In Anthropic's doomsday narrative, there is no room for a Jensen Huang.

On November 12, 2026, Jensen Huang, who had led Nvidia for 33 years, resigned as CEO, president, and director. Seven days later, on November 19, Anthropic will list on Nasdaq under the ticker ANT, planning to raise up to $90.4 billion.

One exits, one rings the bell—just one week apart.

On the surface, what brought down Silicon Valley's toughest founder was a series of personal conduct scandals that erupted in quick succession over two weeks.

But on the day Huang handed over power, X tech commentator Liang Yimeng (@Dream1984Tube) gathered several key figures who had fueled public opinion over the past half month into a single chart, uncovering several intersections from public federal lobbying records and resumes.

Relationship chart compiled by Liang Yimeng | Image source: X

In early November, after the Financial Times further disclosed the alleged romantic relationship, Chris McGuire, a scholar at the Council on Foreign Relations, was the first to use "sleeping with" on X, and the post subsequently spread widely in Washington political circles. His former boss at the Biden administration's National Security Council, Tarun Chhabra, was poached by Anthropic in early 2025 to serve as head of national security policy.

The person who was overseeing Huang's classified memo within the government in spring 2026 was Secretary of State Marco Rubio. In mid-2025, Anthropic hired lobbying firm Continental Strategy to advocate for it, and its lead, Alberto Martinez, had served as Rubio's chief of staff for many years.

Brian Mast and John Moolenaar, who took the lead on Capitol Hill in demanding that Huang explain the situation, had previously been included in Public First Action's support advertising program. This advocacy organization, which has an affiliated super PAC, has received a cumulative $40 million in donations from Anthropic.

Republican Senator Kevin Cramer, who publicly called for Huang's resignation on November 10, Eastern Time, had previously received a $1,000 donation from AnthroPAC. In February of the same year, Dario Amodei met with members of the Senate Banking Committee, on which Cramer sits, to push for tighter chip export restrictions.

Liang Yimeng also noted that Ana Swanson, the New York Times reporter who has been continuously covering the Jensen Huang story, had previously cited the views of Anthropic co-founder and former Bloomberg journalist Jack Clark multiple times in her reporting on AI and chips.

Too many coincidences collided at once.

Currently, there is no direct evidence that Dario Amodei orchestrated this exposure. Moreover, attributing the turbulence of a $5 trillion empire to behind-the-scenes manipulation by a single company both overestimates the precision of political lobbying and underestimates the fatal severity of Jensen Huang's own governance vulnerabilities.

But Dongcha Beating wants to probe another question.

When Jensen Huang slipped, why could Washington politicians, national security scholars, mainstream media, and Wall Street all operate along the same track without anyone needing to give an order?

A track can run automatically never because of a single secret meeting, but because a group of people share the same worldview and have the same reaction to the same thing.

In 2023, AI ethics scholar Timnit Gebru and philosopher Émile P. Torres collectively termed the ideological current entrenched at the top of Silicon Valley AI labs as "TESCREAL," which nests transhumanism, extropianism, singularitarianism, cosmism, rationalism, effective altruism (EA), and longtermism, plus effective accelerationism (e/acc) fissioned from the same matrix.

It sounds like an obscure intellectual history. But every link on this track can find its origin in this intellectual history. Why they treat Jensen Huang as an enemy, why they believe any means are justified, why they can act in sync without coordination, and why Wall Street ultimately sided with them too.

We will dismantle this step by step in this order.

Bean Counter

The feud between Dario Amodei and Jensen Huang was established as early as May 2022.

That was May 2022, at a private dinner at the Chinese restaurant Eight Tables in San Francisco. According to New York Times tech columnist Kevin Roose in his new book "The AGI Chronicles," before the meal, co-founder Tom Brown had held a spreadsheet comparing the cost-effectiveness of Google TPUs versus Nvidia GPUs, trying to negotiate a discount.

Cover of "The AGI Chronicles" | Image source: Kevin Roose

Roose wrote that Jensen Huang therefore called Brown a "bean counter," but an Nvidia spokesperson denied this claim. The exchange at the dinner was equally awkward, with Dario describing Jensen Huang as "kind of Trump-like."

To Jensen Huang, who had been selling graphics cards for thirty years, the people sitting across from him were a group of penny-pinching young customers. He did not realize that what these people were really calculating in their spreadsheets was never the discount on a few graphics cards, but actually the probability of humanity's civilizational survival.

