Sangeen Zeb grew up in Omaha, Nebraska, psychologically wrapped in Warren Buffett’s legend.
“I knew what investing in stocks was at a comically young age, hilariously young,” said Zeb. “I’m eight, my family has no money, and the local news is on: ‘Warren Buffett’s the richest man in the world, and he lives here.’ How did he make his money? Was it chemicals? Is he an inventor? Is it real estate? No, he bought stocks.”
“The most humbling thing is that, no matter how good I am at this job, I will never be the most famous investor from my small hometown,” Zeb jokes. “That brings me down a notch.”
His investments include some of the moment’s most sought-after (and hyped) startups, including OpenEvidence (last valued at $3.5 billion), Harvey (last valued at $5 billion), and Thinking Machines (valued at $12 billion after its seed fundraise). Zeb takes a four-pronged approach when looking at AI startups. You can frame them as questions. The first two are simple: Can the company be big and can that founder actually execute? The next two are more existential. He always asks himself: “If this person fails, would I turn around and hand them a check to go do something else the next day?” And, then, in recognition of just how fast-moving AI is at this point: Can these founders keep pace with the rate of technological change?
“I remember someone once asking me [regarding] Harvey: Is it a land grab?” Zeb told Fortune. “And I said: there’s no land. They’re running on lava. They know the models are changing underneath them, the way we’re interacting with software is changing. Soon, software will be talking to other software… [At Harvey], they were always hyper-aware of all this. They were saying: ‘We have to move faster. We have to push the pace.’”
I asked Zeb, who also serves on the board of the vibe coding platform Vercel, how he feels about the stratospheric valuations (some for extremely early-stage companies like Thinking Machines).
“I do actually understand why valuations are so high—there is scarcity,” he said. “If you look at a few companies, they’re growing like nothing we’ve ever seen. But there are only six of them. So that’s part of the reason we push on these valuations. Even the model companies in the news they’re also growing like nothing we’ve ever seen. We benchmark them against the Mag 7 companies, and we’ve never seen anything like it.”
Term Sheet asked Zeb the seminal question of this series: What’s next?
“The way we think about models will evolve,” said Zeb. “The way that the second generation of model companies are built, I think they’re going to require things beyond scaling and compute. And I do think we’re hitting the limitation on getting the outcomes we want in terms of intelligence… The next set of models is going to advance us beyond large language models.”
See you tomorrow,
