Jensen Huang made a striking announcement this week at the Morgan Stanley Technology, Media and Telecom conference in San Francisco. Nvidia’s investments in OpenAI and Anthropic, he said, are probably done. The company won’t be putting more money into either one once they go public, which is expected to happen later this year.
On the surface, it sounds reasonable. Companies stop being investment targets once they’re public, right? But something doesn’t add up here.
The $100 Billion Promise That Shrank to $30 Billion
Last September, Nvidia announced plans to invest up to $100 billion in OpenAI. It was a massive vote of confidence, the kind of headline that makes people sit up and pay attention. Then in January, when OpenAI actually closed its funding round at $110 billion, Nvidia’s actual contribution was $30 billion. Two-thirds smaller than the original announcement.
That’s not the trajectory you’d expect if everything was running smoothly.
The math on this whole thing was already weird to begin with. MIT Sloan professor Michael Cusumano pointed out last year that the arrangement felt “kind of a wash” because Nvidia was essentially investing in OpenAI while OpenAI turned around and committed to buying massive amounts of Nvidia chips. It’s a circular relationship that starts looking less like strategic investment and more like financial theater when you squint at it long enough.
And people were starting to notice. As concerns grew about whether these kinds of deals were inflating an AI investment bubble, suddenly the commitment got smaller. Much smaller.
The Anthropic Complication
Then there’s Anthropic, where things got genuinely messy.
Nvidia invested $10 billion in Anthropic back in November. Two months later, Anthropic’s CEO Dario Amodei was on stage at Davos making thinly veiled shots at Nvidia, comparing U.S. chip companies selling advanced AI processors to Chinese customers with “selling nuclear weapons to North Korea.”
It was a pretty clear signal that the two companies weren’t exactly on the same page about how AI should be deployed globally.
But that was just the warmup act. Last week, things escalated dramatically when the Trump administration blacklisted Anthropic, barring federal agencies and military contractors from using its technology. The reason? Anthropic refused to allow its models to be weaponized or used for mass surveillance.
Hours later, OpenAI announced a deal with the Pentagon.
Within 24 hours, Claude had shot to the top of Apple’s app store, overtaking ChatGPT. The public had clearly picked a side.
An Exit Dressed Up as Good Timing
So here’s where Nvidia finds itself. It’s holding stakes in two companies that are pulling in completely different directions right now, making very different bets about what AI should be used for and who it should serve. They’re becoming rivals in ways that matter politically and strategically, not just commercially.
Huang’s explanation that the IPO window closes the investment opportunity is… well, it’s not wrong exactly. But it also doesn’t tell the real story. Late-stage private investing happens right up until a company goes public all the time. Companies pile in at the last second searching for more upside. If Nvidia really wanted to keep investing in these companies, IPO status wouldn’t be the deciding factor.
What’s more likely happening is that Nvidia is stepping back from a situation that got complicated faster than anyone anticipated. The company is minting money selling chips to everyone. Why stick around and choose sides in a growing rift between two major AI companies about ethics, weapons, and government involvement?
Maybe Huang saw this coming. Maybe he didn’t. But his stated reason for pulling back sounds more like convenient cover than the actual story.
The real question isn’t whether Nvidia will invest more. It’s whether OpenAI and Anthropic can both claim to be the future of AI when they’re standing on opposite sides of a line that matters increasingly more.
