That’s according to a report Tuesday (Oct. 20) by Reuters, which viewed a draft prospectus for the company’s planned initial public offering (IPO).
Anthropic is expected to go public next month, with some investors now considering a lower valuation of around $1.2 trillion after its most recent financial disclosures, the report said.
The artificial intelligence (AI) startup’s first-half loss, widening from a $13 billion loss last year, included a $72 billion accounting charge related to a rise in the estimated value of bonds that could eventually convert into Anthropic shares, and not cash costs connected to the company’s operation, Reuters said.
Reuters had reported last month that Anthropic saw a net loss of $42 billion for 2025, with most of that figure tied to a similar accounting charge.
We’d love to be your preferred source for news.
Please add us to your preferred sources list so our news, data and interviews show up in your feed. Thanks!
The report noted that these disclosures spotlight the heavy costs of the AI race, even as investors place more scrutiny into whether they’ll see the money they’re giving to data centers and chip spending pay off.
Anthropic has pledged to spend as much as $518 billion on cloud services, computing and infrastructure over the next decade, the report said.
The company spent close to $9 billion on computing capacity and infrastructure in the first half of 2026, up from $2.3 billion a year earlier, the prospectus said.
At an investor meeting earlier this month, Anthropic told prospective IPO investors that its third-quarter revenue surged 45% to $16.8 billion from the prior quarter, Reuters added, citing people familiar with the matter.
While the numbers lessened some concerns about the steep costs of developing AI systems, some investors were let down by the slowing sequential growth rate, as revenue had more than doubled between the first and second quarter.
In other Anthropic news, PYMNTS wrote Wednesday (Oct. 21) about the draft AI safety legislation the company has been floating.
Under those rules, an AI model “could pose a foreseeable risk of $900 million in damage and still miss the bar for federal intervention.”
Instead, the company’s draft calls for the government to block a frontier model only when it presents a foreseeable, imminent risk of more than $1 billion in damage or of killing or seriously injuring more than 50 people. The threat must come from a biological weapon, a cyberattack or a model that escapes its creator’s controls.
Add as Preferred Source
