Why Anthropic's IPO Matters for the AI Race

Monday, October twenty-sixth, twenty twenty-six. Anthropic files a Form S-1 with the SEC. Nasdaq, ticker A-N-T. And buried in the paperwork is a number: in the first six months of this year, the company lost eighty point seven billion dollars.

Eighty... billion. In six months. That's not a typo?

Not a typo. And here's the thing I want you to hold onto for the next ten minutes... that loss is the reason this IPO could be the biggest in history, not the reason it fails. We'll get there.

Okay, you've got me. But start with the basics. Why go public at all? These labs have been swimming in private money.

Right, and Axios made exactly that point back in April. Tech is still in its stay-private-longer era. Primary capital's available, secondary capital's available, Middle Eastern sovereigns writing enormous checks. Anthropic and OpenAI both played that game perfectly.

So what changed?

Scale. Axios's line was... some things are too big to be kept in a private box. Both companies expect to be worth trillions. Plural. Private liquidity is deep, but it isn't bottomless.

And whoever taps the public markets first gets the bigger wallet.

That's Bloomberg's read, June first, when Anthropic confidentially filed and beat OpenAI to the punch. Their framing was blunt: this isn't bragging rights. The first one into US public markets gets an immediate advantage securing chips, data centers and talent. The compute war is a funding war.

So OpenAI staying private isn't strategy, it's just... losing a race?

Reuters quoted an analyst saying Anthropic aimed to beat OpenAI deliberately, for strategic advantage. OpenAI's own IPO is still expected. The two of them are a pair of blockbuster debuts, per the AP. Anthropic just got there first.

Alright. The valuation. Nine hundred sixty-five billion for Anthropic, eight fifty-two for OpenAI. How does the smaller, younger company end up worth more?

Because of where the money actually comes from. Menlo Ventures, mid twenty twenty-five: enterprise LLM spend doubled in six months, three and a half billion to eight point four billion. And Anthropic had thirty-two percent of that enterprise market. OpenAI, twenty-five percent... down from FIFTY percent in twenty twenty-three.

Wait. OpenAI halved its share of the business market? While ChatGPT was the most famous product on Earth?

Consumers love ChatGPT. Companies bought Claude. The turn started with Claude Sonnet three point five in June twenty twenty-four, then Sonnet three point seven, then Claude Code. Coding agents. That's the whole story.

And the revenue race?

This is the part that genuinely stunned me. Start of twenty twenty-five: OpenAI six billion annualized, Anthropic one billion. Six to one. By early twenty twenty-six the gap had WIDENED, twenty billion versus nine.

So Anthropic was losing. Badly.

Then April sixth, twenty twenty-six, Anthropic says it crossed thirty billion annualized. OpenAI had just disclosed two billion a month. Twenty-four billion. Reuters called it a remarkable feat of catch-up. By late May, Anthropic says run-rate crossed forty-seven billion.

From nine billion to forty-seven in about five months. That's not growth, that's a glitch in the matrix.

The S-1 confirms it with audited numbers. Revenue for the first half of twenty twenty-six: sixteen point two nine billion dollars. Up thirteen hundred and four percent year over year. All of twenty twenty-five was four point five nine billion. Bloomberg's chart shows the first half of this year is nearly four hundred times all of twenty twenty-three.

Okay, now come back to the eighty billion dollar hole. Because that still terrifies me.

So about seventy-two billion of that loss isn't operations at all. It's an accounting charge on convertible notes, triggered by the company's own valuation going UP. The more Anthropic is worth, the bigger the paper loss.

So they lost eighty billion dollars because they got more valuable.

Mostly. The real cash burn is still huge — they lost forty-one point nine billion across all of twenty twenty-five, eight point three billion in twenty twenty-four. Compute is brutally expensive. But the headline number is a mirage.

What's the number that isn't a mirage?

Gross margin. PitchBook's Harrison Rolfes told CNBC it's the figure that determines everything — nobody outside Anthropic has ever seen it, and it will, quote, either validate or collapse the entire narrative the private markets have been pricing for three years.

Because if you're selling tokens below what the GPUs cost you, none of this works.

Exactly the risk. Though the customer base looks real: as of September thirtieth, more than eleven thousand nine hundred enterprise customers spending a hundred grand or more a year. No single customer over ten percent of revenue. That's more diversified than Nvidia.

And the bear case?

Gil Luria at D.A. Davidson told CNBC Anthropic appears to have the lead — but Google, Meta, OpenAI are all funded to the teeth, and much of today's usage is trials and experimentation. His words: that may not sustain.

Plus the CEO keeps saying his own product might end humanity, which is an unusual S-1 risk factor.

The AP led with that exact tension. Biggest IPO ever filed by a man warning about existential threat.

So what are we actually finding out next month?

Whether any of this is real. For three years the AI boom has been priced by a small circle of private investors telling each other it's worth it. Anthropic just agreed to open the books to everyone — pension funds, retail traders, short sellers. SpaceX raised a record eighty-six billion in June. This could top it. An eighty billion dollar loss didn't scare them off. The gross margin line might. And after next month, the AI race stops being a story about who has the best model... and starts being a story the market gets to vote on, every single day.