#anthropic预计今日公布s1招股书
Anthropic (the parent company of Claude) publicly filed on October 26 under the ticker ANT, with plans to list on Nasdaq. It had submitted a confidential filing in June; this is the public version. The roadshow could begin as early as the week of November 9, with the listing planned for the week of November 16. The company is aiming to raise up to $100 billion at a valuation of $1.5 trillion to $2 trillion—bigger than SpaceX’s June offering ($86.3 billion raised at a $1.77 trillion valuation).
Revenue is growing incredibly fast: $16.286 billion in the first half, versus $1.163 billion in the same period last year—a 14-fold increase. Second-quarter revenue alone was $11.555 billion. Full-year revenue in 2023 was just $41 million; it has grown more than a hundredfold in three years.
The “$80.7 billion loss” is an accounting figure, not actual cash burn. The company reported a net loss of $80.748 billion in the first half, of which $72.388 billion came from changes in the fair value of convertible notes, derivative liabilities, and SAFEs. The company’s own valuation rose, requiring these financing instruments to be revalued at the new price and recorded as a loss. Excluding these non-cash items, the company was profitable on a non-GAAP basis in the first half, with adjusted net income of $2.671 billion. It also has $112.4 billion in cash and $19.3 billion in short-term investments.
The real good news is on costs: second-quarter operating profit was $974 million, with an 8% margin; a year earlier, the company was still losing money in the same quarter. Compute costs fell from more than twice revenue in the first half of last year to 55%—revenue is growing faster than that fixed bill.
The real risk is the bill still to come: long-term compute purchase commitments total hundreds of billions of dollars—about $161.2 billion for Broadcom TPUs, up to $120.5 billion for SpaceX, at least $114.5 billion for Google, at least $110 billion for Amazon, and at least $44.9 billion for Microsoft, with contracts running as far as 2045. The prospectus itself says future cash flow will remain substantially negative for a long time. The question isn’t “Will it make money?” but “Can it keep raising money?”
The customer base is impressive but concentrated: 11,900 customers pay more than $100,000 a year, including 250 that pay over $10 million, and net revenue retention is 900%. But the top five customers account for about 25% of revenue.
One easily overlooked point: the seven founders control 50.1% of the voting power through Class F shares, and a long-term public-benefit trust can also elect a majority of the board. You buy the stock and take on the risk, but you can’t change the company’s direction. The more than 90 pages of risk disclosures even mention that “AI could pose catastrophic risks.”
In short: this is a prospectus for a company with “explosive growth, improving costs, and a future staked on hundreds of billions of dollars in compute.” A $2 trillion valuation is about 60 times annualized revenue. You’re buying the expectation that “AI will become infrastructure,” not this year’s profits.
Anthropic (the parent company of Claude) publicly filed on October 26 under the ticker ANT, with plans to list on Nasdaq. It had submitted a confidential filing in June; this is the public version. The roadshow could begin as early as the week of November 9, with the listing planned for the week of November 16. The company is aiming to raise up to $100 billion at a valuation of $1.5 trillion to $2 trillion—bigger than SpaceX’s June offering ($86.3 billion raised at a $1.77 trillion valuation).
Revenue is growing incredibly fast: $16.286 billion in the first half, versus $1.163 billion in the same period last year—a 14-fold increase. Second-quarter revenue alone was $11.555 billion. Full-year revenue in 2023 was just $41 million; it has grown more than a hundredfold in three years.
The “$80.7 billion loss” is an accounting figure, not actual cash burn. The company reported a net loss of $80.748 billion in the first half, of which $72.388 billion came from changes in the fair value of convertible notes, derivative liabilities, and SAFEs. The company’s own valuation rose, requiring these financing instruments to be revalued at the new price and recorded as a loss. Excluding these non-cash items, the company was profitable on a non-GAAP basis in the first half, with adjusted net income of $2.671 billion. It also has $112.4 billion in cash and $19.3 billion in short-term investments.
The real good news is on costs: second-quarter operating profit was $974 million, with an 8% margin; a year earlier, the company was still losing money in the same quarter. Compute costs fell from more than twice revenue in the first half of last year to 55%—revenue is growing faster than that fixed bill.
The real risk is the bill still to come: long-term compute purchase commitments total hundreds of billions of dollars—about $161.2 billion for Broadcom TPUs, up to $120.5 billion for SpaceX, at least $114.5 billion for Google, at least $110 billion for Amazon, and at least $44.9 billion for Microsoft, with contracts running as far as 2045. The prospectus itself says future cash flow will remain substantially negative for a long time. The question isn’t “Will it make money?” but “Can it keep raising money?”
The customer base is impressive but concentrated: 11,900 customers pay more than $100,000 a year, including 250 that pay over $10 million, and net revenue retention is 900%. But the top five customers account for about 25% of revenue.
One easily overlooked point: the seven founders control 50.1% of the voting power through Class F shares, and a long-term public-benefit trust can also elect a majority of the board. You buy the stock and take on the risk, but you can’t change the company’s direction. The more than 90 pages of risk disclosures even mention that “AI could pose catastrophic risks.”
In short: this is a prospectus for a company with “explosive growth, improving costs, and a future staked on hundreds of billions of dollars in compute.” A $2 trillion valuation is about 60 times annualized revenue. You’re buying the expectation that “AI will become infrastructure,” not this year’s profits.
