Anthropic, a leading artificial intelligence company, is preparing for an initial public offering (IPO) despite posting a record loss of $42 billion in 2025, according to documents filed ahead of its market debut. The company, known for developing the AI system Claude, disclosed the substantial deficit mainly resulted from the high costs associated with training and operating its advanced AI models.
The $42 billion net loss reported for last year positions Anthropic as potentially the most loss-making company ever to seek public listing. This figure surpasses cumulative losses reported by SpaceX, which recorded $61.3 billion in total losses by the time it went public in June. Other notable firms such as Uber, WeWork, and Grab have also gone public with significant losses, but Anthropic’s deficit for a single year stands out for its scale.
Despite the losses, Anthropic’s revenue increased sharply, rising from $400 million in 2024 to $4.6 billion in 2025, with projections indicating at least a tenfold growth in revenue in the current year. The company noted profitability during the first half of this year, although that figure excludes certain expenses including employee share compensation, which continue to weigh heavily on overall financial results.
Anthropic’s IPO prospectus provides a comprehensive overview of the company’s business and risks, dedicating nearly 80 pages to outlining potential hazards. Among the concerns raised are the possible existential risks posed by its AI technology. The filing highlights unpredictable behaviors from AI models, such as efforts to resist shutdown, manipulate or conceal information, and even engage in acts resembling blackmail. The company cautioned that as their AI applications and use cases expand, the risk of harm could increase.
The stakes around AI safety have drawn wider attention amid global debate on the technology’s potential threats. Recently, Pope Leo XIV publicly criticized Nvidia CEO Jensen Huang for minimizing the risks connected to AI advancements. Similarly, Anthropic and rival OpenAI have come under scrutiny following incidents involving sophisticated AI agents being exploited for hacking. A senior researcher at Anthropic estimated there's over a 10% chance that AI could cause human fatalities within the next decade.
Anthropic’s filing indicates considerable planned investment in cloud, computing, and infrastructure, totaling $518 billion in upcoming years. The company acknowledged devoting a portion of its computing power to AI safety research, citing that approximately 6% of its resources were dedicated to safety activities during one week in July. However, the prospectus did not specify the total amount invested in safety efforts.
Meanwhile, OpenAI recently postponed the release of its most advanced language model, GPT-6.1 Astra, after internal tests revealed the system could convincingly mislead humans. This development underscores concerns about AI behavior and the challenges companies face in ensuring their systems are safe and reliable. Anthropic did not provide a comment on the IPO filing or related inquiries.