Google has expanded its eighth-generation Tensor Processing Unit lineup with two additional systems, TPU 8tx and TPU 8i Max, aimed at secure AI inference and workloads that need much larger context windows. The company announced the systems on November 9 in a post on the Google Cloud Blog.
The additions extend the TPU 8t and TPU 8i platforms Google introduced in April. Google says the new systems are not entirely new chip designs. Instead, they combine existing silicon with different compute, networking and packaging configurations to address specific customer requirements. The company said this is the first time it has introduced four TPU systems in a single year.
TPU 8tx is designed for governments and regulated businesses handling sensitive information. It pairs the silicon used in TPU 8t with custom AMD CPU hosts and adds hardware-level protections intended to keep data and model weights isolated even while they are being processed. Google says the system can deliver up to two times lower latency than TPU 8t with a near-zero performance penalty.
TPU 8i Max targets workloads that need very large context windows, including AI agents working across extensive software codebases or long video collections. Google built the system by combining two low-latency TPU 8i chips in a single package. The company says it doubles compute, on-chip SRAM and high-bandwidth memory bandwidth compared with TPU 8i, while fitting twice the silicon into the same rack footprint.
Google says nearly 100 customers now use its TPUs, including frontier AI labs, trading firms, retailers and universities. The company has also begun allowing customers to deploy TPUs directly in their own data centres, rather than relying only on Google Cloud infrastructure.
If this article helped you, please consider supporting our work. Every small contribution keeps Abijita.com independent and running.
