Anthropic swings back to operating loss as $895 billion compute bill looms over IPO
Anthropic’s revenue jumped 47% to $17 billion (€15.1bn) last quarter, but the developer of the Claude AI models slipped back into an operating loss and revealed close to $900 billion (€804bn) in commitments, mostly for computing, numbers investors will weigh ahead of its Nasdaq listing.
Published on 3 November 2026 - 14:24 GMT+1
Anthropic, the US company behind the Claude AI models, brought in $17 billion (€15.1bn) in the third quarter, up 47% on the previous quarter and more than twelve times its sales a year earlier.
However, it swung to an operating loss and disclosed about $895.4 billion (€800bn) in purchase and lease commitments, mostly for computing, in an updated prospectus filed with US regulators on Monday, ahead of its stock market debut.
The operating loss of $567 million (€506mn) compares with a $974 million (€870mn) profit in the second quarter. Anthropic blamed it partly on $5 billion (€4.46bn) spent on short-term computing capacity, bought when demand outstripped its long-term deals, which the filing said "currently costs significantly more".
It also booked a $2 billion (€1.78bn) non-cash charge for a programme matching staff donations.
On an adjusted basis, operating income came to $2.5 billion (€2.23bn), giving a margin of 15%, down from 19% in the previous quarter.
Its net loss reached $23.3 billion (€20.8bn), against $12.1 billion (€10.8bn) a year earlier.
In the first nine months, Anthropic lost $104 billion (€92.9bn) on revenue of $33.2 billion (€29.6bn), but $94.3 billion (€84.2bn) of that came from non-cash charges as convertible notes, debt that can be swapped for shares, naturally rose in value along with the company.
Growth has also cooled after sales more than doubled from the first quarter to the second, and Anthropic had already shared preliminary figures with investors in October pointing to a slowdown.
The filing estimated that 5% to 7% of the nine-month revenue came from customers breaking its usage rules or commercial terms, warning that its crackdown "may reduce our overall revenue and growth rates in future periods".
Anthropic also said it "may imminently meet" its own threshold for "dramatically accelerated progress attributable to AI", in which AI helps build more capable AI, a process it warned "could also pose a considerable safety risk", though it has not yet designated any current model as crossing it.
What it means for the IPO
Anthropic made its prospectus public on 26 October and plans to trade on Nasdaq under the ticker ANT, though it has yet to set a price range.
US securities rules require companies to publish their prospectus at least 15 days before starting a roadshow. That means Anthropic could start marketing its shares as soon as next week and list from the week of 16 November.
Anthropic's commitments are almost seven times the $131.8 billion (€117bn) it held in cash and short-term investments at the end of September, though they stretch over many years.
About $70 billion (€62.5bn) falls due within 12 months and around $450 billion (€402bn) by the end of 2030, while some $214 billion (€191bn) can be cancelled or depends on conditions being met.
Some investors have reportedly lowered their valuation expectations to around $1.5 trillion (€1.33tn), from the roughly $2 trillion (€1.78tn) floated earlier, after a sell-off in technology stocks fuelled by worries over AI spending.
That would still be well above the $965 billion valuation set in its private funding round back in May.
