Dan Niles warned on May 12 that AI-linked stocks could fall by 30% to 50% in early 2027 following a period of rapid gains and market concentration. "What do I think happens to these stocks sometime early next year from wherever they get to? They probably go down 30% to 50%," Niles said during an interview [1].
Niles compared the current AI boom not to the 2000 dot-com bubble peak but to the internet infrastructure buildout era of 1997-1998. He highlighted "agentic AI," which performs multi-step tasks, as the main factor driving increased computing demand and a prolonged rally before a market reset [1, 2].
He also cautioned that semiconductor stocks are currently overbought, and a near-term pullback would not be surprising [1, 2]. Niles forecast a shift in hardware demand caused by agentic AI, increasing the importance of CPUs relative to GPUs. This change, he said, would benefit Intel and AMD while putting pressure on Nvidia [2].
Among major tech companies, Niles sees Google as best positioned for the AI transition due to its ownership of the full AI asset stack [1].
A JPMorgan survey cited by Niles found 54% of institutional investors expect a more than 30% correction in U.S. stocks during 2026-2027, with 45% anticipating the downturn in 2027. Despite this, 75% of investors see more than 20% upside remaining before a tech bubble peak [2].
Niles advised investors to increase cash holdings and watch oil prices, U.S. Treasury yields, plus AI industry financing and IPO activity for signs of liquidity risks [2]. Meanwhile, Taiwan stock ETFs like Taiwan TOP50 (009816) have seen strong inflows and a 40%-plus gain over three months amid continued optimism about AI and tech growth [2].
Niles’s forecast of a sharp AI stock correction in early 2027 provides a concrete timeline for investors tracking the fast-moving sector.
