and analysts and says things no one expected.
"The $100 billion investment in OpenAI probably isn't going to happen," Huang declared. The reason? OpenAI is preparing to go public. "They're heading for an IPO toward the end of the year," he added. Then came the statement that the $30 billion investment Nvidia had just finalized as part of OpenAI's $110 billion funding round "may be the last opportunity to invest in such a pivotal company."
Huang did not stop with OpenAI. He also mentioned Anthropic, the maker of the Claude chatbot, in which Nvidia invested $10 billion in November 2025. That investment is also unlikely to be repeated. Anthropic is also heading for the stock market.
From $100 billion to $30 billion
Let's go back for a moment. The original September 2025 deal was based on a different principle. Nvidia was supposed to invest gradually as new supercomputing centers were built. The total amount could have reached as much as $100 billion. But then the first signs emerged that something was not working.
In November 2025, Nvidia suggested for the first time in its quarterly report that the deal "may not be completed." In January 2026, The Wall Street Journal reported that the entire deal was "on hold." Nvidia repeated similar language in its February 2026 report.
Then came OpenAI's $110 billion funding round. Nvidia participated with $30 billion, Amazon added $50 billion, and SoftBank another $30 billion. That is significantly less than the original commitment. More importantly, this investment is not tied to any infrastructure deployment milestones. It is a direct equity stake in the company. In return, OpenAI offered Nvidia 3 gigawatts of dedicated capacity for inference and 2 gigawatts for model training on Vera Rubin systems. The deal therefore still makes sense. But it is a different deal from the one originally planned.
Circular deals that no one wants to call a bubble
Professor Michael Cusumano of MIT Sloan described it precisely. When Nvidia announced its $100 billion investment in OpenAI in September 2025, he characterized the entire structure as "a kind of quid pro quo." Nvidia invests in OpenAI, and OpenAI uses the money to buy Nvidia chips. The money essentially moves in a circle.
Hedge fund manager George Noble described the entire structure as "borderline criminal" and predicted that "this cannot end well." According to him, it is hidden vendor financing disguised as venture capital. "Amazon and Nvidia are essentially paying OpenAI to buy their own products," Noble said.
Growing concerns that such deals are inflating an investment bubble may have been one reason why the original $100 billion commitment shrank to $30 billion. Huang himself dismissed the theory of bad blood between Nvidia and OpenAI as "nonsense." But the numbers speak for themselves.
A $10 billion investment in Anthropic
Nvidia's relationship with Anthropic was more complicated from the outset. Nvidia announced the $10 billion investment in November 2025 together with Microsoft. But just two months later, in January 2026 at the World Economic Forum in Davos, Anthropic CEO Dario Amodei made a statement that must have sent a chill through Nvidia. Without naming Nvidia directly, he compared the sale of powerful AI chips to approved Chinese customers to "selling nuclear weapons to North Korea." Ouch. Those are pretty strong words directed at a company from which you have just accepted $10 billion.
And that was not all. Shortly afterward, the Trump administration blacklisted Anthropic as a supply-chain security risk because the company refused to allow its models to be used for autonomous weapons or mass domestic surveillance. Federal agencies and military contractors thus lost access to Anthropic's technology.
Meanwhile, OpenAI quickly struck its own deal with the Pentagon. Anthropic called it "outright lies." Users responded by uninstalling ChatGPT, with the number of uninstalls rising by 295%. Anthropic's Claude shot to the top of the US Apple Store app rankings, even though it had still been outside the top 100 at the end of January.
Nvidia thus holds stakes in two companies that are currently moving in completely opposite directions.
What is behind Huang's statement?
Huang's explanation sounds logical: once a company goes public, the window for private investment closes. But investment experts point out that this is not entirely how it works. Companies routinely invest in businesses shortly before their IPOs, seeking one last opportunity to profit. The website TechCrunch said: Huang's explanation leaves more questions than answers. What seems more likely is that Nvidia is looking for a way out of a situation that became complicated faster than anyone expected.
Nvidia makes enormous amounts of money selling chips. Nvidia's GPUs power the models of both OpenAI and Anthropic. So why would it also need to invest in these companies? Huang himself suggested the answer: "The revenue will come on its own." Nvidia is the biggest winner of the entire AI boom because it makes the hardware without which no large language model can operate.
The AI industry's shift from model training to inference—that is, the rapid processing of user queries—is also changing demand for chips. Nvidia is therefore developing a new chip specifically for inference, and OpenAI is expected to be one of its largest customers. The business relationship therefore continues. Just without multibillion-dollar equity investments.
Could it be said that Huang made a smart move? Perhaps. Nvidia retains a stake in OpenAI at a reasonable price, maintains business relationships with both AI giants, and at the same time extricates itself from the increasingly complex political and ethical chess game between OpenAI, Anthropic, and the US government. And it is doing all this with a smile and the statement that "an IPO closes the door to investors."
Sources: finance.yahoo.com and businessinsider.com
