Google TPU

GOOGL stock fell in premarket trading after a strong breakout week, even as reports pointed to a bigger push inside Alphabet’s AI chip strategy. The move comes as Google is said to be developing an upgraded TPU v9 chip with MediaTek as a key partner. According to supply chain analyst Ming-Chi Kuo, the new chip is codenamed Triggerfish and is intended to build on Google’s current TPU v9 platform, known internally as Humufish. Kuo said the upgrade would focus on stronger inference performance for next-generation AI workloads, including AI agents and reinforcement learning. Why MediaTek Matters The reported collaboration further cements MediaTek as one of Google’s preferred partners for TPU v9 development. It also highlights how the Taiwanese chipmaker is widening its business ties with major U.S. technology companies beyond its traditional smartphone base. MediaTek has also expanded its work with Nvidia and Microsoft, adding to its profile in the custom AI silicon market. Broadcom and Marvell still dominate that space, but MediaTek’s growing role with several large tech firms suggests it is becoming a more credible challenger. What Triggerfish Could Change Kuo said Triggerfish is expected to bring a substantial memory upgrade, with SRAM capacity estimated to be two to three times larger than Humufish. The chip is also expected to support HBM4E memory, compared with HBM4 in the current design. That larger on-chip memory could keep more active workloads closer to the processor and reduce data movement during inference. In practical terms, that may help Google address the “CPU wall” and “memory wall” that often slow AI computing systems. Chip Expected Memory Setup Timing Humufish HBM4 Current TPU v9 platform Triggerfish 2x to 3x larger SRAM, HBM4E Production slated to begin in late 2027 Kuo estimates lifetime shipments of Humufish at roughly 4 million to 5 million units. Triggerfish could add another 1 million to 2 million units, with volume ramping in 2028. Google’s Broader TPU Strategy Google has increasingly relied on custom TPUs as a core part of its AI strategy, using them internally for products such as Gemini and offering them through Google Cloud. That approach puts Alphabet among the few hyperscalers pursuing a large-scale in-house AI chip roadmap rather than depending only on Nvidia accelerators. GOOGL stock was up 17% year to date, making it the best performer in the Magnificent Seven group. By comparison, MSFT, TSLA and META were in the red for the year, while the iShares Semiconductor ETF (SOXX) had gained 111% year to date. On Stocktwits, retail sentiment for GOOGL slipped to bearish from neutral the previous day as traders debated the stock’s performance against the company’s underlying business strength. Even with the premarket dip, the latest chip report adds another layer to Google’s long-term AI investment story. Read more at: finance.yahoo.com

GOOGL Slips Premarket As Google Pushes Ahead With Next-Gen AI Chip, Analyst Says