
The “DeepSeek moment” is back in the chat
David Sacks and Bill Ackman basically looked at China’s new Kimi K3 model and said: uh, maybe the U.S. lead in AI isn’t as comfy as everyone hoped. The model reportedly topped a coding leaderboard, and that was enough to kick off another round of hand-wringing about whether the chip boom is priced for a world that needs endless compute.
Why your chip watchlist is wobbling
Nvidia and Micron both got dragged around by the same story: if AI models get cheaper, faster, and more efficient, the market has to ask whether all those giant GPU and memory orders stay as gigantic. That’s the kind of question that turns a nice green day into “why is my portfolio suddenly doing interpretive dance?”
- Nvidia slid as traders reassessed AI infrastructure demand.
- Micron took a harder hit this month on separate memory-chip worries, then bounced a bit Friday.
- The broader semiconductor trade got caught in the crossfire of policy anxiety, China competition, and a fresh round of model-mania.
The margin math that makes everyone nervous
The real debate isn’t just about bragging rights on a leaderboard. It’s about whether frontier AI labs can keep charging premium prices if model output gets cheaper and cheaper. Some investors are arguing the opposite: that a better, lower-cost model could actually help everyone except the labs sitting on fat margins.
Big picture: this is less about one model launch and more about the market realizing the AI arms race has a second half. The first half was “buy everything with a chip.” The next one might be “prove you still need all those chips.”
