Micron is the worst-performing stock in the US 100 this week (-16%). Why?

Google Research published the TurboQuant algorithm, a method for compressing key-value cache in large language models.

As a result, the potential reduction in memory requirements could be as much as 6x.

The market's logic was straightforward: if AI models consume 6 times less memory, then data centers need 6 times fewer HBM chips - the core product of Micron and SK Hynix. $MU