The AI sector has gotten hotter over the last year. But unlike many of past venture fads — like crypto or web3 — the AI sector had a number of large startups and legacy players already active when the market started to froth.
There have been AI exits and there are even whiffs of potential government regulation. This dynamic makes it a much more complex ecosystem for founders and investors alike — especially considering many of them weren’t paying attention to AI even a year ago.
Entrepreneurs have flocked to the sector, and early-stage investors are trying to cut through the noise to find which startups are merely riding the hype and which have the potential to grow into substantial companies.
One thing, not unlike other sectors, is that investors are looking for companies with a moat, or competitive advantage over rivals. With deep-pocketed players like Microsoft, Google and OpenAI also actively building in the category, investors want to make sure they aren’t backing companies that could be made irrelevant by the actions of one of the larger entities.
Chris Wake, the founder and managing partner at Atypical Ventures, told TechCrunch+ that while his firm is currently taking a step back from AI to see how things play out, he doesn’t see much appeal of startups that are building on top of existing large language models.
“Building on someone else’s model to solve a business problem, you [have to] understand it’s a race to the bottom,” Wake said. “You can create an interesting business but not necessarily a transformative business. For me, that doesn’t seem incredibly interesting.”
Startups building off other companies’ models also run the risk of larger players wiping them out by simply rolling out new features.
Wake added that the startup has to actually be solving a real problem. While that may seem obvious, sometimes investors get a little too eager to actually get into a company during a hype cycle only to later realize it may not have product market fit when the market cools off.
“[AI] is a hammer, not everything is a nail,” Wake said. “Finding the right kind of AI to apply to something is more important than just coming out with an LLM for ‘X.'”
Brian McCullough, a general partner at Ride Home Fund, said he is looking for companies that completely transform a workflow as opposed to just simplifying it. He gave the example that he wouldn’t want to back a company that makes using spreadsheets easier but rather one that would create a new way to complete the task that didn’t involve them at all.
Some companies need to provide solutions that potential customers would need to have, not those that would just be nice to have, said Sarah Guo, founder of VC firm Conviction. She gave the example of a recent investment made by her firm into Harvey, a startup that helps answer legal questions for lawyers run on OpenAI.
“For the sectors where you can do a lot of the work with AI, make people more efficient and change the basis of competition, suddenly sectors that weren’t interesting are now interesting,” Guo said.
She also said that the investment represents two other traits they are looking for in AI investments: startups building sector-specific tech and doing so in markets without strong incumbents or existing technology solutions.
Both Wake and Andrea Funsten, a partner at RTP Global, also think that the attractive opportunities in AI are in areas where companies are targeting a niche market — like legal — or one that is rich with its own proprietary data.
“What I’m being drawn to a lot more is the companies, products and platforms that are really focused on training more hyper-focused language models usually with the help of synthetic data,” Funsten said.
Funsten said that her team is also seeking out repeat founders regardless of whether the founder has built an AI company before. She said they are doing so to find founders who won’t be as tempted to set themselves up for a rocky road ahead by taking on too big of a round or valuation too early.
“We really focus on repeat founders that are entering the AI space primarily because they have gone through the ups and downs of founding a company generally,” she said. “They have likely founded a company in the last five to 10 years and have seen these up and down roller coaster trajectories.”
But all of this could change. The AI market is evolving quickly and regulation down the road has the potential to steer or upend trends. Funsten joked that the market is moving so fast that she might have a completely different answer to what’s attractive by the end of this week.