Google is developing a next-generation AI chip called "Triggerfish" in partnership with MediaTek, according to analyst Ming-Chi Kuo.
Kuo says the chip's headline feature is a substantial memory upgrade. Triggerfish is designed to support next-generation HBM4E memory, the high-bandwidth memory increasingly central to AI workloads. The added capacity is meant to let more active workloads stay on the chip itself rather than being shuffled elsewhere — a design choice that can speed up AI processing.
A separate report carried via Google News, from eciks.org, frames the upgrade as a "3x memory boost," underscoring how large a jump the new design represents over current hardware.
The news moved Google's stock. According to the MSN/Bing News item, GOOGL shares dipped in premarket trading following the report, coming after what the outlet described as a "breakout week" for the stock.
It's worth noting these details come from an analyst rather than an official Google announcement, so specifics could change before any product ships.
Why it matters: memory, not just raw processing power, has become a key bottleneck in running large AI models — so a Google chip built around a major memory upgrade, developed with MediaTek, signals how the biggest tech companies are racing to design their own silicon and reduce reliance on outside suppliers.