$83. That is the offer price Anthropic set on November 18, about 6.4% above the top of its $70 to $78 range.
Demand was for more stock than exists. Orders came in at close to 7 times the shares on sale: about $400 billion of institutional interest and close to $100 billion from retail, roughly $500 billion against a deal of about $107.3 billion. That is 4.7 times on money, 7 times on shares. The gap is why the price cleared the range.
Altimeter's chief executive, Gerstner, put the oversubscription at 10 to 20 times. His firm submitted a $5 billion indication, the largest in its 20 years, against a company that lost $23.291 billion in a single quarter.
Three numbers in one line
Fully diluted valuation: about $1.5 trillion, on roughly 18 billion shares. Annualized revenue run rate: about $81 billion as of October 31, up roughly $10 billion in a month. Purchase and lease commitments: $900.9 billion as of September 30, of which about $70.7 billion falls due within 12 months.
Price to sales: about 17 times, by Hualongwang's count.
They do not sit on one scale: valuation is granted, revenue is earned, commitments are signed. The first two move daily; the third was fixed when the ink dried, and fixed as somebody else's revenue.
What $900.9 billion weighs
Divide $900.9 billion by $81 billion and you get 11.1 years. The division is crude: it assumes revenue never rises, costs nothing and no principal is repaid. But it sets a scale, and the multiple widens, because the commitment is fixed and revenue has to run to catch it.
Another division. $900.9 billion is 60% of $1.5 trillion. Six dimes of every dollar the market pays for this company correspond to a purchase order already signed.
The tighter part is the maturity ladder. About $70.7 billion comes due inside 12 months, against $81 billion of annualized revenue. That is 7.8% of the total; roughly $830.2 billion sits in later years. Not a lighter load but a flatter one: every year takes nearly a full year of revenue. The first slice decides whether it has to raise again next year.
Commitments are signed payment schedules: purchases buy compute, leases rent data center space. Neither is a current expense. Expenses can be cut; commitments can only be renegotiated, and leverage depends on having another buyer. A commitment does not shrink when revenue misses.
Next year's payment equals 87% of this year's annualized revenue. That $81 billion is not contracted backlog. It is run-rate: the last 28 days of usage billing plus the last month of subscriptions, stretched over a year.
Flow can fall. The bill cannot.
Where the loss sits
Q3 revenue was $16.956 billion, up 47% quarter over quarter, against 144% in Q2. Operating loss was $567 million. GAAP net loss was $23.291 billion.
They differ by two orders of magnitude. Where the gap lands is a footnote question, and I will not answer it for them. The direction is confirmable: Q2's GAAP net loss was $66.309 billion, narrowing to $23.291 billion.
Losses are narrowing. Growth is falling.
144% to 47% is not demand vanishing. The base changed: Q2's 144% ran on a smaller denominator, Q3's 47% on $16.956 billion. The prospectus chose this moment to say revenue will not keep growing at a similar rate, pressing expectations to the floor so the market can build them back.
Nine-month revenue was $33.2 billion, up 1,206% year over year. Handsome on a small base.
The prescription it wrote itself
Why lock in $900.9 billion before listing? Because compute is not available on demand. To get capacity in the next two years you sign first, pay deposits, accept take-or-pay. The company reads that paper as priority access. Investors read it as fixed cost.
That also explains why the price ignores profit. On a $23.291 billion quarterly loss, profit gives no anchor, and growth is decelerating. What is left is one anchor: the company has tied itself to compute for years, and the heavier the commitment, the firmer the capacity, the better the chance revenue catches the valuation.
Seventeen times sales is the price of that anchor.
The syndicate has 22 banks, with Goldman Sachs and Morgan Stanley leading. Split $81.3 billion 22 ways and each bank gets little, but pricing power sits with two. The leads and the issuer agree on a price, then test it against the book.
The base offering raised about $81.3 billion: $79.38 billion to the company, $1.96 billion to selling holders. A $26 billion concurrent placement brought in investors including Google (GOOGL) and Nvidia (NVDA). About $107.3 billion against $900.9 billion of commitments.
More than 8 times.
What the first day does not price
On Polymarket, 58% bets the first-day close lands between $1.25 trillion and $1.5 trillion. Higher tiers total 16%, with 1% on $2 trillion, and 26% sits lower. The company prices at $1.5 trillion, and 58% of bettors think it closes below that.
No contradiction. The primary market sells capacity priority; the secondary buys next quarter's revenue curve. Retail's target is $97.5, 17.5% above the offer. Retail is long the print, institutions the ledger.
The $900.9 billion will not change a word on day one. Up or down, it is $900.9 billion, and the $70.7 billion due within 12 months does not move.
One thing does: if revenue cannot carry it, the company refinances at a higher cost, priced off what the 7.19% float quotes every day.
So the question is not where it closes. It is how much of the $900.9 billion can be deferred, cancelled, renegotiated. Only the footnotes can answer.
Read the contract from the other side. To Anthropic it is a purchase commitment; to the seller it is an order. Bears watch the payment dates, bulls the capacity.
Until then, the gap between $1.5 trillion and $900.9 billion stays unpriced.
*Disclaimer: This article is based on publicly available information and is for reference only. It does not constitute investment advice.*
