The New VC Game | CoreWeave Sell-Off Creates Panic | Revenue Multiples Out of Control

The conversation argues that venture capital has become far more concentrated, with the best firms now chasing only the few categories that can plausibly produce trillion-dollar outcomes. The speakers contrast that with a growing base of strong software businesses that still have revenue, profitability, and momentum but struggle to raise money because they do not fit the current AI narrative. They also argue that mega-funds now need very large ownership stakes, which pushes them toward only the biggest outcomes and makes early-stage investing harder to justify on pure capital-efficiency grounds.

Much of the episode focuses on whether the AI trade is cracking. CoreWeave’s sell-off is treated as a warning sign, but the broader view is that the real issue is not demand collapsing, it is how quickly demand must keep growing to justify enormous debt and capex. Anthropic’s slowing growth is discussed as a normal sign of scale rather than a fatal one, even as its implied IPO valuation and revenue predictability raise questions about how public markets will price AI leaders.

The episode also explores where value accrues in the AI stack: code review, build-to-prod tooling, inference infrastructure, and routing decisions inside products like Replit. The speakers argue that the biggest near-term opportunities may be in the less glamorous layers around model usage, while also warning that private-market valuations in companies like Modal, Town, and others may be running far ahead of public comps. Bending Spoons and 37signals are held up as examples of alternative software ownership models that may become more attractive as venture gets pricier and more selective.