Google just sucker-punched these highflying tech stocks — don’t let the relief rally fool you

Google’s TurboQuant, published in April 2025, compresses AI model KV cache by roughly 6x (claimed 83% memory reduction) and runs up to 8x faster with no accuracy loss in benchmarks. The algorithm is free and already ported into popular libraries, prompting a rapid market re‑pricing of memory suppliers: Micron (MU) and peers have seen sharp pullbacks from 2026 peaks. The piece argues TurboQuant could materially cut hyperscaler AI capex, pressuring DRAM/NAND pricing and semiconductor-heavy markets (including parts of Korea’s Kospi), while benefiting companies that deploy AI rather than those selling memory hardware. The author recommends avoiding memory-chip stocks and semiconductor ETFs with memory exposure and favoring hard-to-compress assets (oil, defense, critical minerals). Overall market implication: a bearish outlook for memory suppliers and related ETFs, with potential macro/FX effects for Korea and a deflationary impulse on AI infrastructure spending.