Like this podcast? Create your own with Apisod
Google TurboQuant Shrinks AI Memory 6x
Show Notes
Can a 6x memory breakthrough actually cut AI costs, or just reshuffle the chip winners? Google’s TurboQuant says it shrinks the KV‑cache by up to six times, enough to jolt SK Hynix, Samsung, and Kioxia. We trace where this goes next, from pilot deployments inside Google Cloud to the telltale spread between HBM and NAND prices — and who actually feels it if lab gains turn real.
Efficiency is only step one. Inside Google, “Agent Smith” is so popular access had to be throttled, while Anthropic shows Claude agents running computers. Cool until permission creep, provenance, and leakage hit the red line. Cue a security flex: SentinelOne deepens with Google Cloud, lighting up Frankfurt and Saudi Arabia for sovereignty, but the real test is winning banks and hospitals. And just as scale arrives, an LA jury tags Meta and YouTube with $6M for addictive design, narrowing Section 230 and putting engagement mechanics on notice.
Featuring insights from Cloudflare CEO Matthew Prince and Sergey Brin, plus reporting from Bloomberg and Business Insider.
Powered by Apisod.com