Fact◆Article
Google's 8th generation TPU ships in two flavors: TPU 8t for training (raw throughput) and TPU 8i for inference (latency and chip-to-chip speed)
Google's 8th gen TPU introduces two purpose-built variants: TPU 8t optimized for training throughput and TPU 8i optimized for inference latency. ✦ AI generated
system-design-refresh · ByteByteGo Newsletter · 2026-08-15 · original ↗
At Cloud Next '26, Google unveiled its 8th generation, and for the first time it ships in two flavors. TPU 8t is built for training, where raw throughput wins. TPU 8i is built for inference, where latency and chip-to-chip speed matter most.
Read full article ↗excerpt · fair-use quotation
- ·Unveiled at Cloud Next '26
- ·First-gen TPU shipping in two variants
- ·TPU 8t: built for training, raw throughput wins
- ·TPU 8i: built for inference, latency and chip-to-chip speed matter
Around this claim