From 500 million to 30 billion, SBF, in prison, has invested in the most valuable company in the AI ​​era.

Author: Deep Tide TechFlow

Anthropic is now the most important AI company on the planet, perhaps without exception.

Its Claude large model is deployed in the Pentagon, U.S. intelligence agencies, and national laboratories, and is used by the U.S. military for intelligence analysis and target selection for military strikes against Iran.

Its annualized revenue skyrocketed from zero to $14 billion in less than three years. In February 2026, Anthropic completed a $30 billion Series G funding round, bringing its post-money valuation to over $380 billion. Tech giants like Amazon, Google, Nvidia, and Microsoft lined up to pour money into it.

Over the past few weeks, it has been engaged in a globally watched contest with the Pentagon over the weaponization of AI.

In the early financing history of this company, one name is still talked about today: Sam Bankman-Fried.

In April 2022, ChatGPT didn't even exist, and the AI ​​sector was nowhere near as hot as it is today. SBF, through his hedge fund Alameda Research, poured $500 million into Anthropic's Series B funding round, acquiring 86% of the entire round and approximately 8% of the equity. Seven months later, the FTX empire collapsed, and SBF became the protagonist of one of the biggest fraud cases in cryptocurrency history, sentenced to 25 years in prison. That $500 million was deposits from FTX clients.

But if SBF hadn't been arrested, and if the money had been legally obtained, based on today's valuation of $380 billion, an 8% stake would theoretically be worth over $30 billion. Turning $500 million into $30 billion represents a return of over 60 times; in the history of venture capital, this absolute profit would rank among the highest.

A crypto fraudster serving time in federal prison nearly made the wildest bet in AI investment history.

How did SBF manage to find Anthropic in 2022? Why did they dare to invest $500 million? And why did Anthropic accept this money?

The answer lies in a circle called "effective altruism".

A shared apartment, a sport, a check

In mid-2010s San Francisco, a group of people lived in the same type of shared housing, attended the same type of parties, read the same type of papers, and believed in the same philosophy.

This philosophy is called Effective Altruism (EA). Its core proposition is simple: charity should not be based on feeling, but on calculation. Every dollar should flow in a direction that mathematically "maximizes the good." And according to a significant branch of EA, the greatest existential risk facing humanity is not nuclear war, nor a plague, but out-of-control artificial intelligence.

Dario Amodei is immersed in this circle.

He is the 43rd signatory to the Giving What We Can Pledge, pledging to donate at least 10% of his income. He became a fan of GiveWell back in 2007 or 2008.

He shared an apartment with two other people: one was Holden Karnofsky, co-founder of GiveWell and Open Philanthropy, and one of the most influential fundraisers in the EA movement; the other was Paul Christiano, a key researcher in the field of AI alignment. At the time, Dario and Paul were both serving as technical advisors to Open Philanthropy.

Karnofsky later married Dario's sister, Daniela. After their engagement, the couple lived with Dario for a time. In January 2025, Karnofsky quietly joined Anthropic as a "technical staff member," responsible for security strategy. When Fortune reporters discovered this, Anthropic hadn't even publicly announced the appointment.

This is a close-knit social network.

Amanda Askell, an early employee of Anthropic, is the ex-wife of William MacAskill, one of the founders of the EA movement. She was the 67th signatory to the GWWC, and her doctoral dissertation focused on a core issue in the EA philosophy: how to deal with infinity in ethics.

Anthropic's most important governance body, the Long-Term Benefit Trust, theoretically wields significant control over the company. Three of its four members come directly from the EA system: Neil Buddy Shah, former managing director of GiveWell; Zach Robinson, CEO of the Center for Effective Altruism; and Kanika Bahl, CEO of Evidence Action, a long-term grantee of GiveWell.

The three biggest financial backers in the history of EA were all early investors in Anthropic: Facebook co-founder Dustin Moskovitz, Skype co-founder Jaan Tallinn, and Sam Bankman-Fried.

This is the real path SBF took to find Anthropic. It wasn't some brilliant investment insight or a forward-thinking judgment on the AI ​​field; it was simply a money loop within an industry: EA's money flowed to EA's projects, solving the problems defined by EAs.

SBF adheres to a more radical approach within Expert Advisors (EAs): "earning to give." He resigned from the Wall Street quantitative firm Jane Street to pursue cryptocurrency, publicly declaring that his goal is not personal wealth but "altruism"—to first make as much money as possible and then invest it in areas that will have the greatest positive impact. Anthropic's mission, "safely developing powerful AI," is practically EAs' standard remedy for the inherent risks of AI.

In May 2021, Jaan Tallinn led Anthropic's Series A funding round, raising $124 million, with Moskovitz also participating. In April 2022, SBF took over, leading the Series B round and writing a check for $500 million, representing 86% of the total $580 million raised. Other investors in the round included Caroline Ellison, Nishad Singh, and James McClave of Jane Street.

The list of investors is telling. Caroline Ellison is the CEO of Alameda, Nishad Singh is the Engineering Director of FTX, and Jane Street is the former owner of SBF.

This $580 million Series B funding actually came almost entirely from SBF and the pool of funds controlled by it.

Red Flag and Compromise

Dario Amodei is not stupid.

In a later in-depth interview, he recalled that SBF seemed to be "a person who was optimistic about AI and concerned about security," which aligned well with Anthropic's direction. But then Dario said something crucial: he noticed "enough red flags."

So he made a decision: take the money, but isolate them in the governance structure. SBF received non-voting shares and was excluded from the board of directors. Dario later described SBF's actions as "far, far, far more extreme and egregious than I had imagined," three "much more"s stacked together.

