Nvidia Pulls Back From OpenAI and Anthropic Amid Pentagon Mess

Jensen Huang says Nvidia is pulling back from OpenAI and Anthropic because they’re going public soon.

That’s the official story.

At the Morgan Stanley Technology, Media and Telecom conference, Nvidia’s CEO explained that once these companies IPO later this year, “the opportunity to invest closes.”

But here’s the problem: that explanation doesn’t make sense.

Late-stage private investors pile into companies right up until the eve of their public debut all the time. The IPO window doesn’t suddenly slam shut. If Nvidia wanted to invest more, there’s no structural reason they couldn’t.

So what’s really going on?

This is an exit from a situation that got messy fast.

And the mess has a name: the Pentagon.

Days ago, the Trump administration blacklisted Anthropic after the company refused to allow its AI models to be used for autonomous weapons or mass domestic surveillance. Hours later, OpenAI struck a deal with the Pentagon to provide AI for “all lawful means”—including the surveillance Anthropic rejected.

Now Nvidia holds stakes in two companies pulling in completely opposite directions. One is the principled holdout. The other is the government’s new favorite contractor.

And Jensen Huang? He’s running for the exit.

Let me explain why this isn’t about IPO timing—and what it actually reveals about the AI investment bubble, the Pentagon’s influence, and whether users genuinely care about surveillance.

The $100 Billion That Became $30 Billion: What Changed?

When Nvidia first announced it would invest up to $100 billion in OpenAI last September, an MIT Sloan professor called it “kind of a wash.”

Here’s what he meant: Nvidia invests $100 billion in OpenAI stock. OpenAI uses that money to buy $100 billion in Nvidia chips.

It’s circular. Nvidia gives OpenAI money. OpenAI gives the money back by buying Nvidia’s products. Both companies’ valuations go up. Everyone’s balance sheet looks great.

But it’s also financial engineering that could be creating an investment bubble.

And apparently, someone at Nvidia noticed.

Because when Nvidia finalized the investment last week as part of OpenAI’s $110 billion funding round, the commitment had shrunk to $30 billion—well short of that earlier $100 billion pledge.

So what changed between September and last week?

Two things:

1. OpenAI’s user growth is declining.

ChatGPT isn’t growing at the pace it used to. The hype cycle has plateaued. Competitors like Claude and open-source models are eating into market share.

If OpenAI’s revenue growth is slowing, and if users are starting to churn, then pouring $100 billion into the company looks less like a strategic investment and more like propping up an overvalued asset.

2. OpenAI took the Pentagon deal.

Sam Altman agreed to provide ChatGPT for “all lawful means” right after Anthropic refused. That decision triggered immediate backlash. A “Cancel ChatGPT” movement exploded on Reddit. Claude overtook ChatGPT in the App Store rankings within 24 hours.

And Nvidia? Nvidia suddenly had a problem.

If OpenAI’s Pentagon deal is driving users away, if it’s damaging the brand, if it’s creating a public relations nightmare—why would Nvidia pour more money into that?

The $100 billion commitment was based on OpenAI being the dominant player with unstoppable momentum. But that’s not the reality anymore.

So the investment shrunk from $100 billion to $30 billion. And now Huang is signaling there won’t be any more.

That’s not IPO timing. That’s cutting losses.

Nvidia’s Awkward Position: Invested in Both Sides of a War

Here’s where things get really messy for Nvidia.

The company now holds stakes in:

Anthropic – blacklisted by the Pentagon for refusing to enable surveillance and autonomous weapons
OpenAI – embraced by the Pentagon for agreeing to provide AI for “all lawful means”

These companies aren’t just competitors. They’re on opposite sides of a fundamental question: Should AI companies set red lines on how their technology is used, or should they trust governments to decide?

Anthropic said no to mass surveillance and autonomous weapons. The Pentagon designated them a supply chain risk.

OpenAI said yes to “all lawful means.” The Pentagon signed a deal.

And Nvidia? Nvidia is invested in both.

Is this just awkward, or does it create real problems?

Here’s my take: As long as the balance sheet balances out and the P&L shows profit, Nvidia doesn’t care.