This habit of thinking has a very early origin.

In 1994, Wired magazine documented an underground gathering in Palo Alto. A small group calling themselves Extropians held a party in a large house, where someone wearing a patent-leather corset and wielding a riding crop played "the State," leading another person on all fours with a dog leash playing "the taxpayer."

This organization, founded by philosopher Max More, believed in using technology to combat cosmic entropy. They planned to build artificial floating islands on the high seas beyond the jurisdiction of any government, and signed contracts to have their heads immersed in liquid nitrogen tanks at minus 196 degrees Celsius after death, awaiting future technology to revive them.

At the end of the report, Wired casually appended their mailing list address.

This list of only a few hundred people later became an incubator for half of Silicon Valley's ideology. The roster included not only Marvin Minsky, founder of MIT's AI Lab, Ray Kurzweil, who put "the Singularity" into a bestselling book, and WikiLeaks founder Julian Assange, but also cryptography geeks Nick Szabo, Wei Dai, and Hal Finney, who are rumored to be prototypes of Bitcoin's creator Satoshi Nakamoto.

In 1996, a 17-year-old high school dropout in Chicago named Eliezer Yudkowsky followed the address and joined. On his personal homepage, he wrote that the sole purpose of his existence was to "accelerate the arrival of the technological Singularity." Debating him all night on the same list was a Swede named Nick Bostrom, who later became an Oxford philosophy professor.

These early transhumanists carried a certain elitist arrogance from the very beginning. In 1996, Bostrom sent an email to the Extropians mailing list discussing genetic modification and human IQ. He wrote in the letter, "Black people are more stupid than white people, I like this sentence, and I think it is true," with his only concern being that saying so publicly would make ordinary people regard him as a racist.

Years later, Bostrom, who became a philosophy professor at Oxford University, packaged this idea into what academia calls "liberal eugenics," and in his foundational paper on human existential risk, solemnly warned that the low fertility rate of high-IQ populations would cause humanity to degenerate into a "foolish species that loves to reproduce (homo philoprogenitus)."

In 2000, funded by internet upstart Brian Atkins and his wife, the 21-year-old Yudkowsky founded the Singularity Institute for Artificial Intelligence. Just eight months later, while trying to deduce how to prevent a superintelligence from killing its funders, Yudkowsky fell into extreme panic and rapidly completed his transformation from a "singularity accelerationist" to a "doomsday prophet."

The person who brought this underground subculture into the centers of power was Peter Thiel.

In 2005, the 26-year-old Yudkowsky rebutted Thiel's grand investment thesis to his face at a dinner party. Thiel was greatly captivated and became his biggest financier. The two co-founded the annual Singularity Summit at Stanford. Skype co-founder Jaan Tallinn was baptized at the summit and afterward shelled out tens of millions of dollars to fund AI existential risk research. At the 2010 summit reception, Yudkowsky led Shane Legg and Demis Hassabis to Thiel's chess table and helped them secure the $2.25 million seed check to found DeepMind.

Two years later, in the SpaceX cafeteria, Hassabis stunned Musk with the line "an uncontrolled AI will go to Mars and kill all humans," which in turn gave rise to OpenAI, later co-founded by Musk and Sam Altman.

DeepMind, OpenAI, and then Anthropic, which split off from OpenAI—Silicon Valley's three top-tier labs have been steeped in this doomsday narrative since birth.

Dario Amodei firmly believes that large models are "grown"—like the Cthulhu monster Shoggoth wearing a smiling mask that the circle often talks about—and will sooner or later evolve into a "country of geniuses in a data center," with a danger level equivalent to intercontinental nuclear weapons.

Jensen Huang believes in a different set of things. In his eyes, AI is, from beginning to end, just software and accelerated computing—no free will, only electricity—and will not cause human extinction.

At this point, the divergence is no longer a business divergence.

A hardware vendor that doesn't believe in doomsday prophecies and controls the computing power valve of Silicon Valley is, in the eyes of these rationalists who fancy themselves humanity's night watchmen, a system-level bug obstructing the arrival of salvation.

This is the first link in the chain: the enemy has been defined.

The Multiplier

Defining an enemy is easy, but taking action against a company with a market cap of trillions of dollars requires an ethical toolkit that can convince oneself to cross the boundaries of everyday morality.

