Was Jensen Huang Really Pushed Out by the "Insiders" Who Invested $10 Billion?

Jensen Huang steps down. The founder of the first five-trillion-dollar company in human history has been forced out. This was bound to be a turning point for the AI era, and we keep trying to find the reason.

Amid the flood of coverage, one sentence has been passed around many times. It comes from a November 12 report in Fortune magazine, from a venture capitalist who spoke on condition of anonymity:

If anyone stands to benefit from this, I think it's Anthropic.

Anthropic is an AI model company whose chatbot is called Claude, one of ChatGPT's strongest rivals. Its relationship with Nvidia is complicated: it is a major Nvidia customer, and Nvidia has also invested in it. But on one issue, the two have been at odds for years: whether to sell high-end chips to Country C. Nvidia wants to sell; Anthropic does not.

Many bloggers on X have also noticed that something seems off about Anthropic. Take @Dream1984Tube, for example. On November 12, he published a long post in English titled "Who Is Really Behind the Hunt for Jensen Huang?" In it, he said:

The New York Times reporter who first reported on Jensen Huang's affair happens to have a long-term source at Anthropic; the Financial Times reporter who first described the romantic relationship happens to have a long record of positive coverage of Anthropic; McGuire, who first changed "forwarding the romantic relationship" to "sleep" on social media, happens to be a former subordinate of an Anthropic executive; the first batch of lawmakers to sign on to the congressional investigation happen to be backed by tens of millions of dollars in Anthropic donations; the first senator to call for Jensen Huang's resignation happens to have held a private closed-door meeting with Dario himself. One coincidence is a coincidence—but a series of coincidences?

We gathered all these reports and checked them against donation records, bills, lobbying filings, and public reporting, attempting a reconstruction. What is listed below are the surface facts that can be found in public records. But the human motives behind the operations—no one knows.

Why a hearing matters so much

On November 4, three House committee chairs co-signed a letter to Jensen Huang: Brian Mast, John Moolenaar, and Crawford. These three names will come up again.

The letter asked Jensen Huang to come in for a closed-door briefing on November 18 to discuss two things: how Nvidia prevents foreign adversaries from gaining access; and his dealings with a certain woman. That is, Alice.

The last line of the letter required Jensen Huang to respond by November 11.

He resigned on November 11.

But it didn't stop there. The day after his resignation, November 12, Senate Banking Committee Chairman Tim Scott sent a letter inviting Jensen Huang to appear at a public hearing on December 3. The committee that oversees export controls in the Senate is precisely this one.

From a closed-door briefing to a public hearing—the weight is very different.

In the United States, summoning the head of a tech company to a public hearing on Capitol Hill is itself a ritual. Lawmakers sit higher than he does, the lights shine brighter on him, C-SPAN, the channel dedicated to covering Congress, is filming, and stenographers type every word he says into the congressional record. He must be sworn in; lying under oath is a federal felony punishable by up to five years in prison. If he doesn't show up, first comes a subpoena, then contempt of Congress.

Senate Banking Committee hearing

The more common cost is another kind. Every sentence you say can be cut into fifteen seconds and played on loop for the next ten years. In 2018, Zuckerberg sat on Capitol Hill for two days; the most famous image was the cushion under his rear, and few people remember what he actually answered.

A hearing itself doesn't write law, but it can determine whether a law gets written. In March 2023, TikTok CEO Shou Zi Chew sat in the House for several hours. Thirteen months later, Congress passed a bill requiring TikTok to either sell or be banned.

So a hearing is a watershed in American politics: from that moment on, a company's matter formally becomes the nation's matter.

Jensen Huang has already resigned, but the hearing will proceed as scheduled. What Congress wants to ask is far bigger than one man.

So to be able to send Jensen Huang—Trump's right-hand man, the only person comparable to Musk in sitting on Trump's private plane—to a hearing is no easy feat.

And Anthropic's name keeps coming up.

One dinner table, three clashes

In May 2022, San Francisco, at an upscale Chinese restaurant. This was the first meeting between Jensen Huang and Anthropic founder Dario.

According to an upcoming book by New York Times tech reporter Kevin Roose, things went wrong before dinner even started. Anthropic co-founder Tom Brown brought a spreadsheet to Jensen Huang to negotiate on price, saying that Google's own AI chip, the TPU, was more cost-effective than Nvidia's. Jensen Huang blew up on the spot, calling him a "bean counter"—meaning a bookkeeper who only knows how to pinch numbers—and threatened to cancel the dinner.

