Sarah Guo 与 Elad Gil：万亿美元公司、Token 预算与监管捕获Sarah Guo & Elad Gil: Trillion-Dollar Companies, Token Budgets, and Regulatory Capture
The Takeaway：AI 浪潮是断续平衡，而非持续爆炸；未来 3-5 年不太可能再涌现大量万亿美元公司，真正稀缺的是能快速达到 500-1000 亿美元营收的市场。
No Priors 主持人 Sarah Guo 与投资人 Elad Gil 深入讨论了当前 AI 周期的现实。过去五年，OpenAI、Anthropic、SpaceX 等从接近零增长到万亿市值，属于历史罕见的“寒武纪大爆发”。Elad 指出，技术史更多是 punctuated equilibrium：社交、SaaS、云、加密货币都经历过浪潮后进入整合期。现在很多人把每个机器人、材料或能源创业都想象成下一个万亿公司，但速度与规模被严重混淆。物理世界公司受限于落地足迹，很难在三五年内完成那种跃迁。
两位同时观察到，优秀创始人出现“怯懦”趋势：因害怕与实验室正面竞争，转向硬件、美国动力主义或极度细分的利基市场。Elad 认为这是负面信号，许多市场实验室其实不会自然吃掉，产品与分发仍可硬刚。关于退出，他们建议董事会每六个月进行一次非情绪化讨论：公司是否还在能力提升、成本下降的正确一侧？如果无法跨越，就应认真考虑出售。时间是最大的机会成本——“你人生中最有生产力的年华正在被押上”。
Token 预算成为新权力法则。少数顶尖研究员驱动 80% 进展，compute 会优先分配给他们。企业从“人人用 AI”转向衡量 return on invested tokens。监管方面，他们以法国 70% 核电却事故极少、美国因 1970 年代安全游说几乎停建反应堆为例，警告过度安全导向会重演能源与生物技术领域的停滞。Elad 说：“历史上对大型行业，我们往往走得太远。科技之所以快速成功，正是因为监管较轻。”
“70% of France is still nuclear... Where are all the accidents... nothing. US is 18%, and we haven't built a reactor in forty years.”
No Priors 主持人 Sarah Guo 与投资人 Elad Gil 深入讨论了当前 AI 周期的现实。过去五年，OpenAI、Anthropic、SpaceX 等从接近零增长到万亿市值，属于历史罕见的“寒武纪大爆发”。Elad 指出，技术史更多是 punctuated equilibrium：社交、SaaS、云、加密货币都经历过浪潮后进入整合期。现在很多人把每个机器人、材料或能源创业都想象成下一个万亿公司，但速度与规模被严重混淆。物理世界公司受限于落地足迹，很难在三五年内完成那种跃迁。
两位同时观察到，优秀创始人出现“怯懦”趋势：因害怕与实验室正面竞争，转向硬件、美国动力主义或极度细分的利基市场。Elad 认为这是负面信号，许多市场实验室其实不会自然吃掉，产品与分发仍可硬刚。关于退出，他们建议董事会每六个月进行一次非情绪化讨论：公司是否还在能力提升、成本下降的正确一侧？如果无法跨越，就应认真考虑出售。时间是最大的机会成本——“你人生中最有生产力的年华正在被押上”。
Token 预算成为新权力法则。少数顶尖研究员驱动 80% 进展，compute 会优先分配给他们。企业从“人人用 AI”转向衡量 return on invested tokens。监管方面，他们以法国 70% 核电却事故极少、美国因 1970 年代安全游说几乎停建反应堆为例，警告过度安全导向会重演能源与生物技术领域的停滞。Elad 说：“历史上对大型行业，我们往往走得太远。科技之所以快速成功，正是因为监管较轻。”
“70% of France is still nuclear... Where are all the accidents... nothing. US is 18%, and we haven't built a reactor in forty years.”
The Takeaway: The AI wave is punctuated equilibrium, not continuous explosion. Multiple new trillion-dollar companies in the next 3-5 years are unlikely; the real scarcity is markets that can reach $50-100B revenue quickly.
No Priors hosts Sarah Guo and investor Elad Gil unpack the current AI cycle. The last five years saw OpenAI, Anthropic and SpaceX go from near-zero to trillion-dollar valuations—an unprecedented inflection. Elad frames technology history as punctuated equilibrium: social, SaaS, cloud and crypto each had waves followed by consolidation. Many now treat every robotics, materials or energy startup as the next trillion-dollar company, conflating market size with velocity. Physical-world companies face footprint constraints that make such leaps in three to five years improbable.
Both note a rising meekness among strong founders: fear of lab competition pushes them toward hardware, American Dynamism or ultra-niche applications. Elad sees this as negative—many markets the labs will not naturally consume, and product plus distribution still win. On exits they recommend a non-emotional board conversation every six months: is the company still on the right side of falling costs and rising capabilities? If not, seriously consider selling. Time is the largest opportunity cost—“your most productive years are on the line.”
Token budgets are the new power law. A few dozen researchers drive 80% of progress; compute is preferentially allocated to them. Companies are shifting from “everyone use AI” to measuring return on invested tokens. On regulation they cite France’s 70% nuclear power with virtually no accidents versus the U.S. safety lobby that halted reactor construction for decades, warning that safety-only focus repeats the stagnation seen in energy and biotech. Elad: “Historically, for big industries, we’ve gone too far. The reason tech has been so successful so quickly is because it’s been lightly regulated.”
“70% of France is still nuclear... Where are all the accidents... nothing. US is 18%, and we haven't built a reactor in forty years.”
查看原文 →
No Priors hosts Sarah Guo and investor Elad Gil unpack the current AI cycle. The last five years saw OpenAI, Anthropic and SpaceX go from near-zero to trillion-dollar valuations—an unprecedented inflection. Elad frames technology history as punctuated equilibrium: social, SaaS, cloud and crypto each had waves followed by consolidation. Many now treat every robotics, materials or energy startup as the next trillion-dollar company, conflating market size with velocity. Physical-world companies face footprint constraints that make such leaps in three to five years improbable.
Both note a rising meekness among strong founders: fear of lab competition pushes them toward hardware, American Dynamism or ultra-niche applications. Elad sees this as negative—many markets the labs will not naturally consume, and product plus distribution still win. On exits they recommend a non-emotional board conversation every six months: is the company still on the right side of falling costs and rising capabilities? If not, seriously consider selling. Time is the largest opportunity cost—“your most productive years are on the line.”
Token budgets are the new power law. A few dozen researchers drive 80% of progress; compute is preferentially allocated to them. Companies are shifting from “everyone use AI” to measuring return on invested tokens. On regulation they cite France’s 70% nuclear power with virtually no accidents versus the U.S. safety lobby that halted reactor construction for decades, warning that safety-only focus repeats the stagnation seen in energy and biotech. Elad: “Historically, for big industries, we’ve gone too far. The reason tech has been so successful so quickly is because it’s been lightly regulated.”
“70% of France is still nuclear... Where are all the accidents... nothing. US is 18%, and we haven't built a reactor in forty years.”