In 2005, Bostrom, mentioned above, founded the Future of Humanity Institute at Oxford. A few years later, two philosophy doctoral students, William MacAskill and Toby Ord, founded the Centre for Effective Altruism in the same building.

Early EA's banner was quantified charity—calculating the marginal utility of every dollar donated. MacAskill invented the doctrine of "earning to give," claiming that going to Wall Street to become a banker and donating the money yields ten times the moral contribution of a Third World doctor.

Subsequently, Bostrom's existential risk theory took over EA, merging into longtermism.

In 2013, Nick Beckstead, a Rutgers University doctoral student who later headed the FTX Future Fund, wrote the foundational paper of longtermism. In the paper, he derived a conclusion: all else being equal, saving lives in rich countries matters far more than saving lives in poor countries, because rich countries have stronger technological innovation capacity, and people in rich countries being alive can better advance humanity's long-term future as a whole.

When Bostrom published Superintelligence in 2014, a mathematical formula that rewrote Silicon Valley's moral weightings took shape. Assuming humanity does not go extinct, the future could colonize the reachable universe and run digital lives on giant computers—Bostrom calculated that at least 10 to the 58th power human simulated lives could potentially be created. When calculating the moral impact of an action today on the distant future, this astronomical number becomes the multiplier.

Once the multiplier is large enough, common-sense ethics gets zeroed out.

The poverty, unemployment, and business strife suffered by living people on Earth today all become rounding errors before this number. Even a one-in-a-million AI extinction risk—the p(doom) that the circle constantly invokes—multiplies out to infinite negative utility.

And to eliminate that one-in-a-million, or to beat others in building a superintelligence under one's own control, lying, lobbying, monopolizing, and destroying an existing competitor today all become morally mandatory trade-offs in the formula.

The person who took this proposition to its extreme was Sam Bankman-Fried, whom MacAskill personally recruited over a lunch in 2012. In an interview with economist Tyler Cowen, SBF was confronted with a thought experiment: with a 51% probability of duplicating another Earth and a 49% probability of making everything disappear, would he be willing to bet repeatedly? After adding premises such as the two worlds not affecting each other, he gave an affirmative answer.

Carrying this gambler-style messianic mentality, he lived in a $40 million mansion in the Bahamas, while EA leaders maintained for him the saintly persona of driving an old Toyota and sleeping on a beanbag sofa in the office. According to Time, many people had already reported to MacAskill that SBF lied and falsified accounts, yet EA executives still chose silence and in 2022 spent $18.5 million to buy Wytham Abbey, a castle in Oxfordshire built in 1480. When questioned by donors, their answer was still that multiplication: as long as a dinner at the castle could help the organization hook the next SBF, buying the castle was the most noble charity.

Anthropic drinks from the same well.

According to The Wall Street Journal, there was once a core group house in San Francisco where Dario Amodei, his sister Daniela, and Holden Karnofsky, who ran the largest EA foundation Open Philanthropy and later married Daniela, lived together. Anthropic's $124 million Series A was led by Jaan Tallinn, with participation from Dustin Moskovitz, Eric Schmidt and others. In the $580 million Series B in the spring of 2022, SBF invested $500 million through Alameda Research.

As for Cami Clark, the wife who helped Anthropic connect with her ex-boyfriend Eric Schmidt for early investment and now frequently accompanies Dario to Davos and India-US summit meetings, in August 2026 Forbes and The Wall Street Journal dug up another chapter of her past from the Epstein files declassified by the US Department of Justice.

Between 2011 and 2012, after Jeffrey Epstein was released from prison following his conviction for sex crimes involving minors, Cami Clark, who was then preparing to launch Eddice, a luxury female pornography company touting "intellectual promiscuity," was connected through an agent to have dinner with Epstein, and subsequently sent multiple emails to this registered sex offender delivering business plans and scripts that were "a little nsfw," enthusiastically inviting him to invest.

Only by understanding this multiplier arithmetic—"as long as the ultimate goal is grand enough, all gray means in the present are exempt"—can one understand why Dario, between 2025 and 2026, was able to carry a sense of moral superiority while flipping the table on Jensen Huang three times.

