Google adds TPU 8tx for confidential computing and TPU 8i Max dual-chip system to its eighth-generation TPU line
Both are new system configurations of existing TPU 8t and 8i silicon, first reported by Bloomberg; Google has not been quoted on prices.
Why it matters
Google is turning its in-house AI chips into a broader product line for outside customers, and these two systems target gaps where Nvidia has been strong: confidential computing for regulated buyers and memory bandwidth for agent workloads. Rather than a new chip generation, Google repackaged existing silicon, which it says lets it ship variants in months rather than years. With them it has shipped four TPU systems in one year for the first time and counts nearly 100 external TPU customers. Watch for availability dates, pricing and which cloud customers adopt the confidential 8tx first.
Key facts
- TPU 8tx pairs TPU 8t silicon with a custom AMD host CPU
- 8tx keeps data and model weights protected during processing
- 8tx: up to twice the latency reduction versus TPU 8t
- TPU 8i Max puts two TPU 8i chips in one package
- 8i Max doubles compute, memory and memory bandwidth
- Nearly 100 external TPU customers