Interview, Fireside Chat
How AI Is Rewriting the Power Law of Venture Capital
- The power law in technology investing is projected to intensify compared to the previous 10 to 20 years, driven by capital compounding advantages in AI and the emergence of network effect companies, with economies of scale remaining a critical market driver.
- Late-stage venture-like outcomes are anticipated, with top decile companies like Anthropic and OpenAI potentially reaching $100 billion valuations, contributing to a larger total addressable market estimated at 10x that of traditional SaaS or healthcare IT.
- The $25 trillion of market cap generated in the last cycle is expected to expand into larger subsequent waves, with future $10 trillion to $100 trillion companies likely emerging in new domains such as robotics, defense, manufacturing, and healthcare over the next two tech cycles.
- Labor markets are expected to undergo reinvention rather than elimination, with legal counsel utilization increasing as lawyers focus on high-level strategy while AI handles initial analysis, though full diffusion of AI into non-coding knowledge work will take longer than current hype suggests.
- Loss rates for early-stage venture funds are expected to remain near 60%, whereas growth stage loss rates are projected to fall between 10% and 20%, with fund returning math becoming viable in late stage if the best portfolio company constitutes 5% to 10% of the fund.
- Venture capital consistency is described as extremely difficult, with only 20 out of 3,000 firms achieving consistent 3x net returns over two decades, and the average LP return over the last 10 years estimated at only 1 to 2x net, making average fund allocation less compelling than public or private equity.
- A trend of "death of the middle" in venture capital is expected, where highly specialized small firms and large multi-product funds succeed while the "messy middle" struggles, alongside a competitive landscape where top performers are significantly ahead if they accessed top five to ten companies in the last five to 10 years.
- Deal dynamics are expected to feature larger, faster rounds with traction metrics that may not reflect real product demand, creating confusion where founders choose GPs based on brand and risk mitigation, leading to a flywheel effect for successful partnerships.
- Infrastructure bottlenecks in energy, the grid, and data centers are identified as primary constraints for AI growth, with the US specifically facing challenges in power speed rather than capacity compared to other nations adding 10x more renewable capacity annually.
- Supply-side constraints are expected to generate over $100 billion in opportunities for infrastructure and energy funds, while private equity firms are shifting focus toward "AI native" companies with acceleration, rejecting slower-growing software assets that lack resilience to AI.
- Software companies with legacy models will need to adapt, potentially by reintroducing founders to revamp businesses, as simply adding AI to a website without workflow integration risks customer dissatisfaction and revenue drops.
- The current market is not characterized as a bubble similar to the dot-com or COVID eras, as traction is viewed as real and not ephemeral, though rounds are expected to remain confusing due to metrics failing to reflect actual demand.
- Consumer AI end-use cases are projected to evolve from chatbot interfaces to native systems that execute work on behalf of users, while legal and other professional services see increased billable hours as AI adoption shifts task structures.
- The diffusion of AI adoption is currently very early, with average US companies spending $12 per employee on AI compared to $7,000 for the top 1%, suggesting significant headroom for growth in new product areas.
- Liquidity dynamics favor top 1% venture firms capable of generating faster exits via M&A and IPOs, whereas endowments and family offices generally prefer compounding capital to manage tax and portfolio concerns.
- Private equity exits in 2024 are currently smaller than venture exits, exemplified by acquisitions like Cursor by SpaceX valued at $300 million, with non-AI software assets struggling to find buyers if they do not demonstrate fast growth or acceleration.
- A massive expansion in the number of technology categories is expected, even if specific categories retain winner-take-all dynamics, with open source success not necessarily negating laboratory success, contrasting with previous technology eras.