Google is developing a next-generation AI chip called Triggerfish with MediaTek, featuring significantly larger memory to handle advanced AI workloads like agents and reinforcement learning—a move that positions MediaTek as a credible challenger in the custom AI silicon market alongside Broadcom and Marvell.

Yahoo Finance AI6h ago3 min read

Summaries like this, in your inbox every morning.

3 Key Points

What happened
Google has selected Taiwan's MediaTek as its exclusive development partner for Triggerfish, an upgraded version of its TPU v9 AI chip. The new chip will feature SRAM capacity that is two to three times larger than the current design (codenamed Humufish) and support for next-generation HBM4E memory, allowing more workloads to remain on-chip and reducing data movement during inference. Production is slated to begin in late 2027 with volume ramping in 2028.
Why it matters
Google is one of the few hyperscalers (large cloud providers) pursuing a large-scale in-house AI-chip roadmap rather than relying solely on Nvidia accelerators. This partnership underscores MediaTek's shift beyond smartphones into a strategic role with major U.S. technology companies—Google, Nvidia, and Microsoft—indicating it is becoming a credible challenger in the custom AI chips market, where Broadcom and Marvell remain dominant. The memory upgrades could help address bottlenecks in AI computing known as the "CPU wall" and "memory wall."
What to watch
Triggerfish is expected to add 1–2 million units in lifetime shipments, compared with Humufish's roughly 4–5 million units. GOOGL stock was down 1% in early premarket trading on Monday, following a breakout week with gains of 2.3%.

No discussion yet for this article

Get curated AI news from 200+ sources delivered daily to your inbox. Free to use.

Free · takes 30 seconds · unsubscribe anytime

5 minutes a day. The AI essentials.

200+ sources · Email / LINE / Slack