Hardware Shock: Google TurboQuant Algorithm Crashes Memory Makers' Stocks

The semiconductor industry experienced a "Black Thursday." On March 28, Google Research officially announced TurboQuant—a breakthrough method of mathematical compression (extreme quantization) that reduces RAM consumption during heavy LLM inference by exactly 6 times.

The market reaction was instantaneous and ruthless. Shares of industry giants (Micron, Kioxia) collapsed within 24 hours. The reason for the panic is obvious: if hyperscalers (AWS, Azure, GCP) now require 6 times less hardware to store AI model weights in RAM, global demand for NAND and HBM modules will radically decline. Google has elegantly solved the data center shortage problem not by building new facilities (which the US Congress is currently opposing), but through brilliant software optimization. Software mathematics has once again defeated "iron" physics.

Source: Google Research