Brian Armstrong and Zerodha co-founder Nikhil Kamath warned that soaring private AI valuations could face pressure as cheaper open-source and localized alternatives narrow the gap with frontier models.
Coinbase CEO Brian Armstrong said open-source AI models may be only about six months behind frontier models while reducing inference costs by more than 99%, and Zerodha co-founder Nikhil Kamath said that dynamic could undermine the lofty valuations of private AI companies. Speaking on Kamath’s podcast, the two compared the current AI boom to the dot-com bubble and past crypto market cycles, arguing that expensive proprietary models could lose pricing power as lower-cost alternatives improve. Kamath said he increasingly feels that shorting private AI companies today could be profitable over five years, while Armstrong said fast-rising valuations make him nervous because similar patterns in crypto were followed by corrections before more durable value emerged. Kamath also argued the market could fragment into regional AI ecosystems, with countries building functional domestic versions rather than relying on a handful of American leaders. Armstrong said frontier models may still matter for highly specialized tasks, but cheaper models running on commodity hardware could win a larger share of everyday consumer and business workloads.