Google unveils chips for AI training and inference in latest shot at Nvidia
Google unveils custom AI training and inference chips competing with Nvidia.

“Google is packing ample amounts of static random access memory into a dedicated chip for running artificial intelligence models, following Nvidia's plans.”
The Weekly Silicon take
Google is pairing new chips for training and inference with a dedicated inference part that carries a large amount of on-chip SRAM, a design choice that follows plans Nvidia has signaled. SRAM keeps model data close to compute, which cuts latency and memory traffic, though it costs die area and limits capacity per chip. The split shows the industry treating inference as a separate hardware problem from training. Nvidia faces another large buyer building its own silicon, which matters most for inference spending where cost per query dominates. Google Cloud customers may gain lower-cost or faster serving options if the chips are offered to them, while model developers inside Google get hardware tuned to its own workloads. Foundries and packaging suppliers that build these parts are indirect beneficiaries as custom accelerators multiply. Watch for availability details: when cloud customers can rent the chips, in which regions, and at what pricing relative to Nvidia-based instances. Independent benchmarks on throughput and cost per token, along with named customers moving production workloads, would confirm the design works. Nvidia's product and earnings commentary will show how directly it responds.
AI score breakdown
A composite of 7.00/10 put this story at #6 for Thursday, April 23, 2026, driven mostly by technology breakthrough (8/10) and economic impact (8/10).
Composite is the weighted sum of the five dimensions. How scoring works →
Related stories
- Nvidia stock drops over 30% as Jensen Huang resigns over scandal with alleged espionage ties · 2026-11-20
- Amid Anthropic’s IPO, Google, Microsoft and OpenAI Drop New AI Models · 2026-11-19
- ‘Five-alarm fire’: Nvidia’s fall from grace sets off Washington AI industry scramble · 2026-11-19
- Where was Nvidia’s security team? · 2026-11-15
- ~77% of all new "Success" self-help books on Amazon are likely written by AI, with 1 author, Noah Felix Bennett, publishing a stunning 74 books in mid-2025 alone, at a rate of >1 per day. Richard Trillion Mantey, who has published hundreds of books, was assessed to have used AI for every single book · 2026-04-09