An early-stage VC shares why he's not investing in AI coding startups — but doubling down on these founders

He’s invested in over 600 companies and oversees $72 million in assets. But you won’t find a vibe-coding company in his portfolio of the earliest-stage startups.

Jussi Salovaara, the cofounder and managing partner at startup accelerator Antler Asia, said he doesn’t see much more room for new entrants in the AI-assisted coding space.

“It’s been fantastic, but starting new vibe-coding companies right now feels like it doesn’t make sense anymore,” Salovaara, who is based in Singapore, told Business Insider. “It’s an exciting category, but the guys are already out there.”

He said he sees his colleagues at Antler and founders he has invested in using AI to more efficiently write their code and build tools. The firm’s Europe fund was an early investor in $6.6 billion Swedish vibe-coding darling Lovable.

Still, Salovaara said he doesn’t see vibe coding as a viable category for new, early-stage investments.

He said he’s constantly weighing how any startup would defend itself against industry advancements.

“A million people are going to do the same. How do you differentiate?Anthropic ships your product next week, what happens to you? Model costs go up 5x, what happens to you?”

The future of vibe coding and the businesses it has started to disrupt is one of the most hotly debated topics in tech this year. While professional developers continue to say that AI-coded software poses risks and ultimately requires human input, startups like Lovable, Cursor, and Emergent have raised billions, struck big deals, and seen their valuations soar.

Salovaara said that existing AI coding companies may see some consolidation.

“It’s going to be like a holistic kind of setup with a few winners, and they’re going to stay, they’re going to start acquiring smaller companies and just going to stay on top,” he said.

But there’s one type of AI company the VC of eight years said he’s excited about.

Salovaara said he wants to invest in companies whose founders have deep domain expertise, a longtime focus for Antler.

“They combine this deep AI capability with industry capability. So the customers would be automotive, advanced manufacturing,” he said.

He said he thinks about whether a baseline model getting better makes the product better, or if it replaces it.

An example is one of his portfolio companies IndustrialMind.ai, where ex-Tesla employees are building an AI agent to optimize factory operations like manufacturing and process engineering. Another startup, founded by former professional filmmakers, is building a video editing tool for other professional filmmakers.

“Too often you see cases where people who haven’t spent time building for that space,” he said. “That’s always a bit of a head scratcher, like, does that make sense?”

Not all accelerator programs are pumping the brakes on investing in AI coding startups. The latest cohorts of Y Combinator and Andreessen Horowitz’s Speedrun, which invest in early-stage startups, include several companies building AI coding agents or AI app builders.