Anthropic CEO Amodei admits huge uncertainty about next year, days before a record IPO
Anthropic CEO Amodei admitted "huge uncertainty" about next year at a JPMorgan event on November 13, calling compute the company's deepest dilemma.

Anthropic CEO Amodei said on Friday, November 13 that “there’s a huge uncertainty in what’s going to happen next year,” speaking with JPMorgan CEO Jamie Dimon at a JPMorgan investor event in New York. Per The Decoder’s report of the conversation, Amodei also called compute purchasing “the deepest business dilemma that the company faces” — days before the planned November 19 listing of the Claude maker’s record IPO (per press reporting; the company has not confirmed the date).
Key points
- At the New York event on November 13, Amodei told Jamie Dimon there is “huge uncertainty” about next year and called compute purchasing the company’s “deepest business dilemma” (The Decoder, citing the conversation transcript).
- Bloomberg’s figures: Anthropic has committed nearly $901 billion, mostly to compute and chips, with roughly $71 billion due by September 30, 2027; the IPO targets up to $64.4 billion plus about $26 billion in private placement, and orders are running at roughly six times the shares offered.
- The company’s own run rate: about $81 billion in annualized revenue at the end of October, up roughly $10 billion from late September. Pricing is still planned for Wednesday, November 18, with trading Thursday — nothing is final.
Background
The remark lands on top of a fast-moving filing story we have covered in three steps: when Anthropic’s S-1 went public on October 26, compute commitments stood at about $417 billion; the November 2 S-1/A amendment doubled that to $895.4 billion and showed Q3 revenue of $16.96 billion with operations back in the red; and by the time the roadshow reportedly kicked off last week, buyer valuation talk had slid from roughly $2 trillion in September to $1.1–1.3 trillion. Bloomberg’s number has crept up again since: aiweekly reads the filing at $900.9 billion, above the November 2 figure. The dilemma itself is not new. The Decoder recalls that in February, Amodei said he could buy $1 trillion of compute starting at the end of 2027 and go bankrupt if revenue fell short — adding that some competitors “don’t really understand the risks they’re taking.”
The facts
The checkable numbers from the conversation, separated by attribution; the raise, the oversubscription and the price range are all pre-pricing press figures, not confirmed by the company.
| Item | Figure | Attribution |
|---|---|---|
| Compute commitments total | about $900.9 billion, running to 2045, mostly compute and chips | via Bloomberg; aiweekly cites the S-1 filing |
| Due by September 30, 2027 | about $70.7 billion | aiweekly; The Decoder rounds to $71 billion |
| Scale comparison | the total exceeds the market cap of about two-thirds of S&P 500 members | Bloomberg, via The Decoder |
| IPO raise target | up to $64.4 billion, plus about $26 billion private placement | Bloomberg/The Decoder; Briefs |
| Oversubscription | about six times | Bloomberg, via Briefs and The Decoder |
| Annualized revenue | about $81 billion as of end of October, up ~$10 billion from late September | company figure (The Decoder cites filings; Briefs a company statement) |
| Price range | $70–78 per share, implying up to about $1.4 trillion | aiweekly, citing the filing |
| Planned timing | price November 18, list November 19 | Business Insider and aiweekly, both as plans |
| Cash buffer | over $95 billion cash plus about $36 billion short-term investments (Sept 30) | prospectus, via Business Insider |
The mechanics of the dilemma, in Amodei’s own framing: compute has to be ordered one to two years ahead, and nobody knows how fast the technology converts into business value. Buy too little and you cannot serve customer inference, which costs revenue; buy too much and “you can financially overextend yourself.” His answer is to buy on the low side — “we actually tend to buy on the low side” — and fill the gap on the spot market, which he likened to buying insurance and which “so far has always been available.” The buffers: gross margins he called “quite good,” plus that $95 billion cash pile. aiweekly reads the guidance as planning to roughly the 10th-percentile growth case, and The Decoder notes Amodei likened “under-promise and over-deliver” to an internal model.
What others say
Bloomberg’s comparison is the bluntest: $900.9 billion of commitments outweighs the market value of about two-thirds of the S&P 500. aiweekly pushes the analysis a step further — the filing effectively prices a low-demand scenario, with commitments laid out around the 10th percentile and the bet that spot capacity stays available, but “the filing never names a backstop if that spot-market assumption breaks.” It also notes the bookbuilding excludes investors from mainland China and Hong Kong, and that the $70–78 range’s roughly $1.4 trillion still sits below the ~$2 trillion floated in September. Business Insider’s summary of the strategy: buy “relatively little,” keep cash for the spot market — in Amodei’s words, “for periods of time where things don’t go as well, we’re well covered, and we’re well insulated.” On the IPO itself Amodei sounded unhurried: he is not fixated on a specific month or market conditions, and the advice he said resonates most is to “never let your company be worth a single dollar more than it’s really worth.” The counterpoint sits with buyers: last week’s roadshow talk was $1.1–1.3 trillion, the filing now reaches for $1.4 trillion — the pricing tug-of-war is live, and a six-times oversubscribed book is the sell side’s strongest card.
Our take
Our read: the value of this appearance is not the word “uncertainty” — it is a CEO turning supply risk into an explicit operating method. Low-side buying plus a spot-market hedge is now the stated doctrine, not a stopgap. If you build on the Claude API, treat peak-season rationing and queues as the base case to plan around, because Anthropic itself hedges against demand outrunning its low-percentile plan. If you are looking at the IPO, keep the two big numbers apart: $900.9 billion is a multi-year commitment total (about $70.7 billion of it due in the year to September 2027), while $81 billion is a company-reported, unaudited one-month annualized run rate — dividing one by the other proves nothing. What you can check is the Q3 ledger we recorded on November 3: operations are already in the red. The honest catch: every IPO figure here is pre-pricing reporting and can move before the November 18 pricing window, and “huge uncertainty” doubles as expectation management — it caps the story if demand disappoints and pre-earns credit if growth holds. After listing, quarterly disclosures turn commitment execution into something auditable, which is the most practical thing this IPO changes for the industry.
What to watch
- November 18 (Wednesday): the planned pricing window (Business Insider, aiweekly); a final EDGAR amendment that moves the commitments number again is worth a check.
- November 19 (Thursday): the planned listing day, consistent with the pre-Thanksgiving target we recorded last week.
- Open question: aiweekly’s missing backstop — if the spot market tightens before the roughly $70.7 billion 2027 payment peak, who absorbs it? The filing does not say.