VC skips AI coding startups, backs top founders
Jussi Salovaara, co-founder and managing partner of Antler Asia, has announced a strategic shift in his investment approach, stating that he will no longer fund new startups focused on AI-assisted coding or vibe-coding platforms. Based in Singapore, Salovaara oversees $72 million in assets and has backed over 600 companies. Despite Antler's European fund being an early investor in Lovable, a Swedish vibe-coding unicorn valued at $6.6 billion, Salovaara argues that the window for new entrants in this specific sector has closed. He believes the market is already saturated with established players, making it difficult for new startups to justify their existence or differentiate themselves. Salovaara highlights several critical vulnerabilities in the current AI coding landscape. He questions how a new startup could defend itself against rapid advancements from major tech firms like Anthropic, which could ship a competing product almost immediately. Furthermore, he points to the economic fragility of the model, noting that a fivefold increase in model costs could devastate thin-margin coding startups. He predicts that the future of the sector will likely involve consolidation, where a few dominant winners acquire smaller competitors rather than a wide array of independent companies thriving. This sentiment reflects a broader debate in the tech industry regarding the long-term viability of AI coding tools versus the need for human oversight, even as companies like Cursor and Emergent continue to raise billions in funding. Instead of focusing on generic coding assistants, Salovaara is doubling down on companies that combine deep AI capabilities with specific industry expertise. He advocates for founders who possess extensive domain knowledge in fields such as advanced manufacturing, automotive, or professional video production. His strategy prioritizes businesses where improvements in baseline AI models actually enhance the product rather than making it obsolete. For instance, Salovaara points to IndustrialMind.ai, a portfolio company founded by ex-Tesla employees that develops AI agents to optimize factory operations. Similarly, he supports a video editing tool startup built by former professional filmmakers, noting that successful AI applications in niche areas often require the deep context that only industry veterans can provide. He views it as illogical to see founders without specific sector experience attempting to solve complex industry problems with generic AI tools. While Salovaara's Antler program is stepping back from the general AI coding space, other prominent accelerators remain bullish on the sector. Recent cohorts from Y Combinator and Andreessen Horowitz's Speedrun program continue to include multiple startups building AI coding agents and app builders. This divergence underscores a split in investor sentiment, with some seeing saturation in horizontal tools while others continue to identify opportunities in the vertical expansion of AI capabilities. Ultimately, Salovaara's decision reflects a mature market assessment where the competitive moat for new generic AI tools is perceived as too narrow, shifting the focus toward specialized, domain-driven innovation.
