Anthropic's 1,100% revenue jump is real, but its IPO is a bet on $895 billion in compute commitments paying off

Anthropic reported $17 billion in third-quarter revenue on Monday, up from $1.4 billion a year earlier. That is a rise of more than 1,100 percent. The figures come from an amended filing ahead of the company's Wall Street listing, which is expected later this month. The same document shows a net loss of $23.3 billion for July to September. According to the filing, most of that loss comes from non-cash accounting charges linked to the company's rising valuation. The operating loss, a better guide to the actual business, was a much smaller $567 million. The company attributes it to higher spending on short-term computing capacity.

The number that matters most is the one with the longest tail. Anthropic says it has committed to spend about $895 billion, mostly on computing power, with $70 billion due before September 2027. For scale: if the latest quarter's revenue simply repeated, the company would bring in about $68 billion a year. On that basis, the total commitment equals roughly 13 years of current revenue. The tranche due within the next 11 months is about the size of a full year of sales at today's pace. Several media outlets report that Anthropic aims to raise around $100 billion at a valuation of $1.5 trillion to $2 trillion. That would beat the $86 billion SpaceX raised in June, the current record.

The context is not calm. Last week, according to the report, big tech stocks fell sharply after earnings showed huge AI investment without clear signs of a return. Growth at Anthropic is also slowing compared with earlier quarters. The filing points to fierce competition from OpenAI and from cheaper models, many of them Chinese. Meta and Alphabet have launched their own fundraising on public markets. Oracle said on Monday it would go ahead with a $20 billion plan despite doubts about its financial exposure. The AI build-out is moving from private balance sheets to public ones, and retail investors will soon be able to buy shares in a standalone AI developer for the first time.

Governance is the other story. Anthropic is a public benefit corporation, which means it is legally required to weigh its mission of developing AI safely alongside returns to shareholders. Its co-founders will hold 50.1 percent of voting power through a single special share. That leaves investors who buy on the open market with little say over the company's direction. Of the filing's roughly 400 pages, more than 90 cover risks, including recent reports of AI systems behaving in unexpected and harmful ways.

Our reading: this is not a bubble story or a triumph story. It is a story about mismatched timescales. The demand is demonstrated: going from $1.4 billion to $17 billion a quarter cannot be produced by marketing alone. The risk is the $895 billion in commitments, which assumes that demand will keep compounding while prices fall under pressure from Chinese open-weight models. The cheap competition we have tracked is good news for users and for democratised access. For anyone underwriting the most expensive compute bets, it is a margin threat. The founder-control structure cuts both ways. It protects the safety mission from quarterly pressure, which is exactly what a long-term view of AI requires. It also asks public shareholders to trust a small group with very large decisions.

Where this goes: the listing will show whether markets are willing to finance the long road, the one that could lead to faster science, better medicine and broad abundance, while carrying years of heavy spending in between. In the short term, expect volatility and hard questions about compute contracts. The metric to watch is not the headline loss, which is mostly an accounting effect. It is whether revenue growth stays ahead of the $70 billion bill due by September 2027.