Subscribe
AI
4 min read

Anthropic by the Numbers: The Business Behind the $1.5 Trillion IPO

Anthropic by the Numbers: The Business Behind the $1.5 Trillion IPO Anthropic priced the largest stock market debut on record on Wednesday, November 18, selling 980 million shares at $83 apiece to raise $81.3 billion, with a fully diluted valuation of roughly $1.5 trillion (Reuters via the Jerusalem Post). Deal terms aside, the prospectus tells a sharper story about what the first frontier AI lab to go public actually looks like on paper: revenue compounding month over month, record losses, and

Anthropic by the Numbers: The Business Behind the $1.5 Trillion IPO

Anthropic by the Numbers: The Business Behind the $1.5 Trillion IPO

Anthropic priced the largest stock market debut on record on Wednesday, November 18, selling 980 million shares at $83 apiece to raise $81.3 billion, with a fully diluted valuation of roughly $1.5 trillion (Reuters via the Jerusalem Post). Deal terms aside, the prospectus tells a sharper story about what the first frontier AI lab to go public actually looks like on paper: revenue compounding month over month, record losses, and a near-$1 trillion bill for compute already promised to its infrastructure partners.

Revenue curve against compute cost, Olivetti brochure illustration

Revenue added $10 billion in one month

The numbers in the filing move fast. Anthropic's annualized revenue run rate hit $81 billion as of October 31, up from about $71 billion on September 30. A single month's revenue growth, in other words, roughly equals what some well-known AI startups report as their entire run rate. Quarterly figures show the same shape: revenue rose 47% in the third quarter to $16.96 billion, after preliminary second-quarter revenue passed $11.5 billion (Investing.com, CNBC).

The enterprise base grew with it. The company said it had 11,900 customers generating at least $100,000 in trailing-twelve-month annual revenue as of September 30, against more than 1,500 such customers at the end of 2024. Anthropic ended September with about 5,300 employees, up from roughly 940 at the end of 2024 (The Business Times).

The engine behind that growth is specific. Reuters attributes the surge to demand for Claude models, particularly Claude Code, the coding agent that became a staple for software developers.

Record losses, and what they include

Scale has not yet meant profit. Anthropic reported a third-quarter operating loss of $567 million, after operating income of $974 million in the second quarter. Its net loss narrowed to $23.3 billion from $66.3 billion in the previous quarter, but those headline figures are dominated by non-operating items tied to bonds convertible into shares, a common distortion at companies that raised money in private markets at ever-higher valuations.

Compute is the expense line that matters. The company spent nearly $9 billion on computing capacity and infrastructure in the first half of 2026, close to four times the amount a year earlier, and added another $5 billion in the third quarter. Revenue per dollar of compute is falling as a result, per the filing (Reuters via the Jerusalem Post).

Forward obligations dwarf last year's spend. Anthropic has committed nearly $900 billion in computing spending with infrastructure partners, including Amazon and Alphabet's Google, to train and run its models. The company itself flags the fragility of that chain: its providers "depend on third-party financing and credit support to acquire chips and build capacity," and if that financing tightens, "contracted capacity may be delayed, repriced, reduced, or not delivered."

Data hall grid connected to a single lab building

What the market price implies

At $1.5 trillion, Anthropic lists among the 15 largest publicly traded US companies by market value, near Tesla and ahead of Eli Lilly, per Bloomberg-compiled data (The Business Times). That is well below the $2 trillion floated earlier this year, but above the $1.1 trillion some on Wall Street suggested during the recent tech selloff.

Demand appeared robust despite that backdrop. Orders neared seven times the shares available, Bloomberg reported, and CNBC cited institutional demand around $400 billion with retail demand approaching $100 billion. The $26 billion private placement alongside the IPO drew the chip and cloud incumbents themselves: Nvidia agreed to buy $10 billion of stock, with Alphabet's Google, Microsoft and AMD also participating. Goldman Sachs and Morgan Stanley led a 22-bank underwriting syndicate.

The context for that enthusiasm is unusual: the debut lands while the AI trade is wobbling. The Nasdaq 100 fell 4.4% in the week through Wednesday's close, and Nvidia dropped for a fifth straight session the day after beating earnings expectations, in a market still absorbing the exit of chief executive Jensen Huang.

The bear case in the filing

Competitive data cited by Reuters in September showed OpenAI's flagship GPT-6 Astra accounting for about 13% of enterprise AI spending tracked by the expense platform Ramp, versus roughly 8% for Anthropic's Claude Fable. Meta, one of Anthropic's largest customers, has been working to reduce its use of the company's models as it builds more AI capability in-house. A Capgemini study found 59% of large-company executives plan to increase use of open-weight models over the next year as AI costs exceed budgets, a trend that pressures the price of frontier tokens from below.

Then there is the tension investors are being asked to underwrite: a safety-first lab chasing frontier revenue. CEO Dario Amodei has publicly called for slowing the pace of releasing increasingly powerful models while the company asks the public markets to fund faster training runs. "Our development of highly advanced models, platforms, and applications and expansion of use cases could further increase the risk that our models cause harm," the filing states.

Long-term holders frame the bet as binary. "Our thesis is that the vast amount of value will accrue to whoever is at the frontier," said Peter Singlehurst of Baillie Gifford, which invested roughly $1.5 billion pre-IPO. Skeptics see a new information regime instead. "Once Anthropic reports publicly, it becomes one of the clearest tests of whether the economics of AI are holding up," said Aramina Rosenberg, portfolio manager at Minotaur Capital.

With quarterly disclosures now mandatory and OpenAI expected to follow it to market as early as 2027, Anthropic becomes the first public read on frontier AI economics. Thursday's first trading day under the ticker ANT on the Nasdaq will be the market's first answer.

Sources

AIAnthropicBusiness