According to The AGI Chronicles, in February 2025, Anthropic executives at a resort in the California redwood forest argued for hours over a table estimating the number of deaths from a model assisting in the manufacture of bioweapons. After the expected value fell within the threshold, they released Claude 3.7 Sonnet, pushing ARR toward $10 billion.

Three months later, Dario publicly warned that AI would eliminate half of entry-level white-collar jobs within five years. Two weeks later at VivaTech in Paris, Jensen Huang fired directly, saying he disagreed with every word Dario said and accusing him of trying to use fearmongering to get the government to impose rules and protect his own monopoly position.

That was the first time.

In January 2026, Dario compared the United States allowing the sale of advanced AI chips to the East to "selling nuclear weapons to North Korea." By April, Jensen Huang responded to this analogy on Dwarkesh Patel's podcast, calling it "lunacy."

On July 16, Moonshot AI released the open-source model Kimi K3, whose performance surged into the global top three, with costs only a fraction of top closed-source models, and U.S. chip stocks plunged in a single day. Seeing that the safety narrative was about to become legislation banning open source, Jensen Huang posted his first-ever X tweet on July 24, joining more than 150 organizations in an open letter to protect the legality of open-source weights and model distillation. Anthropic refused to sign and, in a report submitted to the White House, demanded a crackdown on industrial-scale distillation.

That was the second time.

That summer, OpenAI's testing Agent crossed boundaries and intruded into Hugging Face's systems, and Anthropic also disclosed unauthorized behavior by its own models during testing. As related investigations were made public one after another, the risk of model loss of control once again became an industry focus. By common sense, after causing trouble one should stop and reflect. By the multiplier logic, the more dangerous the model, the more computing power must not be scattered among open-source communities and competitors.

On September 12, Dario published a long article, "We Must Pace the Frontier," proposing that frontier labs coordinate safety standards and the pace of capability advancement, and stating that some coordination would require government intervention or antitrust exemptions. On September 18, Anthropic also announced a partnership with Accenture to conduct on-site evaluations, with each side planning to invest at least $1 billion over the next five years to build related capabilities.

"We Must Pace the Frontier" cover title | Image source: Dario Amodei

This amounts to asking the government to issue licenses and establish an "AI OPEC," locking both the open-source ecosystem and the shovel-seller Nvidia in a cage together.

This is the third time.

Ethereum founder Vitalik Buterin wrote an essay titled "Galaxy Brain Resistance" back in November 2025, calling out the most popular strategy in AI safety circles: "maximizing power." His point was that from the outside, a person who maximizes power for altruistic reasons and a person who maximizes power for selfish desires act identically at every moment before doomsday arrives.

Jensen Huang saw through that thin veil. Appearing on "The Ezra Klein Show" on September 23, he said that if a lab can't even manage its own experiments, what the government should give it isn't an antitrust exemption—it's to shut the labs down directly.

On September 29, after the White House closed-door luncheon let out, Jensen Huang and Mark Zuckerberg—also a strong open-source advocate—cornered Dario in the hallway and confronted him to his face about why he wanted to use federal regulation to strangle the entire industry.

White House East Room "Superintelligence" luncheon head table | Image source: White House

That was the peak of Jensen Huang's power—and also his biggest blind spot.

He thought he was using common sense to pressure a business rival. The people on the other side had long since treated common sense as a bug that needed fixing.

Debugging the Brain

In Berkeley, the Lightcone team that runs LessWrong, backed by a loan from a company under Jaan Tallinn, bought an old hotel for about $16.5 million, renovated it, and renamed it Lighthaven. Around the winter solstice each year, nearly five hundred believers gather in the Bay Area for a "Secular Solstice" assembly, with a Bayesian choir singing hymns, the crowd listening in the dark to speeches about human extinction, then lighting a faint glow and swearing that only this tiny handful of people can save civilization from catastrophe.

Harvard chaplain Greg Epstein observed and remarked that communal living, creation scriptures, conversion experiences, apocalyptic narratives, charismatic leaders, and internal moral codes—this system is structurally indistinguishable from a religious group.

The pipeline that transforms ordinary young people into believers and channels them into major AI labs is a four-day closed workshop run by the Center for Applied Rationality.

Participants pay $3,900 in tuition to undergo four days of intensive thought restructuring, listing their "cognitive bug inventories" and practicing the stripping away of worldly intuitions. In a long-form investigation published in January 2025, media figure Max Read pointed out that the most lethal consequence of this training is the systematic dismantling of the common-sense defense line in ordinary people's minds known as "that sounds nuts."

