How AI Is Rewriting the Power Law of Venture Capital

Episode Summary

Executive Summary: The episode argues that AI is reshaping venture capital and broader asset allocation by making the power law even more extreme: capital can now be directly converted into compute, compounding winners faster than ever. The hosts and guest contend that AI’s market spans far beyond software into labor, healthcare, robotics, energy, chips, and data centers, forcing LPs and GPs to rethink concentration, access, and fund sizing.

Main Topics: AI and the intensification of the power law (Priority: 5/5): The discussion opens with the idea that AI is making venture outcomes more concentrated: a few companies capture disproportionate value, and frontier companies can use capital to compound their lead through compute. Portfolio construction and LP asset allocation (Priority: 5/5): The speakers argue AI should be a core allocation, not a satellite one, and that LPs need higher concentration, better access, and larger position sizing to benefit from venture’s dispersion. Why venture funds and firm scale matter (Priority: 5/5): They discuss why only a tiny share of VC firms consistently generate strong net returns, why the middle is disappearing, and why founder demand increasingly favors large platforms with resources across the company lifecycle. AI expands TAM beyond software (Priority: 4/5): The episode repeatedly emphasizes that AI is attacking labor and task value, not just software spend, which makes market size estimates much larger than traditional SaaS or IT framing. Signal quality, traction, and valuation in AI startups (Priority: 4/5): They note that AI has made company evaluation harder because traction can be noisy, rounds move faster, and metrics like ARR may be misleading without deep customer understanding. Impacts on private equity, public markets, and credit (Priority: 4/5): The conversation expands beyond venture to show how AI is pressuring legacy software assets, changing IPO/MA paths, and creating stress in leveraged software buyouts and private credit. Future categories: robotics, autonomy, healthcare, energy, and infrastructure (Priority: 5/5): The speakers identify where the next huge outcomes may come from, especially robotics, autonomy, healthcare delivery, drug discovery, data centers, chips, and power/energy bottlenecks.

Key Arguments: AI is different from prior software waves because capital can be transformed directly into compute, which improves the product and reinforces leader advantages. The AI opportunity is not confined to software; it targets labor, services, coordination, transportation, capital, and healthcare tasks across the real economy. Venture returns are increasingly dominated by access to the few category winners, so LPs should concentrate capital rather than spread it thinly across many managers. Large venture firms can win because founders want brand, lifecycle support, and de-risking help across seed through IPO, not just a small check. Traditional SaaS valuation and growth assumptions may break if products are not AI-resilient; weaker assets may face lower multiples and fewer buyers. The best way to assess AI startups is not just cohort/renewal analysis, but close customer-level understanding of actual demand and usage. Many AI companies are still early in diffusion; usage and spend are tiny relative to the size of the knowledge-work and labor markets they could address. The next major value creation may come from categories that barely exist today, especially robotics, autonomy, healthcare, and AI infrastructure.

Data Points: Venture firms with consistent 3x net returns: 20 out of 3,000 - Data cited on U.S. VC firms achieving consistent 3x net returns over two decades AI revenue milestone: $100 billion - AI reached this revenue level in about four years, compared with SaaS taking 15 years Software-as-a-service revenue milestone comparison: 15 years - Time SaaS took to reach $100 billion in revenue Estimated enterprise value of frontier model companies: $3.5 trillion to $5 trillion - Mentioned for SpaceX, OpenAI, and Anthropic as a cluster of potential value Value created in prior tech cycle: $25 trillion - Referenced as the last cycle’s market cap creation, with expectations that AI will be larger Average venture return: 1x to 2x net over 10 years - Cambridge data referenced for average venture performance Early-stage venture loss rate: ~60% - Referenced as normal for top-performing early-stage funds Growth-stage venture loss rate: 10% to 20% - Referenced as more appropriate for later-stage investing U.S. labor spend vs software spend: ~40x larger - Used to argue AI’s addressable market is much broader than software Healthcare IT spend: $60 billion to $100 billion per year - Compared with AI’s potential to target healthcare labor and tasks Healthcare labor/task market: ~$1 trillion - Claims, billing, and administration cited as AI-addressable labor value Median U.S. company AI spend: $12 per employee per month - Used to show diffusion is still early Top 1% AI spend intensity: $7,000 per employee per month - Used to illustrate how concentrated usage/value capture is today Public SaaS companies trading above 10x revenue: 15 to 20 companies max - Used to show valuation compression and AI resilience screening in public markets Private equity software LBO volume (2021-2022): $200 billion to $300 billion - Referenced as the size of software buyout activity during the prior cycle Debt taken out in those software deals: Over $200 billion - Used to discuss leverage and credit risk in software portfolios Average EBITDA multiple in those deals: 25x to 32x EBITDA - Referenced as prior-cycle entry valuations for software buyouts Current value of those software deals: About half of entry value - Used to explain valuation compression and credit stress Waymo live vehicles in the U.S.: Fewer than 10,000 - Used to show autonomy remains early despite enthusiasm

Pivotal Quotes: "for the first time in my career, you can take capital and throw it at a company and it compounds their advantage." — David George: Explaining why AI makes the power law more extreme than prior technology waves "AI is attacking every facet of the GDP: transportation, labor, services, capital, coordination." — David George: Arguing AI’s addressable market extends far beyond software "the bottleneck in AI today is not demand. It's on the supply side." — Adam D’Angelo: Discussing energy, grid, data center, and chip constraints that may shape the next wave of winners

Implications: Investors should expect higher concentration, bigger winners, and more lifecycle-driven venture franchises. AI is likely to reshape not just startups but public markets, PE, and credit, while creating major new opportunities in infrastructure, robotics, healthcare, and autonomy.