Dave McClure Thinks AI Will Replace Late-Stage Investors First
Venture investor Dave McClure says AI will master the math first, while judging founders stays human for longer.

This Week in Startups · NVIDIA Ousts Jensen, AI Comes for Nerds & Walmart Is A Bubble?!During a debate about what artificial intelligence will do to work, venture investor Dave McClure turned the worry on his own profession.
He was talking with fellow investors Jason Calacanis, Tim Draper, and D.A. Wallach about programmers, experts, and professors who fear being replaced. McClure said investors should be worried too, but not all at once.
He put it this way: “I think AI will disintermediate later stage investors before early stage investors.”
By later stage, he meant investing after a startup already has numbers to study, such as revenue, growth, and cash flow. That work is largely math and finance, and McClure argued AI will be able to “figure out math and finance better than any human.”
Early-stage investing is different, he said, because there is often little to measure. It depends more on a judgment call about the founder, reading intangibles and thinking outside the box, where he believes humans keep an edge a little bit longer.
As evidence of how fast the measurable side is moving, McClure pointed to recent advances in AI-driven mathematics. He said many mathematics academics felt they had lost their purpose, part of what he called a crisis of confidence among technical experts.
Even that early-stage advantage may be temporary. McClure said he expects AI to eventually close that gap as well.
Questions people ask
What is the difference between early-stage and late-stage investing?
Early-stage investors usually write the first checks, when a startup may have little revenue and no proven market. Late-stage investors come in later, when the company has customers, revenue, and financial history, and the bet is more about price and growth than survival.
Why would AI be better at late-stage investing?
Late-stage decisions lean heavily on spreadsheets, comparable companies, and patterns like whether customers expand spending over time. Those are structured data problems, the kind where software can test many scenarios quickly and without fatigue.
Can AI already help judge founders?
Partly, but not fully. Software can now summarize backgrounds, flag inconsistencies, and compare founder traits with past outcomes. It still struggles with live impressions like resilience, honesty, and how someone reacts when a plan falls apart.