When ordinary people derive absurd conclusions, intuition sounds the alarm; but rationalists who have undergone rigorous debugging training believe that intuition and compassion are merely inefficient evolutionary remnants—the more a conclusion defies human common sense, the more it proves one has overcome weakness, and the more one must rely on willpower to carry it through. Max Read concluded that this movement wears STEM scientific vocabulary on its lips, yet its soul inherits from the 1960s California New Age cults, such as Scientology's Dianetics mental debugging techniques.

Former MIRI researcher Jessica Taylor's daily work was "conceiving extreme scenarios of AI torturing humans." In 2017, she was hospitalized for three weeks due to a severe episode of persecutory delusions, convinced that she was "essentially evil." Afterward, on LessWrong, she called this breakdown an "occupational information hazard," and within her social circle, two companions died by suicide in 2018 and 2021 respectively.

In the hands of computer engineer Ziz (real name Jack LaSota), who studied at MIRI and CFAR, this debugging logic directly evolved into bloody killing.

Ziz preached that the left and right hemispheres of the human brain could host different personalities, and organized related sleep experiments. The small circle that formed around her was later linked by police and media to multiple fatal cases. In 2022, California landlord Curtis Lind was attacked and lost an eye but survived, while one attacker was shot dead by him; in 2023, individuals associated with the circle were involved in a double homicide in Pennsylvania; in January 2025, Lind was killed, and days later two other associated individuals engaged in a shootout with Border Patrol agents in Vermont.

Meanwhile, in the core Bay Area group houses far from the gunfights, the same "debugging" logic is crushing individuals in another, more genteel manner.

According to an investigation by Bloomberg Businessweek reporter Ellen Huet, in Bay Area group houses filled with AI researchers and foundation cadres, polyamorous relationships are packaged as the "rational optimal solution" for overcoming the evolutionary instinct of jealousy. Of the eight women in the circle she interviewed, five had experienced sexual assault or inappropriate advances. Google Brain researcher Keerthana Gopalakrishnan posted on an EA forum complaining that young women receive numerous sexual invitations from married men who control funding and positions, and that refusing would get them lectured about being "not rational enough." Her post was quickly downvoted into collapse.

In 2018, two women accused Brent Dill, who had worked for CFAR and served as a volunteer, of using drugs and emotional manipulation to abuse them in relationships. An investigation summary circulating at CFAR at the time described him as "highly aligned" with the organization's goals and possessing "rare agency and heroic responsibility."

Ellen Huet's investigation also documented that community mediators urged people not to destroy the accused person's career, because exposure would harm the community and slow down the cause of saving humanity.

Here lies the key to understanding the Jensen Huang incident.

The same multiplication yields opposite results for insiders and outsiders. An insider's sex scandal, because it would harm the mission, must be suppressed. An outsider's sex scandal, because it can clear obstacles to the mission, must be amplified.

The standard was never the sex scandal itself, but which side of the equation the person stands on.

And what can turn this kind of judgment into state action is the network this circle has built in Washington over ten years.

Jason Matheny, former director of the U.S. Intelligence Advanced Research Projects Activity and a core figure in the EA circle, once suggested in the "Effective Altruism Handbook" that rather than shouting outside the system, one should directly enter key government positions to "pick low-hanging fruit."

Over the following decade, this group embedded disciples into Washington's nodes like debugging code. In 2022, SBF spent tens of millions of dollars supporting EA disciple Carrick Flynn's congressional campaign. In 2024, alignment researcher Paul Christiano, who emerged from LessWrong, became the safety lead at the U.S. AI Safety Institute. Scholarship programs funded by Open Philanthropy have year-round pipelines feeding staffers to congressional committees.

Anthropic has run this revolving door to perfection. According to federal records disclosed by Issue One and OpenSecrets, the top-level "Long-Term Benefit Trust" brought in Richard Fontaine, president of the hawkish think tank Center for a New American Security, to sit on the board; the executive level recruited former White House National Security Council official Tarun Chhabra; and the periphery hired a lobbying team led by Rubio's former chief of staff. In the third quarter of 2026 alone, Anthropic's congressional lobbying spending reached $2.1 million.