At the table, Jensen Huang repeatedly talked about building the world's largest data center. Dario later muttered privately that he was "a bit like Trump."

After that, the two began to clash.

April 30, 2025. Anthropic published a blog post about how Country C gets chips: hidden inside fake pregnant bellies, smuggled in with live lobsters. The next day, an Nvidia spokesperson responded, saying it was a fabricated story. He also added a jab: "Country C has half the world's AI researchers."

January 20, 2026, Davos, Switzerland, the annual gathering of global political and business elites. In an interview with Bloomberg, Dario said that selling Nvidia's H200 chips to Country C was "like selling nuclear weapons to North Korea and bragging about it."

On April 15, Jensen Huang used a word on the tech podcast Dwarkesh: absurd. Then he said the most stinging line: giving up the Country C market would harm America's national security and technological leadership—all for the benefit of one company.

He didn't name names, but everyone could tell which company he meant.

By the fall of 2026, the two met again at the White House. On September 27, Trump had a meal with Dario, and Trump himself later confirmed the meeting. Two days later, the White House hosted a lunch on "superintelligence." According to The Wall Street Journal, citing people familiar with the matter, several tech executives including Jensen Huang asked Dario to his face: why do you describe the risks of AI so alarmingly in public?

Because Dario often says, "AI could destroy humanity."

But the two companies don't entirely avoid cooperation. After all, Nvidia has the most powerful chips at present. On November 18, 2025, Jensen Huang, Dario, and Microsoft CEO Nadella held a joint press conference to announce a three-way partnership: Anthropic would buy $30 billion of Microsoft cloud computing power, using up to 1 gigawatt of Nvidia systems—roughly equivalent to the output of one unit of a large nuclear power plant; Nvidia committed to investing up to $10 billion in Anthropic, and Microsoft up to $5 billion.

Nvidia is Anthropic's shareholder, its supplier, and also the biggest stumbling block to its efforts to restrict Country C's AI technology.

November 2025 three-way partnership press conference

Where the forty million dollars went

On February 12, 2026, Anthropic announced a $20 million donation to an organization called Public First Action. It is a Washington-based organization promoting AI policy, and one of its co-founders is former Republican congressman Chris Stewart.

Anthropic's announcement was quite candid, listing four policy positions, and the third was chips:

Support smart export controls on AI chips so that America stays ahead of authoritarian adversaries.

More than five months later, on July 21, another $20 million. This time it was more direct: we support tightening export controls on advanced chips and semiconductor manufacturing equipment.

Forty million dollars in total—how much is that? By Anthropic's own lobbying spending, about two million a quarter, forty million would be enough to lobby in Washington for five full years.

On the same day the donation announcement was issued, Public First Action ran its first ad since its founding, thanking Senator Marsha Blackburn in Tennessee for her work on children's online safety and AI legislation. On the same day, an affiliated campaign organization ran another ad in Nebraska supporting Senator Pete Ricketts, with the theme "keeping advanced AI chips out of adversaries' hands."

A company gives money to a political organization, and that organization starts running ads the same day.

Besides ads, where else did that $40 million go?

First, let's look at two concepts.

Public First Action is legally a "501(c)(4)," which can be understood as a kind of organization that is only allowed to talk about policy. It can lobby, it can do public outreach, it can spend a lot of money, but it cannot make influencing elections its primary purpose; the advantage is that it doesn't have to disclose who gave the money.

The other kind is called a "super PAC," an account specifically for fighting election campaigns. Its sole purpose is to influence elections, run ads for candidates, or take candidates down. There is no limit on its spending, at the cost that every incoming dollar must be disclosed.

One is not allowed to touch elections but is opaque; the other specifically touches elections but is fully transparent. Anthropic's forty million went into the first kind.

In its July announcement, Anthropic said: the two donations are to be used only for public education and policy, and may not be used to influence the election of any federal, state, or local candidate.

But twelve days after the first money arrived, on February 24, Public First Action transferred $1 million to a super PAC called "Public First." Later, according to a summary of Federal Election Commission filings, it also gradually transferred about $7.9 million to another super PAC, "Defending Our Values PAC."

