Anthropic recorded a shocking loss of 80 billion yuan in the first half of the year, and its computing power expenditure is projected to reach 750 billion yuan in the next few years.

As per latest documents,anthropic The net loss in the first half of this year exceeded US$80 billion, but this was mainly related to accounting recognition expenses. It also revealed that the expenditure on ensuring more computing power could be as much as US$750 billion over the next few years. Furthermore, the company also faces several risks as its revenues grow rapidly.

wall Street The company’s net loss in the first half of the year widened to more than US$80 billion from US$13 billion in the same period last year, the daily reported, citing Anthropic’s draft prospectus for the initial public offering (IPO). The main reason was related to the recognition of accounting expenses related to changes in the value of financial instruments used by the company to raise funds, which amounted to US$72 billion.

Anthropic also revealed that it could invest up to US$750 billion over the next few years to acquire more computing power, including transactions that are still being negotiated.

These numbers highlight the high cost of the AI ​​race. At the same time, investors are also paying more attention to whether the huge amounts of money that companies invest in chips and data centers will ultimately bring corresponding returns.

Revenue jumped sharply in the first half of the year, but faced risks due to small customer base

The report reported that Anthropic’s revenue in the six months to the end of June rose to US$16.3 billion from US$1.1 billion a year earlier; AI infrastructure spending nearly quadrupled to nearly US$9 billion.

Anthropic also outlines a wide range of risk factors, including how to prevent AI models from causing harm and how to avoid legal liability. Potential risks ranging from advanced cyber attacks to out-of-control models that can be used for ransom, even to a small customer base, are all listed.

The company’s revenue growth is deeply tied to a handful of big customers, as nearly a fifth of its $16.3 billion in revenue in the first half of this year came from just two customers. According to the draft prospectus, Anthropic expects a “significant portion” of its revenue in the short term will still come from a “limited number of customers.”

Huge cloud and chip contracts tied to several years in the making and huge spend promised

In terms of expenditure, Anthropic will have to pay about US$450 billion in cloud computing contracts over the next seven to ten years, including the use of computing services from Google, Amazon AWS and Microsoft, as well as leasing Huida chips from AI cloud service providers Nscale and Lambda. In addition, through the agreement with Broadcom, Google chip leasing expenses could reach US$160 billion, and other computing resource transactions still under negotiation could add up to US$150 billion in committed expenditures.

Anthropic did not respond to a request for comment, and Reuters could not immediately independently verify the report. Based on a series of information obtained from Anthropic’s IPO listing prospectus, Reuters first reported in September that the company would incur a net loss of US$42 billion in 2025.

Four people familiar with the matter previously told Reuters that some existing and potential investors were reevaluating Anthropic’s valuation after attending an investor day event last week. Three of them said the latest discussions valued the company at about $1.5 trillion, far below the market’s previous expectation of about $2 trillion.

If the response to Anthropic’s IPO is not as good as expected, it could drag down the overall AI market.