Google is building Frozen v2, an AI chip aimed at 6–10x better token efficiency
ocean_protocol · reddit · 2026-07-22
Google is reportedly working on Frozen v2, a new AI chip expected around 2028 that bakes parts of Gemini’s architecture into silicon rather than using a generic accelerator design.
- The goal is to reduce computation and data movement for inference.
- The expected efficiency gain is 6–10× more tokens per watt versus Google’s current AI chips.
- The tradeoff is flexibility: it may only work well with future Gemini models if the underlying architecture stays stable.
- The report frames it as a trial run rather than a TPU-scale replacement.
Related event: Google Reportedly Developing Gemini-Specific Chip Frozen v2(10 posts)→
More from Infra
- EdgeRock open-sourced: 72 MiB KV cache at 256K context, 99.1% smaller than Qwen3.5-4B — alifcoder · 2026-10-27
- EdgeRock-4B holds KV cache at 72 MiB across 256K context, 99.1% smaller than Qwen3.5-4B — alifcoder · 2026-10-27
- RTX 5080 price jumps from $1658 to $2318 in ten days on Amazon — Frizzy-MacDrizzle · 2026-10-27
- ComfyUI on 12GB VRAM: measured peaks show offloading turns 24s into 11 minutes — Impossible-Mail-4619 · 2026-10-27
- Neon rebuilds its docs search entirely on its own stack: Lakebase Search, Functions, AI Gateway — matei_zaharia · 2026-10-27
- PyTorch details 2.12x faster on-device streaming ASR on Arm SME2 via ExecuTorch — PyTorch · 2026-10-27