Ex-Greylock GP Sarah Guo Announces Conviction Partners, Her New AI Fund

Venture capital is prone to hype cycles, like the funding glut for Web3 last year and for artificial intelligence in the late 2010s. Now, AI has returned to the zeitgeist but venture capitalist Sarah Guo is confident the hype is here to stay—so much so that she’s staking her new VC fund on that bet.

Guo announced Tuesday the first close of a $101 million debut fund for her new investment firm Conviction Partners. After spending nine years at Greylock, where she rose through the ranks to become the Silicon Valley firm’s second female general partner, Guo left the firm in June. With Conviction, she plans to invest in AI startups up to the Series A stage with check sizes between $1 million and $10 million. Guo is starting out as a solo investor, but plans to build a team around her that’s tailored to AI’s research-heavy nature.

For example, she has recruited a visiting scholar to conduct AI research, and says she plans to hire more academics who might build new models and collaborate on ideas with startups Conviction invests in. “We’re not going to assume the traditional knowledge of what a venture firm thinks belongs in house and out of house,” she told Forbes in an interview. She pointed out that for generalist firms, hiring a team of AI researchers “doesn’t make sense.”

Guo grew up in Wisconsin as the daughter of Chinese immigrants who worked at Bell Labs before becoming tech entrepreneurs. She became immersed in technology as a teenager, building a website for her parents’ company. After graduating from the University of Pennsylvania, she worked two year stints at broadband network startup Casa Systems and on tech investment banking at Goldman Sachs. She joined Greylock in 2013, and credits the generalist firm for giving her access to investments in spaces like health tech and crypto.

But it was AI that stuck with her most of all. She says she seriously considered quitting Greylock to start a company with Google Brain cofounder Andrew Ng in 2016 (also the year she was named to Forbes’ 30 Under 30 list for Venture Capital). Still, like Web3 last year, AI had a VC hype cycle of its own which fizzled out once the technology and revenue proved to be growing too slowly. “A couple of very smart investors asked me over dinner in 2018, in the trough of disillusionment [about AI], what I’d gotten wrong about bots and natural language,” she said. “I abashedly admitted that the underlying technology wasn’t quite there.”

Now, recent rapid-fire technical breakthroughs from companies like Hugging Face and Stability AI indicate that the pace of research is progressing up an exponential curve, Guo says. Instead of blockchain-based Web3, Guo believes the foundational technology of the coming years is AI. She says she left Greylock to invest in this type of technology in line with the “classic startup story of wanting to do something that’s really different—often, that’s much easier to do starting from scratch.”

Guo says her fund took just 10 weeks to raise, and features both institutional investors like university endowments (though she said she cannot name names) and about 50 founders and executives who are bullish on AI. She counts herself lucky to have raised the fund during a downturn because it gives her access to higher quality companies at less inflated prices. “It would be really problematic to be starting a venture fund at the peak of the bull market,” she says. “Capital was insanely abundant and that affected the actual quality of investing and the quality of decision making.”

Conviction Partners will likely make a maximum of 20 investments into startups building with AI from the ground up. “You have to design a company differently to do that: the talent you hire is different and the challenges to solve in the product are different,” Guo said. For example, these companies will need to spend much more on computing costs compared to incumbent software giants that are built on infrastructure like Amazon Web Services.

“There are plenty of smart investors I respect that say, ‘it’s a hammer looking for a nail’ or ‘no machine learning company I’ve ever invested in has made money,’” Guo said. “But they’re wrong this time. Lots of things don’t work until they do.”