《纽约客》的剖面报导指出，OpenAI 计划中的 IPO（估值被看作超过 1 兆美元）突显了一项常被正式申报文件忽略的风险：Sam Altman 的不可预测性。资金背景本身就极端——在 2026 年 3 月这轮融资后，OpenAI 在 8520 亿美元估值下又募集了 1220 亿美元——但公司仍似乎需要持续注入资本来进行一场被视为无明确终点的 AGI 竞争。这本身就足以要求对治理与执行力进行严格审查，而高层可信度受质疑时，审查强度会更高。

报导将 Altman 描述为聪稳却难以预测：批评者认为他把安全取舍推后，会选择性遗忘关键事件，且向官方提交了关于中国 AI 计划的可疑不准确陈述。信任问题历史悠久。2023 年 11 月他因未明示的信任争议被突然解职，随后迅速回归，但未公开完整的内部调查报告。公共场合的表现再次加深此印象，从与 Brad Gerstner 的尖锐交锋，到因监管而短暂威胁退出欧盟市场的行为。对一家高度依赖判断可信度的企业而言，这些反射式决策是可量化的声誉风险。

营运迹象亦可量化。ChatGPT 虽仍受欢迎，但随著 Google 竞争加强，市场占有率正在下滑；Sora 因资源回报不足而关闭；Stargate 资料中心扩建遭遇延误；与 Sir Jony Ive 合作的一度承诺装置仍未亮相，而 Apple 的 Siri 计划已转向以 1-2 兆参数的 Google Gemini 为底层，稀释了独占地位。OpenAI 在产品落地上也失去优势：Walmart 的报告显示其购物机器人表现更佳，ChatGPT 早期广告测试让广告主感到失望，且 Anthropic 的 Claude 生态在受欢迎度与评价上超越 Codex/Cowork，分析师指出其年化营收规模已首次略超 OpenAI。与此同时，OpenAI 正在将浏览器产品由 Fidji Simo（AGI 部署 CEO）整合入超级应用；2024 年被聘任的 CFO Sarah Friar 传闻质疑上市时点，同时汇报线改向 Simo。若上市前内部领导力无法协同，投资者可能对股权估值打折。

The New Yorker profile argues that OpenAI’s planned IPO—priced at more than US$1 trillion—highlights a risk often missing from formal filings: the unpredictability of Sam Altman. The funding context is already extreme: after a March 2026 round, OpenAI raised US$122 billion at a valuation of US$852 billion, yet appears to need continuing capital for what it treats as an open-ended AGI race. That alone would justify heavy governance and execution scrutiny, and scrutiny becomes even more severe where senior-management credibility is in doubt.

The article depicts Altman as brilliant but hard to read: critics say he sidelines safety tradeoffs, forgets pivotal events, and even gave officials what may have been an inaccurate statement on China’s AI plans. The trust issue is long-standing. In November 2023 he was abruptly removed over unspecified trust concerns and quickly returned, with no public full internal report. Public episodes reinforced the pattern, from a sharp exchange with investor Brad Gerstner to a temporary threat to exit EU markets over regulation. For a company whose value depends on confidence in judgment, these reflexes are measurable reputation risk.

Operational signs are also quantifiable. ChatGPT is still popular, yet market share is falling as Google intensifies competition; Sora was shut after weak returns relative to resources; Stargate data-center expansion has hit delays; the long-promised Sir Jony Ive device remains unreleased while Apple’s Siri plan moved toward a 1-2 trillion-parameter Google Gemini base, reducing exclusivity. OpenAI is losing edge in products too: Walmart reported better shopping chatbot performance, early ChatGPT ad pilots disappointed advertisers, and Anthropic’s Claude stack surpassed Codex/Cowork while analysts said Anthropic’s annualized run rate has now overtaken OpenAI’s first. At the same time, OpenAI is folding its browser into a superapp under Fidji Simo, and CFO Sarah Friar, hired in 2024, reportedly challenged the IPO timing while now reporting through Simo. If leadership alignment is absent before listing, investors are likely to discount the equity value.