Anthropic’s giant loss puts the AI cloud bill under the microscope

Anthropic’s march towards the public markets is running into a very expensive question: how much does it really cost to build and run frontier AI?

A CTech report dated 21 October 2026, citing a draft prospectus reviewed by Reuters, says the Claude maker recorded a first-half net loss of US$80.7 billion. That is a brutal headline number, but the detail matters.

According to the Reuters account carried by CTech, most of the loss was not day-to-day operating spend. It said US$72 billion came from an accounting charge linked to an increase in the estimated fair value of convertible bonds, which may eventually convert into Anthropic shares.

The same report said Anthropic’s first-half loss was up sharply from US$13 billion in the same period a year earlier. It also said investors were reassessing the company’s valuation, with some considering a figure of around US$1.5 trillion after the latest financial disclosures.

The infrastructure bill is the other big number.

CTech, again citing Reuters and the draft prospectus, reported Anthropic has committed to spending at least US$518 billion over a decade on cloud services, computing and infrastructure. In the first half alone, the company reportedly spent nearly US$9 billion on computing capacity and infrastructure, up from US$2.3 billion a year earlier.

TechCrunch reported on 20 October 2026 that Anthropic’s cloud partnerships are getting messy, citing an S-1 filing leaked to The Wall Street Journal. That report said Anthropic believes it could spend more than US$750 billion on cloud contracts, including some deals still under negotiation.

TechCrunch also reported, based on the Wall Street Journal account of the leaked filing, that Anthropic is having trouble securing enough cloud capacity from AWS, Google and Microsoft. Those hyperscalers are increasingly seen as competitors as well as suppliers, according to the report.

To plug the gap, TechCrunch said Anthropic is leasing its own data centres and working with third-party data centre operators. The leaked filing reportedly warned those operators have a “limited operating history”, which could make them less reliable than the hyperscalers.

Revenue is rising fast, but investors appear to be watching the curve closely. CTech reported that at an October 14 meeting with prospective IPO investors, Anthropic said third-quarter revenue rose 45 per cent from the previous quarter to US$16.8 billion, citing Reuters and people familiar with the matter. It also said second-quarter revenue had more than doubled from the first quarter to US$11.6 billion.

TechCrunch separately reported the company is targeting a November IPO. CTech said Anthropic did not respond to a request for comment, and TechCrunch said it did not immediately receive a response.

Startup Daily carried a separate headline saying Anthropic posted a US$114 billion half-yearly loss and that its valuation had plunged, but the supplied extract did not include supporting detail for those figures.

For now, the story is less about whether demand for AI is real, and more about whether the infrastructure economics can keep up with the hype.