There are some people who feel that building startups at this point might be futile, because AI would essentially do everything in the coming years, but Y Combinator President Garry Tan believes this isn’t the case.
Speaking on a podcast, Tan offered a counternarrative to the doom-and-gloom predictions about artificial intelligence rendering entrepreneurship obsolete. The Y Combinator chief, who has overseen some of Silicon Valley’s most successful startup accelerations, shared observations that paint a more optimistic picture for aspiring founders. His comments come at a time when debates about AI’s trajectory and its implications for human enterprise have reached fever pitch.
“Out there, you open up the Wall Street Journal and you read AI, AI, AI, and it’s very easy to say, oh, none of this is real. This is just Silicon Valley doing its thing again,” Tan observed. “And then when you come here and hear the stories of people working 80-hour weeks trying to build these new businesses, but they’re working and they’re awesome.”
The YC president’s remarks highlight a stark disconnect between media narratives and on-the-ground realities in the startup ecosystem. While headlines often focus on existential threats posed by artificial intelligence, Tan emphasized witnessing entrepreneurs who remain undeterred, pouring their energy into building innovative companies.
But it was Tan’s analysis of AI development trends that provided the most encouraging news for startup founders. “Even the dynamics of what’s happening in AI, I think that’s the good news. The best news I heard from this conference is actually that startups have a chance,” he said. “There’s this other view of AGI and super intelligence, artificial super intelligence coming, that there’s gonna be fast takeoff and suddenly nobody’s gonna have any jobs or nobody can have a company or startup anymore.”
However, Tan believes the reality may be quite different from these apocalyptic scenarios. “I think that we might be entering this moment where the scaling laws are not — they’re sort of slowing down a little bit and then that gives everyone a little bit of a break,” he concluded.
This assessment aligns with recent industry observations about AI development hitting certain technical and economic constraints. AI scaling laws, the methods and expectations that labs have used to increase the capabilities of their models for the last five years, are now showing signs of diminishing returns, according to several AI investors and industry leaders. While recent research suggests that scaling laws may have reached a point of diminishing returns, this trend could paradoxically create opportunities for nimble startups rather than spell their doom.
But that doesn’t mean that AI won’t eat some startups’ lunch. There are plenty of AI wrapper companies whose core products have been included into the AI models themselves, such as companies building try-on clothes apps, or those building grammar tools. But if the pace of AI is slowing down, that gives startups room to create value in areas that AI models themselves can’t do well. Rather than being displaced by AI, many startups can successfully leverage these technologies to accelerate their own development and market penetration. The slowing of AI scaling laws, rather than representing a ceiling for innovation, may actually democratize access to powerful AI capabilities, allowing smaller companies to compete more effectively with tech giants who have traditionally dominated through sheer computational resources.