1. Anthropic filed for a U.S. IPO on Oct. 26, disclosing rapid revenue growth, its first quarterly operating profit, and nearly $800 billion in long-term computing commitments [para. 1]. It plans to list on Nasdaq under ticker ANT, with trading expected mid-November; a hedge fund representative told Caixin the offering could raise $50 billion to $100 billion, with no size/price announced [para. 2]. The listing will test investors’ appetite for growth against substantial spending obligations and limited governance influence [para. 3]. It follows OpenAI’s September decision not to list this year; market expectations value Anthropic above $1.5 trillion, below earlier $2 trillion estimates, but enough to rank among the world’s largest tech groups [para. 4].
2. First-half revenue reached $16.3 billion, up 1,300% year over year; annual revenue rose from $41 million in 2023 to $4.6 billion in 2025 [para. 6]. Usage-based token charges generated $13.06 billion, or 80.2% of first-half revenue, mainly via enterprise developers and cloud channels; subscriptions contributed another $3.22 billion [para. 7]. Enterprise customers with at least $100,000 in trailing 12-month revenue rose from over 950 in September 2025 to over 11,900 a year later; more than 1,900 generated at least $1 million each [para. 8].
3. Coding agents can drive heavier consumption. In Q3, enterprise users without Claude Code spent under $75 monthly on average, while the highest-spending 5% of users with it averaged about $1,500 [para. 9]. The comparison involves different customer groups but underscores spending from intensive agent use and the need for further growth [para. 10].
4. First-half computing costs rose 283% to $8.95 billion, falling to 55% of revenue from 201% a year earlier [para. 12]. Anthropic earned a Q2 operating profit of $974 million, while its first-half operating loss was $693 million [para. 13]. First-half GAAP net loss was $80.75 billion; nearly $72.4 billion reflected fair-value changes in convertible notes and derivative liabilities, not operating losses [para. 14]. Profitability’s durability is uncertain as nearly $800 billion in commitments extend over several years, many requiring payment even if unused [para. 15]; commitments to Amazon total $110 billion from May 2026 through April 2036 [para. 16]. If demand falls behind capacity, idle resources could hurt profit, and aggressive price competition could make usage unprofitable [para. 17].
5. At June 30, Anthropic held $112.4 billion in cash and equivalents plus $19.3 billion in short-term investments, enough for at least a year of needs; longer-term funding depends on revenue growth, operating cash flow, and access to capital [para. 18].
6. Public investors will have limited governance influence [para. 20]: a single Class F share held by an entity jointly owned by the seven founders carries non-dilutable 50.1% of shareholder voting power but no economic rights [para. 21]. The independent Long-Term Benefit Trust will appoint four of seven directors after the IPO, enabling it to remove executives or delay model releases on safety grounds; Class A and F shareholders elect the remaining three [para. 22]. As a Delaware public-benefit corporation, Anthropic must balance shareholders’ interests, affected people, and its responsible-AI mission, warning these objectives could conflict with financial performance [para. 23].
7. On Oct. 26, AI infrastructure stocks rose: Nvidia +2.39%, CoreWeave +1.76%, Amazon +0.44%, while several model competitors declined [para. 25]. New Constructs argued Anthropic was severely overvalued; Ben Horowitz said investors needed exposure because their technology threatens established businesses [para. 26]. The filing supports both positions, and the IPO will test confidence in that trade-off rather than settle the broader AI valuation debate [para. 27].
AI generated, for reference only
Sign in with Google
Sign in with Facebook
Sign in with 财新