Gavin Ezekowitz of BFA Global Investors says Anthropic has a strong shot at being one of the major AI companies, even though his firm's early bet came when OpenAI seemed likely to win everything.
Ezekowitz says his understanding of Anthropic's valuation moved from about $170bn to $380bn, and he now weighs team, milestones and enterprise wins above current revenue projections.
David Weisburd says he underwrote Anthropic assuming it would be worth at least $100bn, while his fund runs about 20 positions to capture asymmetric venture returns.
Why venture's best returns defy spreadsheets, indexing and early exits
Venture returns come from a few convex outcomes that cannot be predicted from current revenue or tidy market-size models. Investors who do well hold concentrated positions, read talent and team signals, and avoid indexing or selling winners early. Both guests say discipline and candor about failure matter as much as picking companies.
Takeaways
Early AI winners are hard to predict. Ezekowitz says the winner of a long technology cycle is often not the obvious company. Anthropic looked very different before Claude, when OpenAI seemed set to win everything. Bidders for Anthropic shares in the FTX bankruptcy were bidding only $10m or $15m.
Valuation rests on team and milestones, not revenue. Ezekowitz admits BFA did not focus on revenue at first, and his view of Anthropic moved from $170bn to $380bn. Outcomes are convex, and Uber's market-size analysis would have missed how it expanded its market.
Venture is a macro bet on rare outliers. Exits are rare and capital is committed for years. Founders Fund told its LPs it needs a $100bn outcome per vintage. Weisburd underwrote Anthropic at $100bn or more, and says fear more than greed keeps first movers ahead.
Talent is the most honest signal. Weisburd argues top employees learn the truth about a company quickly, and Ezekowitz agrees. Capital magnets, founders who raise easily, are sometimes not honest, yet following them is not a terrible strategy. Ezekowitz still prizes integrity and transparency.
Indexing venture destroys its convexity. Ezekowitz says BFA's strategy uses 10 venture funds, and spreading across the whole asset class loses the asymmetric returns. Weisburd holds about 20 positions and calls it a myth that any manager can get lucky into the best companies.
Selling early to return capital caps returns. Ezekowitz is concerned that managers selling positions in the secondary market cut off portfolio upside. Weisburd says GPs want 2x to 3x to raise another fund, and poor LPs can push them to sell too early, which he calls LP capture.
Institutional quality and candor sustain firms. Generational transfer rarely works in venture, and Weisburd says most people lack the discipline to build an excellent organization. A manager with $100m needed a Big Four auditor. Ezekowitz gives investors a full account of his worst decisions.
Calls made and missed
Ezekowitz passed on xAI's early $25bn round.
Ezekowitz missed a chance to invest in Cursor at a $10bn valuation.
Weisburd invested in Anthropic at a $4.1bn valuation and says he still holds roughly 90% of his SpaceX position.
Tap a name to go there · tap a line to see why they are linked
On this episode 2
Show 1
Ezekowitz credits Elon Musk with building data centers quickly when nobody else was building at 100,000 GPUs.