Anthropic faces a difficult timing problem: the computing capacity needed to serve customers and develop AI must be arranged well before the company can know how much revenue that capacity will generate. CEO Dario Amodei described the uncertainty shortly before the company’s planned Nov. 19 IPO, saying, “There's a huge uncertainty in what's going to happen next year.”
Capacity decisions come before the payoff is clear
At a JPMorgan Chase investor event in New York on Friday, Amodei called compute “the deepest business dilemma that the company faces.” AI technology is moving quickly, he said, while its future economic value remains difficult to predict.
The mismatch is built into the timing of the business. Anthropic needs to purchase compute one to two years in advance. By the time that capacity is ready, the company may still not know whether customer demand and revenue will be enough to justify the commitment.
That leaves Anthropic balancing two costly possibilities. If it secures too little capacity, it may not be able to serve customers or pursue new technologies, putting revenue at risk. If it commits to too much, the spending could stretch the company’s finances beyond what its business can support.
Amodei has previously described how severe that exposure could become. In February 2026, he warned that buying $1 trillion of compute starting at the end of 2027 could lead to bankruptcy if revenue fell short. He said being off by just one year could mean bankruptcy, and expressed doubt that some competitors had fully accounted for the risks.
Large commitments put the trade-off in focus
According to Bloomberg, Anthropic has committed nearly $901 billion, with most of that amount tied to compute and chips. Roughly $71 billion is due by September 2027. Bloomberg noted that this sum exceeds the market capitalization of about two-thirds of S&P 500 companies.
Those figures show why forecasting demand is not an abstract planning exercise. A long-term capacity commitment can be difficult to adjust after it is made, while the revenue that is supposed to support it remains uncertain. The company has to plan for future needs without knowing exactly how quickly its products will translate into sales.
Buying less upfront can reduce the risk of carrying unused capacity, but it may leave the company short if demand grows. Buying more can give Anthropic room to serve customers and develop products, but it raises the possibility of financial overextension. The central challenge is choosing a level of capacity that leaves room to respond in either direction.
Cash and flexible buying offer some protection
Amodei pointed to gross margins and cash on the balance sheet as ways Anthropic can manage the dilemma. He described the company’s gross margins as “quite good,” especially compared with other players, and said it has a large cash balance while seeking to raise more.
The company also buys relatively little compute upfront, according to Amodei. If business performs better than expected, it can seek additional capacity in the spot market. He compared this approach to buying insurance: Anthropic can be more protected if results disappoint, yet still invest further if demand and business performance are strong.
This flexibility does not make the underlying forecast easier. It gives the company options as conditions become clearer, while limiting how much capacity it has to lock in before it knows what customers will need. The size of its cash reserves also matters, because more available cash gives Anthropic a wider margin for error between buying too little and buying too much.
The planned IPO is intended to widen that margin
Amodei said the compute bets were the biggest reason for going public. Anthropic plans to raise up to $64.4 billion through the IPO and another $26 billion through a private placement, Bloomberg reports. Together, those amounts would make it the largest IPO ever, and the offering is reportedly about six times oversubscribed.
The fundraising is meant to give Anthropic a larger buffer if spending and revenue do not line up as hoped. The company’s annualized revenue run rate reached $81 billion at the end of October, up $10 billion from late September, according to company filings. Bloomberg says the popularity of AI products, including coding models, has helped drive that growth.
Amodei said he was not fixated on a particular month or market conditions for the IPO. He also shared advice he said had resonated with him: “never let your company be worth a single dollar more than it's really worth.” He linked that idea to Anthropic’s internal approach of “under-promise and over-deliver.”
The tension in Anthropic’s plans is clear: demand and revenue are growing, but the capacity behind that growth has to be committed in advance. The IPO would provide more cash to absorb uncertainty, while the company’s future still depends on how well its long-range compute decisions match the business that actually develops.