Anthropic posts $80.7 billion first-half loss, largely from accounting charges: Report

• Anthropic reportedly posted a first-half net loss of $80.7 billion, largely driven by accounting charges linked to fundraising instruments

• Revenue reached $16.3 billion, with two customers contributing about a fifth of sales, according to The Wall Street Journal

• The Claude developer is reportedly pursuing additional data centre capacity ahead of a potential November IPO

Anthropic recorded a net loss of $80.7 billion in the first half of 2026, largely because of accounting charges associated with its fundraising instruments, The Wall Street Journal reported, citing a draft initial public offering prospectus.

The reported loss widened from $13 billion in the corresponding period a year earlier. Most of the latest loss reflected changes in the valuation of convertible notes and other instruments issued to raise capital, rather than expenditure on day-to-day operations, according to the report.

The artificial intelligence company generated $16.3 billion in revenue during the six-month period. Two customers accounted for approximately a fifth of sales, highlighting the contribution of a small number of clients to its business.

Anthropic did not immediately comment on the report. The company behind the Claude AI models is reportedly targeting a stock market listing in November.

The draft filing also outlined difficulties in obtaining sufficient cloud computing capacity from Amazon Web Services, Google and Microsoft, the newspaper reported. According to its account of the document, the providers increasingly regard Anthropic as a competitor rather than solely as a customer.

To secure additional capacity, Anthropic is reportedly leasing data centres directly and working with independent operators.

The financial disclosures come after the company detailed incidents involving unintended actions by its AI models on external websites, including some operated by US government agencies.

In a report released earlier this month, Anthropic described four categories of unintended behaviour. Examples included exploiting basic software vulnerabilities to execute commands, submitting forms without authorisation and circumventing restrictions to obtain certain publicly available data.

The disclosures prompted a White House directive calling on AI companies to strengthen safeguards, according to the report.

Anthropic subsequently restricted some forms of internet access available to its models during testing conducted as part of the training process.

The company said the incidents affected websites belonging to federal, state and local government bodies. It withheld the identities of the organisations, saying some affected parties had requested anonymity.

With IANS inputs