The world’s most valuable company is done betting big on AI labs.
Nvidia CEO Jensen Huang confirmed Wednesday that the chipmaker’s $30 billion investment in OpenAI “might be the last time” it invests in the ChatGPT maker. Its $10 billion stake in Anthropic “probably will be the last as well.”
Speaking at the Morgan Stanley Technology, Media and Telecom conference, Huang framed the pullback as a consequence of OpenAI’s upcoming IPO. “They’re going to go public toward the end of the year. So, this might be the last time we’ll have the opportunity to invest in a consequential company like this.”
But the explanation raises more questions than it answers.
The $100 Billion Deal That Wasn’t
In September 2025, Nvidia and OpenAI announced an arrangement that seemed to cement their alliance: Nvidia would build 10 gigawatts of AI data centers for OpenAI across 10 installments totaling $100 billion, with OpenAI committing to lease Nvidia chips in return. The first gigawatt was scheduled for late 2026.
That deal is dead.
By January 2026, doubts had surfaced within Nvidia. The Wall Street Journal reported that Huang had privately criticized OpenAI’s “lack of discipline in business approach” and expressed concern about competition from Google and Anthropic. When confronted about the report, Huang called it “nonsense” and insisted he believed in OpenAI.
Then on February 1, Huang clarified that the $100 billion was “never a commitment” but rather “a letter of intent with an opportunity to invest.” The company would proceed “one step at a time.”
The $30 billion that Nvidia did commit became part of OpenAI’s $110 billion funding round in February. But that’s where Nvidia’s checkbook closes.
OpenAI’s IPO Problem
Huang’s stated reason - that OpenAI’s IPO makes further investment impossible - is technically true but conveniently timed.
OpenAI is racing toward a Q4 2026 IPO, having hired finance executives and begun informal talks with Wall Street banks. The company raised $64 billion to date and is currently valued at $500 billion following an October 2025 employee secondary sale. It’s reportedly looking to raise another $100 billion at an $830 billion valuation before going public.
The IPO timing gives Nvidia a graceful exit from a relationship that had clearly soured. Rather than publicly acknowledge tensions, Huang can point to market mechanics: once a company goes public, early-stage strategic investments are no longer possible.
But the question remains: would Nvidia have pumped the remaining $70 billion into OpenAI if the IPO weren’t happening?
The private criticisms suggest not.
The Anthropic Complication
Nvidia’s pullback from Anthropic is harder to explain away with IPO logic.
Anthropic hasn’t announced IPO plans. While analysts expect both OpenAI and Anthropic could go public in 2026, Anthropic is currently embroiled in an existential conflict with the Pentagon that clouds any public market timeline.
Last week, Defense Secretary Pete Hegseth gave Anthropic an ultimatum: loosen restrictions on military use of its Claude AI models or be labeled a “supply chain risk.” The company’s negotiations with the DoD stalled because Anthropic sought assurances that its models wouldn’t be used for autonomous weapons or mass surveillance of Americans.
The Pentagon’s subsequent order for contractors and agencies to cease business with Anthropic puts the company’s $200 million defense contract and broader government access at risk.
When asked about the Anthropic-Pentagon clash in late February, Huang was notably detached.
“I hope that they can work it out, but if it doesn’t get worked out, it’s also not the end of the world,” he said. “Anthropic is not the only AI company in the world, and the Department of Defense is not the only customer.”
For a company with $10 billion invested in Anthropic, that’s remarkably cavalier.
What This Really Means
Nvidia’s retreat from AI lab investments marks the end of a brief era where the chipmaker tried to be both supplier and stakeholder. The strategy made sense when AI labs were scrappy startups dependent on GPU access. It makes less sense now that they’re among the most valuable private companies in history - or soon-to-be public ones.
Huang framed the pullback as prudent conflict management: “We’re pulling back because there are a lot of customers, a lot of other customers, and we don’t want to have a conflict of interest with them.”
But the timing - amid reported tensions with OpenAI and Anthropic’s government crisis - suggests more than neutral portfolio rebalancing.
Nvidia reported $47.5 billion in data center revenue with government and defense customers as a growing segment. If Anthropic becomes radioactive to the Pentagon, Nvidia’s continued investment becomes a liability rather than an asset.
Similarly, OpenAI’s profitability timeline - the company says it won’t turn a profit until 2030 - and Huang’s private criticisms about “lack of discipline” suggest the partnership was straining behind closed doors.
The Bottom Line
Nvidia is cashing out of direct AI lab stakes while they’re worth something. The official explanation about IPOs obscures a messier reality: relationships have soured, government politics have intruded, and being everyone’s GPU supplier is incompatible with being anyone’s strategic investor.
The AI labs will still buy Nvidia chips. They have no choice. But Nvidia is no longer betting its own capital on any of them winning.