Analysis
Benchmark's five general partners say they're still making just two to three new investments each per year, even after the firm closed $2 billion across two funds this summer -- only the second time in its 31-year history it has raised meaningfully more than its traditional size, TechCrunch reported after interviewing the partners on stage at TechCrunch Disrupt, continuing a string of 2026 coverage of the firm's shifting strategy.
The raise splits into a core early-stage fund, historically held near $500 million for about 20 years, and a new $1.25 billion growth vehicle -- Benchmark's first dedicated late-stage fund, according to Pulse 2.0. Partner Peter Fenton called keeping the core fund flat for two decades without adjusting for inflation "probably a tactical error."
Five partners, one Monday meeting
The firm's investing process has barely changed: Peter Fenton, Eric Vishria, Chethan Puttagunta, Jack Altman (who joined in February) and Ev Randall (who joined late last year from Founders Fund, Bond and Kleiner Perkins) meet every Monday with no agenda, available "at their disposal" when founders need help, Puttagunta said. Fenton described the firm's judgment of founders as "awe" rather than a winner-or-loser scorecard, and said a single polarizing personality in the room can sink a deal.
Benchmark's case study is Cerebras, where it led the first round in spring 2016; the chipmaker's shares priced at $185 at its IPO earlier this year, climbed to $350, and have since drifted back near $185, with the company having raised $6 billion in total. Partner Ev Randall also described Character.AI -- which filed for bankruptcy this month -- as a company "already sold for $2.3 billion," a reminder that even Benchmark's losses can carry a real exit number attached. Shopify and Ramp were cited as proof that founding outside San Francisco can still work, even as Fenton said more than 90% of the firm's historical capital returns trace back to San Francisco-founded companies.
The partners' read on public-market turbulence at CoreWeave, Nebius, Datadog, DoorDash and Amazon was that the swings reflect exaggerated or inaccurate data points, not unhealthy AI fundamentals, in Randall's words. Vishria said direct listings are unlikely for the current generation of AI companies because they need too much capital to skip a traditional raise -- a view that cuts against more bullish direct-listing chatter elsewhere this year.
The discipline carries a real risk, though: three of the four partners interviewed three years ago have since left the firm, and a five-person partnership making two to three bets each will need outsized conviction -- and outsized check sizes -- to deploy $1.25 billion into just four to six, maybe seven or eight, growth-stage companies without drifting into the kind of late-stage prices Benchmark has historically avoided.
Jack Altman, the newest partner, says he's already made several chip, robotics and infrastructure bets and credits the other four for helping him win competitive deals -- the clearest sign yet that Benchmark's bet is on its five-person bench, not on writing more checks.
