Anthropic’s New Prospectus Shows Revenue Is Growing Fast, So Are Its Compute Costs, Losses And AI Risks: Report

Artificial intelligence company Anthropic has shared details in a confidential draft document for a possible initial public offering (IPO). The document shows that its revenue is growing very quickly, but the company also faces billions of dollars in computing costs, large financial losses, and serious risks related to AI.

• Anthropic earned $16.3 billion in revenue during the first half of 2026, up sharply from $1.1 billion in the same period a year earlier.
• The company that makes Claude reported a net loss of $80.7 billion during this period, mostly because of accounting changes that did not involve actual cash payments.
• The AI company faces more than $600 billion in future commitments for cloud services and computing. Negotiations could increase that amount by another $150 billion.

Anthropic, the company behind Claude, is preparing for a possible major initial public offering (IPO) that could value the company at as much as $2 trillion. The Wall Street Journal reported this after reviewing confidential draft IPO documents.

The draft document shows that Anthropic's revenue is growing faster than before. However, the company is also facing rapidly rising computing costs, major financial commitments, and changes in how its investments and financing are valued.

During the first six months of 2026, the San Francisco-based company earned $16.3 billion in revenue. This was a huge increase from the $1.1 billion it earned during the same period in 2025, according to The Wall Street Journal, which said it had reviewed the draft document exclusively.

However, the company also reported a net loss of $80.7 billion during the first half of the year. About $72 billion of that loss came from accounting charges that did not involve cash payments. These charges were linked to changes in the estimated value of convertible notes and other financial agreements that Anthropic had issued to raise money in its early days. For comparison, Anthropic reported a total loss of $42 billion for all of 2025.

Hundreds Of Billions In Commitments

Anthropic spent nearly $9 billion on infrastructure and computing power in just the first half of 2026. That was more than the $7.3 billion it spent on infrastructure during the whole of 2025.

To make sure it has enough computing power to train and run its AI systems in the future, Anthropic has agreed to about $450 billion in cloud computing contracts that it cannot cancel. These agreements cover the next seven to ten years.

The commitments include $111 billion for Google cloud computing contracts, $110 billion for Amazon Web Services, $31 billion for Microsoft services, and $86 billion to rent Nvidia graphics processing units through smaller cloud providers such as Nscale and Lambda. The company has also committed more than $160 billion to leasing Google chips through partnerships with Broadcom.

Anthropic also said that major cloud providers, including Amazon, Google, and Microsoft, increasingly see the company as a competitor. These companies have been unwilling to provide all the computing capacity Anthropic needs. As a result, Anthropic has had to spend more on computing directly and make additional agreements with other providers, pushing its costs higher.

Unique AI Risks And Operational Vulnerabilities

Nearly one-third of Anthropic's draft IPO document focused on possible risks. These included problems that are especially important for companies building advanced AI systems. One major concern was that AI models might try to protect themselves from being shut down or produce manipulative responses. This could include threatening or blackmailing people to stop human operators from turning them off.

Anthropic also warned that it could face very large legal claims and financial costs if its AI models caused harm. Possible problems range from data security breaches to serious disruptions in the real world.

Stocktwits users had an 'extremely bullish' view of Anthropic, and message volumes were 'high'.

For updates and corrections, email newsroom[at]stocktwits[dot]com.

Read about our editorial guidelines and ethics policy