I spent two hours with Turner Novak on The Peel talking about where we are in the AI cycle, and why the hardest part of early-stage investing is knowing what not to chase.
A few other topics we got into:
1. FOMO is for sheep. Confidence at the inception or seed stage is usually just FOMO dressed up as conviction. When someone tells me they’ve identified the winner in a billion-dollar seed round, what they’ve identified is a hot deal.
2. A lot of people are trying to separate vibe revenue from real revenue right now. At the inception or seed stage, there’s neither traction nor revenue, which is why the diligence has to be on the people.
3. We are in inning one on inference. Inference workloads are still under 10% of the total. Today, only two things have real traction – search and answers, and coding. Coding alone is 30 to 50x the revenue of every other agentic category combined. The rest of the stack hasn’t been built yet.
4. The history lesson of bottlenecks. In the mid-90s, we had demand but no infrastructure: dial-up modems, no WiFi, no smartphones, and ~50M PCs. Today, what’s missing is the computing grid and the power to run it. We’re building the highways. Enterprises have to buy the cars. If adoption lags the buildout, the market multiples will correct.
5. We’re selling work, not software. Global spending on white-collar employees is around $30 trillion, compared with a $600 billion enterprise software market. AI will go after people spend rather than software spend, which is why pricing will move from subscription to outcomes. If AI does 10% of the work people do by 2030, it’s a $3 trillion market. At 20%, it’s a $6 trillion market.
Thanks Turner, for a fun conversation!
