Y Combinator's Bet on AI: How YC Shaped the Anthropic IPO

October seventh, twenty fifteen. Sam Altman, thirty years old, running Y Combinator, puts ten million dollars of his own money into a new research lab. No product. No company. He tells MIT Technology Review he wants work with a twenty to thirty year time horizon.

Thirty years. From the guy whose day job is three-month startup bootcamps.

Right. And the first project he announces? He won't say what it is yet. Two months later, it has a name.

OpenAI.

December eleventh, twenty fifteen. The blog post goes up on Y Combinator's own site: "The first group affiliated with YC Research is OpenAI."

Hang on, so OpenAI was literally a YC project? Not just Altman's side thing?

It was YC Research's first group, and YC Research was one of the donors. Alongside Altman, Elon Musk, Greg Brockman, Reid Hoffman, Jessica Livingston, Peter Thiel and Amazon Web Services. A billion dollars committed on paper.

Jessica Livingston being one of YC's own founders.

Yes. And the stated point was that this money came with no strings. The launch post says the goal is to benefit humanity "unconstrained by a need to generate financial return."

That line is going to age interestingly.

It ages fast. In twenty nineteen, OpenAI takes a billion dollars from Microsoft and creates a for-profit subsidiary. And in December twenty twenty, after nearly five years, Dario Amodei, the VP of research, walks out.

With his sister.

Daniela, who'd held a bunch of roles there, plus a handful of other senior people. Seven co-founders in all. By May twenty twenty one, Anthropic exists with a hundred and twenty four million dollars, led by Jaan Tallinn, the Skype co-founder.

Okay, but I want to be careful here. Is Anthropic a YC company? Because the title of this episode sort of implies a straight line.

It isn't, and that's the honest answer. Anthropic never went through a YC batch. YC didn't fund it. The link is one step removed: the lab the Amodeis left was the thing YC Research started.

So YC's bet on AI pays off... in a company YC doesn't own a share of.

And by twenty twenty, YC had formally cut ties. YC Research became OpenResearch, no longer affiliated. YC's own update says they're proudest helping founders side by side in the batch program.

They let go of the thing that worked?

They let go of the structure. Altman went on to run OpenAI. The research lab experiment, as a YC arm, was over.

All right, so what's the number now? What is Anthropic worth on paper?

Guess. Last private round, May twenty twenty six.

Three hundred billion? Four?

Sixty five billion raised, at a nine hundred and sixty five billion dollar post-money valuation.

I was off by a factor of two and a half.

Series H, co-led by Altimeter, Dragoneer, Greenoaks, Sequoia, Capital Group, Coatue. Samsung, SK Hynix and Micron came in too. TechCrunch reported one institutional investor pledged five billion dollars just to get a meeting with the CFO.

Just for the meeting. That's not investing, that's a cover charge.

Then on October twenty sixth, this week, Anthropic filed publicly. It'll list on the Nasdaq under the ticker ANT. Analysts and investors have it above one and a half trillion. CNBC reports the company's reportedly targeting one and a half to two trillion.

Which would make it instantly one of the biggest tech companies in America. Off a company founded in twenty twenty one.

And the prospectus cuts both ways. Thirteen hundred percent annual sales growth. Also roughly eighty billion dollars in net losses in half a year, and more than ninety pages of risk factors.

Eighty billion in six months. How do you even price that?

Nobody knows. CNBC calls it uncharted territory: the first pure-play AI company going public, with rates rising and people nervous about the size of the buildout.

So the thing Altman built to be free of financial return is now the ancestor of the biggest IPO ever attempted.

The people are the same, the money is completely different. Dario has a physics PhD from Princeton. Daniela came from international development. She told Lightspeed last year they're both bad at what the other's good at.

Seven people to how many now?

Eight hundred plus within four years, and that was before this stretch. The prospectus opens with: "We believe AI is one of the most consequential technologies in human history."

Which is almost word for word the twenty fifteen pitch, except now there are shareholders.

And a roadshow. They still have to go sell it, set a price, allocate shares.

Ten million dollars of one guy's money to this. I'd love to know what Altman thinks when he sees the ticker.