A Better Newspaper

Anthropic PBC, the AI company behind the Claude model family, is preparing for a potential IPO and has been briefing prospective investors on its financials. Ahead of the offering, Anthropic disclosed that revenue grew 45% sequentially (nearly 50% according to another report) to $16.8 billion in the third quarter, and that it achieved profitability on an adjusted basis for the first time (Financial Times, via SiliconANGLE, October 14-15). The disclosures were made at a private meeting between Anthropic executives and investors interested in buying pre-IPO shares (SiliconANGLE, October 15). The Financial Times noted that this pace of growth is slower than some had anticipated, amid broader fears about an AI trade bubble rattling markets (FT, October 15). Separately, Anthropic has offered Blackstone a double-digit yield as part of its latest chip financing arrangement (FT, October 17). The report notes that high borrowing costs are complicating Anthropic's goal of achieving investment-grade credit status (FT, October 17). This financing detail suggests the company is paying a steep premium to fund compute infrastructure even as it touts improving profitability to IPO investors — a tension likely to draw scrutiny from prospective shareholders. Anthropic has also said it will commit 5% of its shares to a 'public benefit' fund, which could be used for donations to NGOs and 'direct distributions to specific populations' (Al Jazeera, October 19; FT, October 19). The FT frames this move as coming amid growing public backlash over AI safety and economic concerns (FT, October 19). One commentary source puts the figure at 20% of shares donated to the Anthropic Public Benefit Fund, a discrepancy with the 5% figure reported elsewhere — the exact percentage remains unclear across sources (Substack/chatgptiseatingtheworld, October 19, vs. Al Jazeera/FT, October 19). SiliconANGLE reports the pledge is for "at least" an unspecified amount of shares worth billions of dollars, framing it alongside a separate, unrelated pledge by Elon Musk to gift SpaceX shares to every child in America, which Musk announced via X shortly after Anthropic's disclosure (SiliconANGLE, October 19). The move comes as the company courts pre-IPO investors and positions itself around its public-benefit corporation structure. In a separate legal development, the D.C. Circuit reportedly sided with the Department of War (per the source's characterization) in classifying an Anthropic model as a national security risk under 41 U.S.C. § 4713, a ruling the source describes as contrary to an earlier decision on 'supply chain risk' under 10 U.S.C. § 3252 by a Northern California court (Substack/chatgptiseatingtheworld, September 25). This claim comes from a single newsletter source and has not been corroborated by mainstream outlets; it should be treated as unverified pending confirmation. On the product side, Anthropic released Claude Haiku 5.5, which cuts list prices by 90% versus prior pricing, adds adaptive thinking, and introduces a 1M-token context window (KDnuggets, October 19). Anthropic also released a product called Claude Design, which reportedly triggered market speculation that it could threaten Figma's design-tool business; Figma CEO Dylan Field pushed back on this narrative, arguing "it's easy to mistake model progress for product progress" and noting that building a tool like Figma has taken a long time to get right (YouTube/Figma CEO interview, October 20). Anthropic has also drawn public criticism from venture capitalist Ben Horowitz, who said Anthropic and Bill Gates represent 'the most dangerous thing in AI' due to their influence on AI-risk discourse and political credibility (YouTube/Ben Horowitz interview, October 18). In competitive developments, diffusion language model startup Inception AI Inc. released its flagship model Mercury 3, claiming it beat Anthropic's Claude Haiku 5.5 (and OpenAI's GPT-6 Luna) on eight of 12 benchmark tests at four to eight times their speed, per Inception's own benchmarks (SiliconANGLE, October 20). This is a competitor's self-reported claim and has not been independently verified. Separately, Anthropic published research describing methods to use Claude to reduce the attack surface of code, a security-focused research effort from the company's own research team (Anthropic, October 21). Anthropic's IPO process and financial disclosures are likely to be closely watched as a bellwether for AI sector valuations and investor sentiment.