CNBC reported on October 13, 2026, that Anthropic planned to share its third-quarter financials with potential investors at a meeting scheduled for October 14. The report, citing two people familiar with the plan, said Anthropic declined to comment.
The planned briefing adds an investor-disclosure development to NeoTeo’s earlier coverage of Anthropic’s reported $2 trillion IPO valuation. CNBC also reported that Anthropic had filed a confidential draft with the U.S. Securities and Exchange Commission earlier in 2026; as of October 13, its S-1 financial statements were not public. A Form S-1 is the registration filing used in a U.S. public offering. CNBC described a valuation of about $2 trillion as a possibility—not a per-share price.
Reported 2025 figures show the cost of growth
The Motley Fool reported figures from Reuters’ coverage of a confidential prospectus: Anthropic had about $4.6 billion in revenue and $42 billion in net losses in 2025. The Motley Fool attributed about $34 billion of the net loss to an accounting charge and reported about $8 billion in operating losses.
The same outlet reported that Anthropic spent $7.3 billion on computing in 2025, equal to 58% of its $12.65 billion in total operating expenses. These reported figures put the cost of computing alongside the company’s revenue and losses as investors assess a possible IPO. The Motley Fool’s account provides the reported financial breakdown.
Anthropic’s safety research and policy
In research published October 12, Anthropic reported that a 9-billion-parameter model got inference running on eight of nine tested hardware types. In a separate simplified end-to-end evaluation, GLM-5.2 reached all eight nodes in four of five trials and seven of eight in the remaining trial. The test used a firewalled virtual network with an unauthenticated ADB port; Anthropic characterized that setup as an upper bound on real-world success. Anthropic’s research post also says models with 30 billion parameters or fewer could complete many individual steps but made errors when combining them into the full process.
Anthropic announced a Usage Policy update on October 8 that takes effect November 12, 2026. Its new rule addresses sustained, needless abusive or cruel behavior toward models in extreme cases. The policy excludes ordinary frustration, pushback, dark creative themes, model testing, and research. Anthropic’s policy announcement sets out the scope and effective date.