Nvidia isn’t making moral judgments. They’re selling chips. If OpenAI wants to work with the Pentagon, fine—they’ll buy more GPUs for government contracts. If Anthropic refuses, fine—they’ll buy GPUs for commercial applications.

Nvidia’s business model works regardless of which side wins this debate.

But here’s the problem: customers and partners are getting dragged into this fight.

If you’re a company deciding between Claude and ChatGPT, you’re now also deciding between a provider that’s blacklisted by the U.S. government and one that’s working directly with military intelligence.

If you’re a developer building on these platforms, you’re picking a side in the surveillance debate whether you intended to or not.

And if you’re Nvidia, holding stakes in both, you’re trying to stay neutral in a conflict where neutrality is becoming impossible.

That’s uncomfortable. And uncomfortable situations are exactly what Jensen Huang is trying to exit.

Dario Amodei Called Nvidia’s China Sales “Selling Nuclear Weapons to North Korea”

Two months after Nvidia announced a $10 billion investment in Anthropic, Dario Amodei took the stage at Davos and—without naming Nvidia directly—compared U.S. chip companies selling high-performance AI processors to approved Chinese customers to “selling nuclear weapons to North Korea.”

Ouch.

That’s not exactly the kind of thing you say about a company that just gave you $10 billion.

So would that comment make you pull your investment if you were Nvidia?

Honestly? No.

Here’s why: Business is business.

Amodei was making a geopolitical point about AI proliferation risks, not attacking Nvidia specifically. And Nvidia knows that. They’re not going to kill a strategic investment because a CEO made an uncomfortable comparison at a conference.

Plus, Jensen Huang is of Asian origin. He understands the geopolitical complexities of selling advanced technology in a world where China, the U.S., and other powers are racing for AI dominance. He’s not naïve about these tensions.

If anything, Amodei’s comment just confirms what Nvidia already knows: AI companies are getting pulled into geopolitical conflicts whether they want to be or not.

And that makes them risky investments.

Not because of what they say at conferences, but because governments are now forcing AI companies to pick sides.

Anthropic refused to work with the Pentagon on surveillance. They got blacklisted.

OpenAI agreed to work with the Pentagon. They got backlash from users and developers.

There’s no safe position anymore. And that’s exactly the kind of environment where Nvidia would rather be selling chips than holding equity stakes.

Claude Overtook ChatGPT in 24 Hours—Users Actually Care About Privacy

Here’s something interesting: within 24 hours of the Pentagon announcements, Anthropic’s Claude shot to the top of the free-app rankings on Apple’s U.S. App Store, overtaking ChatGPT.

At the end of January, Claude was outside the top 100, according to Sensor Tower data.

Now it’s #1.

So does that prove users care about surveillance concerns, or is it just temporary virtue signaling?

I think people genuinely care about privacy.

And here’s the thing: a lot of people have experienced serious hallucinations with AI at one point or another.

Everyone who uses AI regularly has had moments where the model just gets things spectacularly wrong. It hallucinates facts. It misunderstands context. It confidently produces outputs that are complete nonsense.

So when you tell users that the same AI that hallucinates might also be analyzing their data for government surveillance, they take that seriously.

It’s not virtue signaling. It’s a rational response to real risk.

If ChatGPT is now feeding data to Pentagon systems for “all lawful means”—which includes mass surveillance under certain Patriot Act provisions—then users who care about privacy have a legitimate reason to switch platforms.

And they did. En masse. Within 24 hours.

That’s not a temporary blip. That’s a structural shift.

Anthropic’s principled stance wasn’t just morally defensible. It was also good for business.

Meanwhile, OpenAI’s Pentagon deal—despite bringing in government contracts and funding—cost them their #1 ranking and triggered a user exodus.

And Nvidia? Nvidia watched this unfold and realized: one of our portfolio companies just lost significant market share because of a political decision.

That’s exactly the kind of risk that makes late-stage investors nervous.

What Nvidia’s Balance Sheet Really Cares About

Let me be clear about what drives Nvidia’s decisions:

As long as the balance sheet balances out and the P&L shows profit, they’re good. Otherwise, they’re not.

Nvidia isn’t making ethical judgments about surveillance or autonomous weapons. They’re not taking sides in the Pentagon debate. They’re selling chips and managing their investment portfolio.

