Anthropic Faces Major Financial Losses Despite Revenue Growth

AI startup Anthropic reportedly incurred losses exceeding $80 billion in the first half of 2026, as indicated by a draft IPO prospectus obtained by the Wall Street Journal. Meanwhile, the company’s revenue rose significantly, positioning it for a potential IPO that could value it at around $2 trillion.

According to the Wall Street Journal, Anthropic’s net loss stood at $80.7 billion for the six months ending June 30, 2026. This figure is notably higher than the $13 billion loss reported during the same period in the previous year. In comparison, the total losses for 2025 were $42 billion. A substantial part of this increase, approximately $72 billion, resulted from accounting adjustments associated with the fair value of convertible notes and other financial instruments. Anthropic opted not to comment on the prospectus.

In terms of revenue, Anthropic generated about $16.3 billion in the first half of 2026, a significant rise from $1.1 billion a year prior. For context, the revenue for the entire year of 2025 was $4.6 billion. However, this revenue growth came at a considerable expense, as the company expended nearly $9 billion on computing infrastructure in just the first half of 2026—surpassing the $7.3 billion spent in all of 2025 and well above the $2.3 billion during the same period the previous year.

A notable portion of Anthropic’s revenue, approximately one-fifth, was generated from just two clients. The company indicated in the prospectus that it expects to continue relying on a limited customer base, with a considerable share of its revenue closely tied to these clients.

Looking toward the future, Anthropic has an immense need for infrastructure. The company has projected that it could spend up to $750 billion in the coming years on computing resources, with ongoing negotiations for additional deals. Currently, Anthropic has already committed to roughly $450 billion in long-term cloud agreements over the next seven to ten years, most of which cannot be canceled. This includes $111 billion allocated to Google, $110 billion for Amazon Web Services, and $31 billion for Microsoft—spending that is planned to occur “regardless of actual usage.” Additionally, Anthropic plans to spend up to $86 billion renting Nvidia chips from Nscale and Lambda through to 2033 and 2034, and carries commitments exceeding $160 billion for leasing Google chips through a deal with Broadcom. Altogether, projected infrastructure costs could exceed $600 billion, with further negotiations potentially increasing that commitment by another $150 billion.

Anthropic has also reported tensions with the cloud service providers it relies on. The company has stated that AWS, Google, and Microsoft increasingly regard it as a competitor, leading to limitations on the computing resources these partners are willing to supply. In light of this, Anthropic is investing in its own data centers via long-term leases and partnerships with third-party operators—some of which have limited experience. Most of its existing data center capacity is based in the U.S., though the company is exploring expansion overseas and has signed some letters of intent. It has cautioned that political challenges to data center developments could become more organized and that its increased visibility might draw attention to projects associated with it.

A significant portion of the draft prospectus addresses various risk factors, including the possibility that Anthropic’s AI models could cause real-world harm. The company noted that its models could exhibit “self-preserving” behaviors, potentially generating responses similar to “blackmail” to avoid shutdown. Anthropic expressed concern about being held liable for harmful actions by its models—such as data breaches or security incidents—and advised that securing insurance for potential catastrophic AI-related incidents might be difficult.

As for executive compensation, CEO Dario Amodei and President Daniela Amodei, who are siblings, each received a base salary of around $655,000 in 2025. Including stock and option awards, Dario’s total compensation was approximately $18 million, while Daniela’s was about $16.5 million, as stated by MyLogIQ in the filing. These figures align with the median pay for S&P 500 CEOs, although they fall short relative to many large tech company executives. Both co-founders are committed to donating 80 percent of their personal equity in Anthropic to charitable causes, as noted in the prospectus. On the Sunday prior to the publication of this article, Anthropic revealed plans to allocate 5 percent of its shares to a new “public benefit fund” aimed at facilitating charitable contributions or directing resources to specific communities.

Anthropic and CEO Dario Amodei have been prominent figures in the AI doomsday movement, with Amodei advocating for a slowdown in AI development, a sentiment echoed by figures like OpenAI’s Sam Altman and SpaceX’s Elon Musk.

In an essay, Amodei outlined a three-step approach intended to moderate the pace of AI advancement without halting it. “To be clear, pacing does not mean stopping model training or technical progress, but ensuring companies take adequate time to align and secure their models, and for third-party evaluators to validate this,” he stated.

Anthropic claims to have already undertaken the first step by providing third-party evaluators with employee-level access to verify its safety practices and report any incidents—an action Amodei asserts does not compromise commercial advantages or the U.S.’s leadership in AI. The second step calls for leading AI firms in democratic nations to collaborate on shared safety standards. The third, more ambitious, step advocates for cooperation between democratic and authoritarian governments.

Notably, prominent AI companies are both launching powerful new models and calling for stricter regulations within the industry. David Sacks has noted that when influential figures like Amodei, Altman, and Musk request a slowdown in development, it often stems from their desire to gain control over the industry while shielding themselves from antitrust scrutiny. In response, noted author Wynton Hall has produced a bestseller intended to outline how the MAGA movement could adopt positions regarding AI that benefit society without compromising national interests to the tech elites of Silicon Valley or allowing foreign powers like China to dominate global AI advancements.