人工智能正在借由加速初创公司的成长，来挑战传统的风险投资幂律，促使 Benchmark、Thrive Capital、Andreessen Horowitz 和 Sequoia 等大型公司募集高达先前资金规模两倍的成长阶段基金。虽然早期投资传统上能获取最高的报酬，但后期押注现在提供了快速的倍数增长。例如，在 2025 年初以 600 亿美元估值支持 Anthropic 的投资人，现在持有的股份价值已达其初始投资的 15 倍。这与历史上的成长阶段报酬形成对比，例如 Facebook 在 2009 年以 100 亿美元的估值募集资金，随后在 2012 年进行了 1000 亿美元的首次公开募股。

这种加速是由前所未有的销售增长所推动的，Anthropic 本月有望超越 500 亿美元的年化收入运行率，将其估值推向接近 1 兆美元。Accel 合伙人 Miles Clements 指出，在 2021 年以 600 亿美元估值支持 SpaceX 的交叉投资人（他们通常以 15% 的内部报酬率为目标），现在在其预期的首次公开募股前正迎来 30 倍的报酬。Clements 表明，当达到 1.7 兆美元的成果是可实现的时候，成长基金的准入估值会变得宽容得多。

来自 Coatue Management 的数据表明，后期投资的报酬可能会超越早期投资：仅有 8% 估值为 10 亿美元的初创公司实现了 10 倍的估值增长，而估值在 100 亿美元或以上的公司中只有 13% 达到 1000 亿美元，相比之下，接近三分之一的 1000 亿美元公司实现了 10 倍的增长。虽然像 Primary Venture Partners 的 Jason Shuman 等早期投资人认为，从数学上讲，获取 100 倍的报酬优于获取最后的 10 倍增长，但这种新模式的最终验证取决于这些后期投资退出时所实现的现金报酬。

Artificial intelligence is challenging traditional venture capital power laws by accelerating startup growth, prompting major firms like Benchmark, Thrive Capital, Andreessen Horowitz, and Sequoia to raise growth-stage funds up to twice the size of previous pools. While early-stage investing traditionally captured the highest returns, late-stage bets now offer rapid multiples. For example, investors who backed Anthropic in early 2025 at a $60 billion valuation now hold a stake valued at 15 times their initial investment. This contrasts with historical growth-stage returns like Facebook, which raised funding at a $10 billion valuation in 2009 before its $100 billion initial public offering in 2012.

This acceleration is driven by unprecedented sales growth, with Anthropic on track to surpass a $50 billion annual revenue run rate this month, pushing its valuation near $1 trillion. Accel partner Miles Clements noted that crossover investors backing SpaceX at a $60 billion valuation in 2021, who typically target a 15% internal rate of return, are now looking at 30-fold returns ahead of its expected initial public offering. Clements suggests that when outcomes reaching $1.7 trillion are attainable, entry valuations for growth funds become significantly more forgiving.

Data from Coatue Management indicates that late-stage returns may eclipse early-stage ones: only 8% of startups valued at $1 billion achieve a 10-fold valuation increase, and 13% of those at $10 billion or more reach $100 billion, whereas nearly one-third of $100 billion companies achieve a 10-fold increase. While early-stage investors like Jason Shuman of Primary Venture Partners argue that capturing 100-fold returns mathematically outperforms capturing the final 10-fold growth, the ultimate validation of this new model depends on the cash returns realized when these late-stage investments exit.