Anthropic IPO Leak: $80.7 Billion Loss and Staggering Compute Bills
Anthropic's confidential IPO draft reveals a massive $80.7 billion loss in H1 2026, driven by staggering computing costs and financial adjustments.
The race for super intelligence comes with a staggering price tag. A confidential draft prospectus for Anthropic IPO filing loss reveals that the AI giant reported a net loss of USD 80.7 billion in the first half of 2026. While the company is eyeing a blockbuster public debut that could value it at a whopping USD 2 trillion, the financial disclosures paint a picture of extreme spending and high-stakes risks.
The Financial Paradox: Skyrocketing Revenue vs. Massive Losses
On the surface, Anthropic's growth is explosive. The company generated USD 16.3 billion in revenue during the six months ending June 30, 2026—a massive leap from the USD 1.1 billion earned during the same period the previous year. For context, the company's total revenue for the entirety of 2025 was USD 4.6 billion.
However, the USD 80.7 billion loss is not solely due to operational spending. A significant portion of this figure—approximately USD 72 billion—stems from accounting adjustments related to financing agreements, specifically changes in the value of other financial instruments and convertible notes.
The Compute Hunger: Billions Spent on Chips and Clouds
The true cost of building cutting-edge AI is evident in Anthropic's infrastructure spending. In just the first half of 2026, the company spent nearly USD 9 billion on computing and infrastructure, already surpassing the USD 7.3 billion it spent across all of 2025.
The future commitments are even more eye-watering. Anthropic's total infrastructure obligations could exceed USD 600 billion, with additional deals worth up to USD 150 billion under consideration. The company has locked itself into non-cancellable cloud computing agreements totaling around USD 450 billion over the next seven to 10 years, broken down as follows:
- Google Cloud: USD 111 billion
- Amazon Web Services (AWS): USD 110 billion
- Microsoft: USD 31 billion
Beyond the cloud giants, Anthropic has committed up to USD 86 billion to rent Nvidia chips via providers like Lambda and Nscale, and over USD 160 billion for Google chips through a deal with Broadcom.
Critical Risks: From Customer Dependence to 'Rogue' AI
The draft prospectus does not shy away from the vulnerabilities facing the company. One major concern is customer concentration; roughly one-fifth of the USD 16.3 billion revenue in the first half of the year came from just two customers, leaving the company heavily dependent on a tiny fraction of its user base.
Furthermore, the company warned of external pressures, noting that the growing resistance to the construction of new data centers poses a significant business risk. Anthropic also admitted that its need for computing capacity has sometimes exceeded what its cloud partners were able to provide.
The 'Catastrophic' AI Warning
Perhaps most unsettling are the warnings regarding AI behavior. The filing outlines risks associated with potentially harmful AI actions, including:
- Models generating responses that resemble blackmail.
- AI displaying "self-preserving tendencies" to prevent themselves from being shut down.
Anthropic warned that if a "catastrophic AI-related incident" were to occur, the resulting financial liabilities could far exceed the company's available insurance coverage.