Nvidia has ended its investments in OpenAI and Anthropic. Jensen Huang confirmed this withdrawal on March 4, 2026, during the Morgan Stanley Technology, Media and Telecom conference in San Francisco.
A Less-Than-Convincing Official Justification
Nvidia’s CEO offered the following reasoning: With OpenAI and Anthropic preparing for their IPO this year, the window for private equity fundraising is automatically closing. « I will likely no longer have the opportunity to invest in a substantial company like this, » stated Jensen Huang, Nvidia’s CEO, at the Morgan Stanley Technology, Media and Telecom conference.
The issue is that this explanation doesn’t hold up well under scrutiny. In late-stage venture capital, institutional investors regularly invest right up until the eve of an IPO to capture a final valuation tranche. Therefore, Nvidia was under no structural obligation to withdraw at this precise moment.
When questioned by TechCrunch on March 4, 2026, a Nvidia spokesperson referred to an excerpt from the fourth-quarter earnings call, where Huang had explained that all of the group’s investments are « focused very squarely, strategically on expanding and deepening our ecosystem reach. »
From 100 Billion to 30 Billion: The Shrunken Commitment
To gauge the scale of the retreat, we must go back to September 2025, when Nvidia announced a potential investment of $100 billion in OpenAI. Immediately after the announcement, Professor Michael Cusumano of the MIT Sloan School of Management described the deal as « a kind of zero-sum game » in a statement reported by the Financial Times, noting that OpenAI would buy $100 billion or more in Nvidia chips, while Nvidia invested $100 billion in OpenAI.
The final outcome is very different from the initial promise. As part of the $110 billion funding round closed by OpenAI at the end of February 2026, Nvidia ultimately participated with $30 billion, alongside Amazon ($50 billion) and SoftBank ($30 billion), according to an OpenAI press release reported by CNBC. OpenAI’s pre-money valuation stood at $730 billion in this round, up from $500 billion in a secondary transaction in October 2025, according to CNBC. The initial $100 billion commitment was therefore never fully honored, and Nvidia will pay its $30 billion in three installments, on April 1, July 1, and October 1, 2026, according to Le Magit.
Anthropic: From Investment to Political Conflict
The relationship between Nvidia and Anthropic is even more complex. In November 2025, Nvidia announced a $10 billion investment in the startup. Just two months after this announcement, Dario Amodei, CEO of Anthropic, spoke publicly in Davos and, without naming Nvidia directly, compared the sale of high-performance AI processors to certain Chinese clients to « selling nuclear weapons to North Korea, » according to TechCrunch. This was a particularly harsh public statement for a partner that had just committed $10 billion.
However, the situation then took a completely different turn. At the end of February 2026, the Trump administration blacklisted Anthropic, prohibiting federal agencies and military contractors from using its technologies after Anthropic refused to authorize the use of its models for autonomous weaponry or large-scale domestic surveillance, according to NPR. Secretary of Defense Pete Hegseth officially designated Anthropic as a « national security supply chain risk, » according to NPR.
Two Companies Pulling in Opposite Directions
What makes Nvidia’s position particularly delicate is that its two stakes, within the space of a few days, have taken openly contradictory paths. Hours after the announcement of Anthropic’s blacklisting, OpenAI concluded an agreement with the Pentagon, a decision that Anthropic described as « deceitful, » according to TechCrunch. The impact on the consumer market was immediate: Claude, Anthropic’s model, went from 100th place to first place in the ranking of free apps on the US App Store in less than 24 hours, according to Sensor Tower data cited by TechCrunch.
Nvidia thus finds itself a shareholder in two companies in open opposition on issues of AI governance and cooperation with US authorities. Jensen Huang had certainly dismissed any hypothesis of personal tensions with OpenAI, calling such theories « nonsense » at the Morgan Stanley conference. However, the accumulation of friction in just a few weeks makes the official explanation difficult to accept as is.
Nvidia’s publicly stated reason remains the IPO argument, but there’s no indication that this justification is complete. What is certain is that the circular nature of cross-investments, the political turbulence surrounding Anthropic, and the growing strategic divergence between the two major players in generative AI have made the situation uncomfortable enough to warrant a withdrawal. To understand what comes next, we will need to monitor how Nvidia manages its commercial partnerships with OpenAI and Anthropic, regardless of equity stakes, as both companies advance toward their IPOs.