Google Takes Aim at Nvidia With New AI Chips Built for Privacy and Massive Memory

Google has unveiled two systems in its eighth-generation Tensor Processing Unit (TPU) range, targeting privacy-sensitive and memory-intensive AI workloads. The launch expands Google’s challenge to Nvidia by offering businesses more specialised hardware and software options, potentially giving AI developers greater choice and bargaining power.

The TPU 8tx uses hardware-level confidential computing to protect data and model parameters during processing; Google says it can cut latency by up to half compared with the earlier TPU 8t. The TPU 8i Max combines two chips in one package, doubling computing power, on-chip memory and memory throughput versus TPU 8i, while Google says its open-source TPU Sync software can reduce some transfer delays and response startup times by more than 90%. These comparisons are with Google’s own earlier hardware; pricing and availability dates were not disclosed.