No one can prove that the $1 million transferred out came from Anthropic's $20 million. But by November, Public First Action had raised about $80 million in total, and Anthropic alone accounted for half. To put it another way, for every $1 it sends toward the election side, about 50 cents comes from Anthropic.

Now look at the people at Anthropic.

On May 4, 2026, Dario personally donated $1 million to the super PAC "Public First"; on the form, the employer field was filled in as Anthropic and the occupation field as CEO. In June, five more Anthropic employees donated to the same super PAC, totaling more than $2.15 million.

Counting Dario, this super PAC received more than $3.15 million from people at Anthropic, more than sixty percent of its total fundraising for the whole cycle (about $4.9 million).

The company says its money doesn't touch elections, yet its CEO and employees became the largest donors to the same campaign account.

Anthropic also has its own political donation account, called AnthroPAC; the money is almost all $5,000 checks written by employees, totaling just over $400,000. By November, it had donated more than $195,000 to candidates of both parties, with a list that was almost half Democrats and half Republicans. Mast received $2,500 and Crawford $1,000—Mast and Crawford being two of the three who wrote to Jensen Huang.

On July 17, Public First Action, which had received Anthropic's $40 million, announced a $15 million project called the "Conservative AI Consensus." It opened by spending more than $7 million on ads promoting four Republican lawmakers; over the following month, it would also gradually build momentum for another 12 Republican lawmakers, and the list included Mast and Moolenaar.

More than three months later, it was precisely Mast and Moolenaar who co-signed that letter to Jensen Huang.

The purpose of this $15 million project was stated plainly: "Conservatives built this country's technological advantage, and it will not be handed over to a handful of companies that answer to no one."

But that's not all—these are not the only details.

Further upstream

The institutions that advise Washington are called think tanks. When lawmakers write bills and letters, many of the talking points and data come from their reports.

On the question of "how America should manage technology toward Country C," three think tanks matter.

One is Georgetown University's Center for Security and Emerging Technology, or CSET. When it was founded in 2019, Georgetown's press release headline was "America's largest AI and policy center lands at Georgetown." The money came from a philanthropic fund called Open Philanthropy: $55 million initially, then two more grants, totaling $102 million. The center has fifty-one people, which works out to about two million dollars per person.

Another is RAND Corporation, one of America's oldest think tanks, long a contractor for the Pentagon. From 2020 to 2026, it received 17 grants from the same fund, totaling $86.956 million.

The third is the Washington-based Institute for Progress, or IFP, which received three grants totaling $21.457 million.

The money for all three came from the same pocket.

That pocket changed its name on November 18, 2025, and is now called Coefficient Giving. Its founding partners are the foundation of a couple: Dustin Moskovitz and his wife Cari Tuna. Moskovitz is one of Facebook's co-founders and one of Anthropic's earliest investors. According to Forbes in February 2026, citing Cari Tuna, the couple transferred their Anthropic shares into a nonprofit in early 2025, and those shares were worth about $500 million at the time.

The letter to Jensen Huang is precisely the product of this research pipeline over the years. And at the source of this pipeline stands an Anthropic investor.

Who benefits

All of the above is surface appearance. No single operation can show that Jensen Huang's downfall was engineered single-handedly by Anthropic, but what the people behind each operation were really thinking—no one knows.

Anthropic has what is now widely recognized as the best model, and among all of Silicon Valley's top frontier AI labs, it is also the most aggressive in banning accounts from China. Search Claude on any social media platform, and guides on how to use the Claude model domestically will appear, but the end result of every guide is basically a ban. Anthropic even recently updated its account-ban policy specifically targeting China.

Of course, there is also the opposing view.

Some argue that if Anthropic really wanted to bring down Jensen Huang, waiting until after its own IPO to make a move would be more advantageous. Jensen Huang's departure casts a dark cloud over the entire AI industry, and Anthropic's own valuation is threatened as a result.

This is also the most abstract part.

In that three-way partnership, Nvidia committed to investing up to $10 billion in Anthropic. According to Fortune, Nvidia also plans to subscribe to roughly $10 billion of stock when Anthropic goes public.

They don't believe that Anthropic, valued at $1.4 trillion, would sabotage its own major shareholder before its IPO.

And no one will ever know the truth.

Original link