On November 13, Anthropic CEO Dario Amodei sat down with JPMorgan Chase CEO Jamie Dimon at the bank's New York headquarters, closing out the company's record-breaking IPO roadshow. Facing investors in the room, Amodei did not dodge the sharpest question: compute is the company's "deepest business dilemma," and "there's a huge uncertainty in what's going to happen next year." The conversation came less than a week before Anthropic's planned pricing date, with a Nasdaq listing set for November 19 under the ticker ANT.
Where the $901 billion compute commitment comes from
The compute problem is a timing problem. Amodei's explanation is that AI is advancing so fast that nobody knows how much economic value today's spending will return next year, yet compute has to be bought one to two years ahead. Buy too little and customers and new technology go without resources, hurting revenue; buy too much and the company risks overextending itself financially. In sharper terms he used in February, if Anthropic started buying $1 trillion of compute from the end of 2027 and revenue fell short, being off by a single year could mean bankruptcy. Bloomberg's tally shows Anthropic has already committed nearly $901 billion to cloud, chips and data centers, with about $71 billion due by September 2027 — a figure larger than the market value of roughly two-thirds of S&P 500 companies.
How much room for error the IPO buys
Amodei offered three buffers: gross margins that are "quite good," a large cash balance, and a buy-low purchasing strategy — buy less upfront, then top up from the spot market when needed, like buying insurance. He was blunt that the compute bet is the biggest reason to go public: the more cash on hand, the wider the margin for error between too little compute and too much. According to Bloomberg, the IPO aims to raise up to $64.4 billion, plus a $26 billion concurrent private placement — $90.4 billion combined, which would make it the largest IPO ever — and it is reportedly already about six times oversubscribed. Company filings show Anthropic's annualized revenue reached $81 billion at the end of October, up $10 billion from late September, driven largely by the popularity of its coding models. The valuation has nevertheless cooled from expectations of about $2 trillion a few weeks ago to roughly $1.4 trillion, weighed down by higher borrowing costs, political resistance to data center construction, and the turmoil at Nvidia. Another report surfaced during the roadshow: Anthropic's IPO reportedly turned away orders from Chinese and Hong Kong institutions, with export controls drawing the line.
The coming week is the decisive window: pricing is expected on November 18, followed by the Nasdaq debut on November 19. Amodei said he is not fixated on a particular month or market conditions, reminding investors that the company grew from nothing to its current size in six years on an internal creed of under-promising and over-delivering. In the same conversation he said he hopes AI will cure cancer and Alzheimer's within five to 15 years — a vision that likewise came with no firm timetable. Whether the cooler valuation spreads to AI startup funding more broadly will depend on the subscription results, as discussed in After Jensen Huang's exit, is the funding tap for AI startups still open?.