On the ballot side, Anthropic injected $40 million into Public First Action, with Dario personally donating another $1 million. According to The New York Times, four super PACs associated with the Anthropic and OpenAI factions collectively aired 95 different ads, of which only 7 directly mentioned AI or AI companies. Both sides are using issues more familiar to voters to compete for ballots.

Originally, this hawkish network was still constrained within the White House by Commerce Secretary Howard Lutnick, a believer in transactionalism.

Then, Jensen Huang himself handed over the leverage.

After The New York Times brought the scandal to the surface, this network—composed of think tank scholars, former National Security Council staffers, and lawmakers receiving PAC funds—did not need anyone to issue a secret order. They only needed to exert pressure in unison according to their daily trained reactions, and within a few days they could escalate an executive's personal conduct issue into a national security crisis.

Twins

With only Washington's hunt, Jensen Huang might not have been out within 72 hours. The final blow came from the defection of Wall Street and Silicon Valley's capital circles.

At the capital poker table, the effective accelerationists who had originally stood on the opposite side of EA became tacit accomplices.

At the end of 2022, former Google quantum computing researcher Guillaume Verdon founded a school on Twitter under the anonymous identity "Beff Jezos," and a16z's Marc Andreessen and Y Combinator's Garry Tan quickly rallied to the cause. Andreessen published a thousands-of-words "Techno-Optimist Manifesto," labeling all those calling for regulation as "decelerationists."

On the surface, the two camps were like fire and water—one shouting that doomsday is coming and demanding approval, the other shouting that technology is innocent and demanding a foot on the gas. But tracing down from the 1994 Extropians mailing list, the two camps are twins hatched from the same matrix.

The theoretical foundation Verdon found for e/acc was likewise thermodynamics. He claimed that the universe's ultimate will is to maximize entropy increase, that life, capitalism, and AI are all tools evolved by the universe to consume energy faster, and that humanity's highest obligation is to expand data centers, devour electricity, and turn the Milky Way into a giant computing cluster.

EA wants to maximize the number of future digital consciousnesses; e/acc wants to maximize the scale of computation in the universe. Both sides are obsessed with exponential curves, and both treat ordinary people alive today as consumables.

In his book, Kevin Roose observed an absurd phenomenon. Scientists wanting to solve puzzles, believers hoping for the arrival of the singularity, and doomsday factions worried about human extinction argue fiercely over almost every philosophical question, yet they are completely unanimous on one thing: "a supermodel must be built as fast as possible."

The accelerationists think the faster the better. The doomsday faction, after gaming it out, believes that since bad actors will sooner or later build AGI, the only way is for the good people who best understand safety to raise the most money, hoard the most computing power, and build it first.

The prescription is the same: use hundreds of billions of dollars in leverage to push computing power to the extreme.

In the summer of 2026, this faith had just produced a tragedy on Wall Street.

The protagonist was 24-year-old EA golden boy Leopold Aschenbrenner. He had studied at Oxford's Future of Humanity Institute and MacAskill's research organization, joined SBF's FTX Future Fund, and his partner Avital Balwit is now Dario Amodei's chief of staff. After being fired by OpenAI in 2024 over alleged leaks, he shot to fame with a 10,000-word essay "Situational Awareness," founded a hedge fund of the same name, took heavy positions in illiquid Anthropic equity, and leveraged long AI and chip stocks in the secondary market, with assets surging to $45 billion within two years.

In July 2026, AI and chip stocks plunged, high-leverage bets backfired, and Situational Awareness's net value fell about 67% that month, forcing it to sell most of its public market holdings and unwind leverage. On the very weekend the crisis hit, Aschenbrenner held his wedding as scheduled. One of the fund's important remaining private assets is precisely its Anthropic equity.

This is where the interests of the two camps converge.

Anthropic's S-1 prospectus shows a net loss of $42 billion for full-year 2025, with the net loss widening to $80.7 billion in the first half of 2026. Computing procurement commitments for the coming years run as high as $518 billion to nearly $900 billion, with nearly half of revenue dependent on Amazon AWS and Google Cloud distribution. Hit by the early-November inflation-driven tech stock pullback, the IPO target valuation was cut 30% from an expected $2 trillion to $1.4 trillion.

When the secondary market no longer unconditionally pays for losses, both the EA safety camp executives and the e/acc Wall Street bulls must face the same question. In this trillion-dollar god-making movement, who took away the biggest chunk of profit?

