How AI Is Rewriting the Power Law of Venture Capital

最近一期 A16Z 播客节目强调了风险投资和技术投资领域正在发生的深刻转变，这主要由人工智能的出现所驱动。David George、Ram Viridian 和 Jen 参与的讨论强调，风险投资中的“幂律”现象比以往任何时候都更加极端，在过去二十年中，美国 3,000 家风险投资公司中只有 20 家能够持续实现 3 倍的净回报。 一个核心的启示是人工智能前所未有的市场影响力。Ram 指出，人工智能在短短四年内实现了 1000 亿美元的收入，而 SaaS 达到这一成就花费了 15 年。人工智能目前正在“攻击 GDP 的每个层面”，影响着交通、劳动力、服务、资本和协调——这些领域总计高达 30 万亿美元。这种广泛的影响表明，人工智能不仅仅是“软件的下一次演进”，而是一种根本不同的技术范式。 David George 解释说，在他的职业生涯中，这是第一次将资本投入一家人工智能公司能够直接通过资助计算资源来增强其优势，而不仅仅是增加人手（这在历史上会造成协调问题）。这种“对推理的无限需求”在人工智能市场中培养了真正的规模经济，从而使领先公司实现大规模、快速增长。发言者预测，在人工智能创造巨大价值的潜力驱动下，将出现市值达 100 万亿美元的公司。 风险投资行业本身正在转型。Ram 和 David 将其描述为不再是“家庭手工业”，而是一个 5-6 万亿美元的资产类别，公司保持私有化的时间更长，即使在后期也能产生“风险投资式的回报”。顶级的退出，曾经大约在 100 亿美元左右，现在经常超过 400-1000 亿美元。这导致了风险投资中的“中间地带的消亡”：只有高度专业化、利基的早期基金或像 A16Z 这样能够从种子轮投资到 IPO 的大型全生命周期公司才能蓬勃发展。小组成员强调，强大的早期阶段特许经营权对于后期阶段的成功至关重要，因为它能提供资源、信息和关系。 从有限合伙人 (LP) 的角度来看，情况充满挑战。持续高表现的风险投资公司稀缺，这意味着有限合伙人必须集中投资，而不是过度分散。考虑到流动性锁定，平均风险投资回报（1-2 倍净回报）与私募股权或公开市场相比通常不具吸引力。有限合伙人还面临激励措施错位的问题：如果他们遵守基准，很少会因为错过下一波大潮而受到惩罚，但如果做出糟糕的投资则面临被终止的风险。人工智能快速、高估值的融资轮次进一步增加了他们的尽职调查难度。然而，顶级的普通合伙人 (GP) 可以迅速实现资金回笼的流动性，有时甚至比私募股权更快，从而减轻了对“流动性时间线”的担忧。 展望未来，机遇是巨大的。David 强调了人工智能尚未涉足的领域：消费级人工智能（超越聊天机器人）、机器人技术、自主性、医疗保健（占 GDP 的 18%），以及对物理世界的重塑（制造业、国防、数据中心）。Ram 指出人工智能的供应侧瓶颈——能源、电网、数据中心和芯片——是巨大的未开发投资机会，暗示解决这些问题可能会开启数千亿美元的市场。小组成员总结说，目前关于人工智能的“零和思维”，即技术堆栈中的某一层必须“胜过”另一层，是一个误区。市场如此巨大，以至于“一切都将奏效”，这将导致类别的大幅扩展和前所未有的价值创造时代。

A recent A16Z podcast episode highlighted a profound shift in venture capital and technology investing, driven primarily by the advent of artificial intelligence. The discussion, featuring David George, Ram Viridian, and Jen, underscored that the "power law" in venture capital has become more extreme than ever, with only 20 out of 3,000 U.S. VC firms consistently achieving 3x net returns over the last two decades. A core revelation is AI's unprecedented market impact. Ram noted that AI reached $100 billion in revenue in just four years, a feat that took SaaS 15 years. AI is currently "attacking every facet of the GDP," impacting transportation, labor, services, capital, and coordination – areas totaling an astounding $30 trillion. This widespread influence suggests that AI is not merely the "next evolution of software," but a fundamentally different technological paradigm. David George explained that, for the first time in his career, pouring capital into an AI company directly compounds its advantage by funding compute resources, rather than just increasing headcount, which historically created coordination issues. This "unlimited demand for inference" fosters real economies of scale in the AI market, leading to massive, rapid growth for leading companies. The speakers predict the emergence of $100 trillion market cap companies, driven by AI's potential to create enormous value. The venture capital industry itself is transforming. Ram and David described it as no longer a "cottage industry," but a $5-6 trillion asset class where companies stay private longer, resulting in "venture-like outcomes" even at the late stage. Top-tier exits, once around $10 billion, now regularly exceed $40-100 billion. This has led to the "death of the middle" in venture capital: only highly specialized, niche early-stage funds or large, full-lifecycle firms like A16Z, which can invest from seed to IPO, are thriving. The panelists emphasized that a strong early-stage franchise is crucial for success in later stages, providing access, information, and relationships. From an LP (Limited Partner) perspective, the landscape is challenging. The scarcity of consistently high-performing VC firms means LPs must concentrate their investments rather than over-diversifying. The average venture return (1-2x net) is often uncompelling compared to private equity or public markets given the liquidity lock-up. The LPs also face an incentive misalignment: they are rarely penalized for missing out on the next big wave if they stick to benchmarks, but risk termination for making a poor investment. AI's rapid, high-valuation rounds further complicate their due diligence. However, top-tier GPs can achieve fund-returning liquidity quickly, sometimes even faster than private equity, mitigating the "timeline to liquidity" concern. Looking ahead, the opportunities are vast. David highlighted areas where AI has barely scratched the surface: consumer AI (beyond chatbots), robotics, autonomy, healthcare (which is 18% of GDP), and the re-imagining of the physical world (manufacturing, defense, data centers). Ram pointed to the supply-side bottlenecks in AI – energy, the grid, data centers, and chips – as massive untapped opportunities for investment, suggesting that addressing these could unlock multi-hundred-billion-dollar markets. The panelists concluded that the current "zero-sum thinking" about AI, where one layer of the tech stack must "win" over another, is a pitfall. The market is so immense that "everything will work," leading to a significant expansion of categories and an era of unprecedented value creation.