kaifaai.app AI 快讯 · 2026-10-28

Anthropic提交335页IPO招股书，7名联创控50.1%投票权

Anthropic正式递交IPO文件，被称“史上最大AI公司IPO”和“史上最亏AI公司”，7位联合创始人仍掌握超半数总投票权。

Anthropic已向美国SEC正式提交335页S-1招股说明书，披露其2023年营收达2.1亿美元（同比增长超400%），但净亏损扩大至12.2亿美元；客户中年消费超10万美元的企业超200家，其中超半数同时使用Claude平台与第一方应用。治理结构上，7名联合创始人通过特殊股份安排锁定50.1%的总投票权，确保长期技术路线主导权，该设计远超常规双层股权架构强度。对早期AI创业者而言，这标志着资本市场正接受‘高亏损+强控制+重使命’的新型AI公司范式——无需盈利即可上市，但必须用可验证的治理机制锚定长期价值。团队应优先构建不可稀释的技术决策权结构（如协议层控制、模型权重托管条款），而非过早让渡投票权换取融资；同时需在产品设计中嵌入‘企业级交叉使用钩子’（如API+UI+Agent协同入口），以复刻Anthropic跨产品线绑定客户的增长路径。

Anthropic files 335-page IPO prospectus; 7 co-founders retain 50.1% voting control

Anthropic's 335-page S-1 filing reveals $210M in 2023 revenue (up >400% YoY) but a $1.22B net loss; over 200 enterprise customers spend >$100K annually, and more than half use both the Claude Platform and Anthropic's first-party apps. Its governance structure grants the 7 co-founders super-voting shares securing exactly 50.1% of total voting power — a deliberate, legally embedded safeguard against mission drift. For early-stage AI founders, this signals that public markets now reward durable technical sovereignty over near-term profitability. Rather than optimizing for valuation multiples, teams should prioritize enforceable governance mechanisms — e.g., on-chain model access logs, irrevocable inference policy clauses, or multi-sig deployment keys — and design product architecture to drive cross-product usage (e.g., shared context stores, unified agent registries) that locks in enterprise adoption before scale.

来源：量子位 — 刚刚，Anthropic公开335页招股书：史上最大IPO，史上最亏AI公司