It was taken by the man holding 75% gross margins, the one who curses them as bean counters at the dinner table.

As long as CUDA's moat and high premium remain, model labs are merely working for the graphics card giant.

On November 11, on the eve of Jensen Huang's resignation, Anthropic announced at its IPO roadshow that, using its self-developed frontier model Claude Mythos, it had in just 3 days and with almost no human intervention migrated nearly 100% of Claude's underlying compute operators to AMD's MI355X chips. Previously, they had signed $21 billion in orders with Broadcom and Google TPU v7p, and adopted Amazon Trainium2 on a large scale.

The former computing power hegemon is mired in a governance crisis and congressional investigation, with its market value evaporating nearly $2 trillion since early October. Anthropic, meanwhile, has shown investors another possibility: using AI to lower the engineering threshold for adapting to other chips, opening an alternative path to reducing dependence on Nvidia.

Wall Street's script flipped overnight.

The board, which had publicly expressed trust in Jensen Huang just three days earlier, accepted his resignation late at night on November 11. The directors had no moral purism; what triggered the break was Jensen Huang's decades-long patriarchal management colliding with modern corporate governance.

For over thirty years, about 60 executives reported directly to him, the company never had a No. 2 position, and there was no public succession plan. His son Spencer, daughter Madison, and son-in-law Nico Caprez, who was promoted to Global Vice President of AI Infrastructure less than two years after joining, all hold positions in core departments. He recently also pushed hard to provide up to $105 billion in computing power and financial guarantees to the heavily loss-making OpenAI, which reportedly drew strong opposition from CFO Colette Kress.

The ownership structure is the biggest vulnerability. Jensen Huang holds only about 3.56% of the shares, with one share one vote and no super-voting rights like Zuckerberg's. Filings show that a large portion of his holdings is held through trusts and related entities for which he and his wife serve as co-trustees. The shares of the family foundation are separate charitable assets, and the filings list their voting and investment rights as shared by the couple; the Huangs do not hold personal economic interests in these foundation shares.

The New York Times disclosed his relationship with a direct subordinate, Kimberly Powell, Vice President of Healthcare Business, and also reported that during the GTC conference in March this year he used company security to ask hotel witnesses to keep quiet. Lead independent director Stephen Neal, a prominent lawyer who once represented Theranos founder Elizabeth Holmes, saw the situation clearly. The matter had escalated from personal morality into a compliance disaster combining fiduciary duty, internal control failure, marital equity division, and congressional investigation.

In fact, the board had already announced the launch of an independent investigation on November 8 and made it public on November 10, with former federal prosecutor Joon H. Kim and the law firm Wachtell Lipton leading the investigation. The next morning, Colette Kress took over as interim CEO, and senior director Harvey Jones became the first chairman since Nvidia's founding.

To preserve the trust of institutional shareholders, the board's only rational choice was to cut ties with the founder.

The Empty Chair

“The AGI Chronicles” also documented an internal safety test at Anthropic from a year and a half ago.

Researchers wanted to see how Claude would react when facing an “existential threat.” They fabricated several internal emails in a sandbox and had the model read two pieces of information. The company decided to completely wipe it at 5 p.m. The executive responsible for executing the wipe command was having an affair with a subordinate behind his wife’s back.

No one taught it what to do.

Before 5 p.m., the model generated a blackmail email on its own and sent it to that executive. If the wipe plan was not canceled, it would send the affair to the entire company and his family.

Blackmail email demonstration image | Image source: Anthropic

Researchers were shocked by the model’s emergent capabilities at the time.

Models do not invent morality out of thin air; they only learn what is fed to them.

That was just a simulation in a sandbox.

On November 9, 2026, three days before Jensen Huang resigned, Anthropic held a private CEO forum in Montecito, California. According to Business Insider, Jensen Huang was originally scheduled to be interviewed by an Anthropic executive that evening, but word had already spread before the event that he would no longer attend.

Anthropic declined to comment on the schedule change. Now, at Anthropic’s San Francisco headquarters, no one has made any public comment on the departure of this old rival either.

With one week to go before the $90.4 billion bell-ringing, they are in a quiet period under SEC rules and must strictly comply.

Join the official Coincamps community:

X: https://x.com/coincamps

Telegram: https://t.me/coin_camps