This decision later proved to be extremely wise. However, it also left a sharp question: if the warning signs were so numerous that isolation measures were necessary in the governance structure, why was it still taken?

You could argue that the AI ​​funding environment in early 2022 was far less vibrant than it is today. Anthropic needed a large sum of money to build computing power, and finding an investor willing to put up $500 million at once, regardless of how many "red flags" they had, was difficult.

But there's an even more subtle reason: in the operating logic of the EA (Expert Advisor) circle, the "cleanliness" of the funding source is never a priority. What matters is the "effectiveness" of the funds—whether they can help you do more. SBF's entire wealth narrative is built on this foundation: making money is the means, doing good is the end, so the way you make money doesn't have to be so particular, as long as the final "good" output is large enough.

This logic was taken to the extreme of crime by SBF, but at the time he invested in Anthropic, it seemed to be just a radical but not illegal philosophical choice.

After the Collapse: A Dark Comedy

The rest of the story is well-known in the crypto community.

In November 2022, CoinDesk exposed Alameda's balance sheet, and Changpeng Zhao announced the sale of FTT, triggering a run on FTX and the collapse of the empire within nine days. SBF was arrested, extradited, tried, and sentenced to 25 years in prison in March 2024. Anthropic's 8% stake, along with all its assets, was frozen in bankruptcy liquidation proceedings.

One episode during the trial that was excluded by the court is worth mentioning.

SBF's defense lawyers attempted to use Anthropic's investment as evidence of "foresight," arguing that "he didn't just squander money; he made an investment decision that multiplied the valuation several times over."

Prosecutor Damian Williams responded firmly: whether the investments were profitable or not was completely irrelevant to the fraud charge. Even if you steal someone else's money to invest and it makes a profit, it's still stealing. The judge accepted the prosecution's argument, and Anthropic's name was excluded from the trial.

The prosecution added insult to injury: Isn't FTX itself the best example of what not to do? It was valued at $18 billion in 2021, $32 billion in 2022, and is now worthless.

Then comes the liquidation auction.

In March 2024, the company completed its first round of financing, valuing the company at $884 million.

The largest buyer, Abu Dhabi's sovereign wealth fund Mubadala, invested $500 million, exactly the same amount SBF invested in the past. The second largest buyer was Jane Street, the former employer of SBF and Caroline Ellison. Craig Falls, head of quantitative research at Jane Street, even personally invested $20 million. SBF's first job after graduating from MIT was as a trader at Jane Street, and now his former employer has bought back the shares that a former employee purchased with embezzled funds.

A total of 1.34 billion was recovered in two rounds. This money flowed into FTX's creditor compensation pool, becoming an important source of funds for affected users to recover their deposits.

What if the liquidation team doesn't sell?

In February 2026, Anthropic completed a $30 billion Series G funding round, bringing its post-money valuation to $380 billion. Without considering dilution, that 8% theoretically went from $1.34 billion to $30 billion. The liquidation team, of course, didn't choose this path; their responsibility was to liquidate assets as quickly as possible to repay creditors. However, this numerical difference—$1.34 billion versus a potential value exceeding $30 billion—is key to understanding why this story is still being discussed today.

It is the biggest regret in the entire FTX bankruptcy case.

EA's collective amnesia

Anthropic's size and influence today are self-evident, but an interesting phenomenon is that the company is systematically distancing itself from the EA movement.

Its seven co-founders have pledged to donate 80% of their personal wealth, which, based on current valuations, is worth approximately $38 billion. Nearly 30 Anthropic employees registered for the EA conference in San Francisco, more than double the combined attendance of OpenAI, Google DeepMind, xAI, and Meta Superintelligence Labs.

But Daniela Amodei said in a Wired interview, “I’m not an expert on effective altruism. I don’t agree with that term. My impression is that it’s a bit outdated.” Her husband, one of the most influential fund allocators in the EA movement, had just joined her company.

This stance of "taking EA's money, using EA's people, living in EA's shared housing, but not acknowledging that they are EA" became understandable after the SBF case. FTX's collapse caused the reputation of the EA movement to plummet. Anthropic needs to distance itself from this label, just as any smart company would cut ties when its brand is negatively associated.

But the facts are there: Anthropic's founding logic stemmed from the core arguments within the EA (Expert Advisor) community regarding the inherent risks of AI; its early funding came almost entirely from funds within the EA network; and its governance structure was controlled by people within the EA system.

Parallel Universes in Prison

Sam Bankman-Fried is currently in federal prison. He is not expected to be released until 2049 at the earliest, when he will be 57 years old.

During his imprisonment, the AI ​​company he invested in with embezzled funds has grown to a valuation exceeding $380 billion and is engaged in a high-profile battle with the Pentagon over the weaponization of AI. Its founder has become a regular feature in The New York Times and on Capitol Hill. If everything is legal, that $500 million bet is enough to make SBF one of the most profitable venture capitalists of our time.

SBF's "making money and donating money" and Anthropic's "safe development of AI" share the same underlying operating system, and are willing to bear unusual means and risks in order to achieve sufficiently large good results.

SBF pushes this logic beyond the boundaries of crime. Anthropic operates on the safe side of this line, but its core proposition—"We must build the most powerful AI ourselves to ensure the safety of AI"—is itself a grand, almost self-justifying gamble.

They grow in the same soil.

In that land, Dario and SBF attended the same parties, believed in the same philosophy, and lived on different nodes of the same social network. One went on to build an AI empire valued at $380 billion, while the other ended up in federal prison.

The $500 million check that connected them remains one of the most bizarre chapters in Anthropic's history.