Sourcery: USV’s $900M Fund Signals AI’s Next Venture Wave

SPEAKER_1 (0:01)

When a venture firm raises its biggest fund ever, the real story isn't just the number, right? It's what that number says about the market ahead. Exactly. And USV's new $900 million fund is telling us a lot. This episode features Molly O'Shea's conversation with Fred Wilson, Rebecca Kaden, and Michael Mignano, and they're unpacking not just the fund size, but the firm's evolving thesis and how AI is fundamentally changing venture capital. Fred Wilson was careful to point out this is "by far the most USV has ever raised in a new fund cycle," but the goal isn't vanity. They're planning to invest in about 30 to 40 companies, with more seed-stage bets than before. That's interesting because you'd think a bigger fund means bigger checks, but they're actually going earlier. Why the shift?

Because building is easier now, competition is fiercer, and new startups are appearing everywhere. But USV remains intentionally concentrated, bottom-up, and thesis-driven. And that thesis now centers on four areas: physical AI, AI applications, AI infrastructure, and energy. The partners made a compelling comparison to the early internet era. Infrastructure came first, then the application layer exploded later. They believe AI is entering that same phase now, with agents, consumer tools, and enterprise systems still early but moving fast. One phrase kept coming up in their conversation: "obliterate, don't automate." What do you make of that?

It's crucial. Rather than just speeding up old workflows, they see AI as reorganizing entire markets. Consumers may gain access to intelligence once reserved for expensive professionals, whether in finance, healthcare, or other services. But it also means startups face new pressure from inference costs, crowded markets, and the need for sharper go-to-market execution. There are big opportunities here, but also big questions. The group talks about how every company may eventually have an agent, and how users will want a single system that can talk to many service-specific systems on their behalf. And they stressed that network effects still matter, but may emerge in new ways as single-player AI behaviors become truly multiplayer. Security came up as one of the most urgent frontiers too. Right, because as agents take on more responsibility, users need better protection, clearer trust, and more control.

They held up Coinbase as a model, described as "the most trusted brand in the sketchiest industry ever." That same logic may apply to AI, where trusted brands could matter as much as raw capability.

Building on that point, USV closed by returning to founders, pointing to long-term wins like Twitter, Etsy, Indeed, MongoDB, Twilio, Stripe, Coinbase, and Duolingo.

The message being that great companies are rarely straight lines. In the end, this episode is really about patience, conviction, and backing the people who can turn a thesis into a generational company.