Hedge Fund Star Ken Griffin Thought AI Was "Garbage." Now He's Changing His Tune.

This reads as a dispersion signal, not a clean sector call: the market is still overpaying for anything with an "AI" label while underpricing the difference between real monetization and narrative. The first victims are the lowest-quality proxies — small-cap quantum/AI-adjacent names with no visible unit economics — because their multiple depends almost entirely on concept premium. NVDA is less exposed than the headline tone suggests; if anything, skepticism about "AI washing" should concentrate capital into the few names with measurable throughput and pricing power rather than collapsing the whole stack.

The next 1-3 months matter most around earnings and capex commentary. If enterprises cannot show AI-driven margin expansion or incremental revenue in the next print cycle, the market is likely to rotate out of broad AI beta and into a narrower infrastructure trade. The contrarian miss is that the long-run winner set may be broader than semis: workflow automation, data tools, and vertical software can monetize productivity even if the model layer itself is commoditized.

Over 6-18 months, the key is whether AI translates into visible labor substitution and operating leverage; without that, speculative names should see multiple compression even if usage stays high. A meaningful reversal would be a second leg of hyperscaler capex growth or hard evidence from enterprise surveys that AI is replacing headcount rather than just improving workflow hygiene. Until then, this is a "show me" market, and the weakest balance sheets with the loudest AI branding are most vulnerable to de-rating.