And right now, the portfolio is getting messy.

Anthropic is blacklisted by the U.S. government but gaining market share with users who care about privacy. Their commercial business is growing, but they’re cut off from lucrative government contracts.

OpenAI has Pentagon contracts but is losing users and facing a “Cancel ChatGPT” movement. Their government revenue is up, but their consumer brand is taking damage.

From Nvidia’s perspective, both companies are now riskier investments than they were six months ago.

Anthropic’s blacklisting limits their total addressable market. OpenAI’s Pentagon deal creates reputational risk and user churn.

Neither outcome is ideal for an investor.

So what does Nvidia do?

Exit.

Not immediately. Not in a panic. But strategically, over time, by declining to invest more and letting their stakes dilute as these companies raise additional funding.

Huang’s explanation—that IPOs close the investment window—is just cover. The real reason is simpler: the situation got complicated, and Nvidia would rather sell chips to both companies than own equity in either one.

Is Huang Making the Smart Move, or Missing the Long-Term Opportunity?

So is Jensen Huang making the right call by pulling back?

In the short term? Absolutely.

Nvidia doesn’t need to own equity stakes in AI companies to make money. They’re the picks-and-shovels provider. Whether OpenAI, Anthropic, Google, Meta, or anyone else wins the AI race, they all need Nvidia’s chips.

Owning stakes in specific companies creates concentration risk. It forces Nvidia to care about which AI provider wins. It exposes them to political conflicts, reputational damage, and regulatory uncertainty.

Selling chips to everyone is a much safer business model.

But in the long term?

If one of these companies becomes the dominant AI platform—the equivalent of Google in search or Amazon in cloud—then Nvidia’s decision to exit early could look short-sighted.

Early investors in Google and Amazon made generational returns. If OpenAI or Anthropic achieves similar dominance, Nvidia’s stakes would be worth far more than any chip sales.

But here’s the key question: which company will dominate?

Right now, it’s unclear. OpenAI has scale and government contracts, but they’re bleeding users. Anthropic has momentum and user trust, but they’re blacklisted by a major customer.

And there are dozens of other competitors—Google, Meta, Mistral, open-source models—all fighting for market share.

In that environment, Huang’s decision makes sense: exit the messy equity game and focus on selling chips to whoever wins.

It’s the smart, conservative play.

But it’s also an admission that Nvidia doesn’t know how this shakes out.

And when the world’s leading AI infrastructure company can’t predict which AI platform will dominate, that tells you everything about how uncertain and volatile this market really is.

What This Really Means

Jensen Huang is pulling back from OpenAI and Anthropic. His official explanation—that IPO timing closes the investment window—doesn’t hold up under scrutiny.

Here’s what’s actually happening:

This is an exit from a messy situation. The Pentagon forced AI companies to pick sides. Anthropic refused and got blacklisted. OpenAI complied and triggered user backlash. Nvidia is invested in both and would rather not be.

The $100 billion became $30 billion for a reason. OpenAI’s user growth is declining, and the Pentagon deal accelerated the exodus. Nvidia cut the investment because the risk-reward calculation changed.

Business is business. Amodei’s “nuclear weapons” comment wouldn’t make Nvidia pull out. Huang understands geopolitical tensions. But forced political alignment? That’s a different risk entirely.

Nvidia only cares about the balance sheet. As long as P&L shows profit, they’re good. But when portfolio companies get blacklisted or lose users over political decisions, that’s bad for returns.

Users genuinely care about privacy. Claude overtaking ChatGPT in 24 hours isn’t virtue signaling. People have experienced AI hallucinations and don’t trust these systems with surveillance. That’s rational, not performative.

The smart move is to sell chips, not own equity. Whether OpenAI or Anthropic wins, they both need Nvidia’s GPUs. Owning stakes creates risk. Selling picks-and-shovels is safer.

But it’s also an admission of uncertainty. When Nvidia—the company best positioned to see where AI is going—can’t predict who wins, that tells you how volatile and unpredictable this market is.

Jensen Huang’s pullback isn’t about IPO timing.

It’s about getting out before the situation gets worse.

And given how fast things are moving—Pentagon deals, user revolts, government blacklists—that’s probably the smartest decision he could make.