Marc Andreessen on Evaluating Founders and AI's Consumer Surplus

What It Covers

Marc Andreessen speaks with Harry Stebbings on the 20VC podcast about evaluating founders across three dimensions, why venture capital's "scalded stove" bias destroys returns, how AI is reconcentrating the tech industry within a 20-mile Silicon Valley radius, and why roughly 99% of AI's economic value will flow to users rather than the companies building the technology.

Key Questions Answered

• •Scalded Stove Phenomenon: Venture investors who lose money in a category carry that emotional scar into future meetings, causing them to pass on the next great company in that same space. Andreessen argues mistakes of omission — missing the next Google — cost 10x more than mistakes of commission. Actively naming and releasing past losses is how a16z keeps partners evaluating each deal fresh.
• •Three-Part Founder Framework: Andreessen screens founders for IQ first (measured by how many notes he takes during a meeting), then courage (the willingness to confront problems directly rather than avoid them), and finally primal drive — a compulsion to build something of one's own that persists when conditions become genuinely miserable. Credentials and polish are explicitly not part of the filter.
• •Overfunding Kills Companies: Andreessen cites Don Valentine's principle that more companies die from indigestion than starvation. Every funding round sets a valuation hurdle the next round must clear, and no investor voluntarily leads a down round in someone else's company. Founders should treat excess capital as a structural risk, not a safety net, because financial discipline collapses when headcount scales too fast.
• •AI Geographic Concentration: Despite COVID-era optimism about remote work decentralizing tech, Andreessen observes that AI has reversed that trend sharply. Close to 100% of high-value AI companies now sit within a 20-mile radius of Silicon Valley. Founders building AI outside that corridor face a meaningful talent and network disadvantage, with Eleven Labs, Black Forest Labs, and Mistral cited as rare exceptions.
• •Consumer Surplus Dominates AI Economics: Drawing on Schumpeterian economics, Andreessen argues that roughly 99% of AI's total economic value will accrue to users, not to the companies building models — mirroring the Internet and smartphone patterns. Apple and Google captured approximately 1% of smartphone value; users captured the rest. Investors should size AI application opportunities against this user-surplus dynamic, not just model-layer revenue.
• •Large Company Overstaffing Is Structural, Not AI-Driven: Current tech layoffs are misattributed to AI. Andreessen estimates most large companies are overstaffed by 50%, with many at 75%, due to zero-interest-rate hiring binges during COVID. AI was not capable enough to replace those roles until late 2024, making it a convenient cover story. The actual driver is interest rates rising from 0% to 5% forcing financial replanning across every major organization.