Nvidia CEO Jensen Huang has signaled that the chip giant is effectively done investing in OpenAI and Anthropic, even as both AI startups remain among its most important customers and partners.

Speaking at a major technology investment conference in San Francisco, Huang said Nvidia’s recent stakes in the two companies are likely to be its last. He framed the decision as a simple matter of timing: once OpenAI and Anthropic go public, as many on Wall Street expect, the window for Nvidia to participate as a private investor will close.

On its face, that explanation tracks with standard venture practice. Nvidia already profits enormously from selling the GPUs that power both firms’ models and cloud infrastructure. It hardly needs equity upside to justify the relationships.

Yet Nvidia has offered little detail beyond that. Asked to elaborate, the company pointed back to its latest earnings call, where Huang described Nvidia’s startup investments as tools to “expand and deepen” its ecosystem. By that measure, the original bets on OpenAI and Anthropic have already done their job, cementing Nvidia’s chips at the center of the current AI boom.

But the structure of these deals has raised eyebrows. When Nvidia’s prospective OpenAI commitment was first floated at up to $100 billion, critics noted the circularity: Nvidia would buy OpenAI equity, and OpenAI would, in turn, spend roughly the same amount on Nvidia hardware. The final investment, reportedly closer to $30 billion as part of OpenAI’s massive new funding round, underscored how fluid those numbers were.

Nvidia’s Anthropic stake has been no less complicated. After Nvidia announced a multibillion-dollar investment, Anthropic CEO Dario Amodei publicly likened the sale of advanced AI chips to certain Chinese customers to handing nuclear weapons to a rogue state, a pointed rebuke of the very supply chain Anthropic depends on.

Since then, the two AI labs have diverged sharply in their approach to government and military work, with OpenAI embracing a Pentagon deal that Anthropic has condemned. User sentiment has swung accordingly, fueling a surge in downloads of Anthropic’s Claude chatbot.

Nvidia now finds itself holding positions in two rivals moving in opposite ethical and strategic directions. Against that backdrop, Huang’s claim that looming IPOs alone explain Nvidia’s retreat feels incomplete. More plausibly, the company is quietly disentangling itself from a web of alliances that has become politically charged, commercially circular, and increasingly hard to manage.