How AI Is Rewriting the Power Law of Venture Capital
This episode delves into how AI is fundamentally altering the power law of technology investing, making capital a compounding advantage for AI companies and expanding the total addressable market far beyond traditional software. The guests discuss rethinking portfolio construction, the challenges of evaluating true AI traction amidst hype, the struggles of legacy SaaS businesses, and future trillion-dollar opportunities in robotics, autonomy, healthcare, and physical infrastructure, emphasizing that the real bottleneck for AI's growth is often on the supply side, particularly compute and energy infrastructure.
This discussion, featuring a16z’s Jen Kha and David George with Accolade Partners’ Aram Verdiyan, explores how Artificial Intelligence is fundamentally reshaping the landscape of technology investing, particularly in venture capital. The core argument is that the traditional power law in venture, where a few investments generate the majority of returns, is becoming even more extreme due to AI.
AI’s Impact on the Power Law and Economic Landscape
David George explains that for the first time in his career, capital can directly compound an AI company’s advantage by purchasing compute, which in turn enhances products and businesses. This differs from traditional startups where excessive capital could lead to coordination issues. AI is now attacking every facet of the GDP, including transportation, labor, services, and capital coordination—sectors representing an estimated $30 trillion in GDP. This unprecedented breadth suggests AI’s Total Addressable Market (TAM) is vastly larger than traditional software or even healthcare IT, potentially 10x or more, by targeting the economic value of tasks performed by labor. Aram Verdiyan highlights the rapid growth of AI, reaching $100 billion in revenue in just four years, compared to SaaS which took 15 years to achieve the same milestone. This rapid penetration and demand make a strong case for AI as a “core” or “super core” allocation in investment portfolios, rather than a mere “satellite position.”
Rethinking Portfolio Construction and Investment Strategy
The conversation stresses the need for investors to rethink portfolio construction from a blank slate. The speakers note that venture-like outcomes are increasingly observed in late-stage companies, with top outcomes growing from $10 billion to $40 billion, and potentially $100 billion. This expansion suggests that subsequent technology waves are larger than prior ones. While the market may not be “winner-take-all” across all categories, there’s a strong belief in the power law within a given category, where the dominant player captures most market share. This implies that investors must prioritize identifying and backing the leading company in credible categories. Despite the high-risk nature of venture (with top-performing funds having a 60% loss rate on early-stage investments), the few winners can yield 10x or more returns, making consistent access to category-defining companies crucial.
Challenges in Evaluating AI Companies and LP Sentiment
Evaluating AI companies presents unique challenges. The rapid pace of growth (e.g., zero to $5 million ARR in a month) can be misleading without established renewal cycles or true product-market fit. Distinguishing real traction from hype requires deep customer understanding and qualitative assessment beyond mere financial metrics, as demonstrated by the example of Harvey, where usage initially was low but soared post-reasoning models. LP sentiment is mixed; while some worry about market overheating or “catching a falling knife,” others, like CalPERS, are significantly increasing their allocation to venture and growth. A significant challenge for LPs is the “incentive misalignment”: GPs risk being fired for missing the next big company, while LPs often face less pressure for inaction, creating a resistance to embracing change. Only 20 out of 3,000 venture capital firms in the US have achieved consistent 3x net returns over the last two decades, underscoring the difficulty of consistency and the need for LPs to concentrate investments in top-tier funds to avoid average venture returns, which are often inferior to private equity or public markets.
The Legacy SaaS Problem and Future Opportunities
The “Legacy SaaS Problem” refers to pre-ChatGPT software companies, often acquired by private equity, that are not AI-native, show slow growth (e.g., 10-20%), and struggle with dramatically contracted valuations (from 15-20x EBITDA to 2x revenue). These companies face challenges in finding buyers and showing resilience from AI without significant re-invention, which often requires a “founder mentality” and tough decisions, as exemplified by Intercom. The idea of “AI Private Equity” is not a panacea; simply infusing AI without fundamental workflow changes can lead to customer churn. The conversation identifies vast future opportunities beyond current AI applications, including robotics, which is predicted to be “bigger than the language stuff” within the next decade; autonomy (e.g., robo-taxis); healthcare (which is 18% of GDP, with nascent progress in care delivery and drug discovery); and the reimagining of physical infrastructure, defense, manufacturing, and data centers. The speakers agree that the current bottleneck for AI is not demand, but supply-side factors like energy, grid infrastructure, data centers, and chips, presenting a massive opportunity for investments in “the machine age.”