Google announced two additions to its eighth-generation Tensor Processing Unit (TPU) portfolio on November 9, 2026: the TPU 8tx and the TPU 8i Max. These new systems aim to address specific enterprise AI computing needs, namely confidential data processing and the handling of large context windows. This marks the first time Google has introduced four TPU systems within a single year, following the launch of the TPU 8t and TPU 8i in April 2026.The TPU 8tx is designed for confidential computing, ensuring data remains protected even during active processing. This system integrates Google's TPU 8t silicon with fifth-generation AMD EPYC processors and Google's Boardfly network, which is optimized for low latency. Google states this configuration aims to reduce latency by up to 50 percent compared to the TPU 8t, with minimal performance impact from the added security features. The hardware-level security aims to isolate sensitive data and model weights, a critical concern for businesses adopting AI.The TPU 8i Max focuses on addressing the "memory wall" challenge in AI, where processing power can be bottlenecked by insufficient memory bandwidth. Google engineers combined two TPU 8i chips into a single package to create the TPU 8i Max. This dual-chip configuration reportedly doubles the compute performance, on-chip SRAM capacity, and high-bandwidth memory (HBM) bandwidth compared to a single TPU 8i. By fitting twice the silicon into the same rack footprint, the TPU 8i Max allows for greater compute density in data centers. This increased memory capacity and bandwidth is intended for AI applications that process extensive documents, large software projects, or lengthy video sequences, which demand very large context windows.Amin Vahdat, Google's chief technologist for AI and infrastructure, emphasized that these new systems are built around existing eighth-generation TPU chips, with modifications to packaging, memory, and networking. This approach allows Google to rapidly deliver specialized supercomputers to meet customer and research needs in months rather than years. Google's strategy involves co-designing its TPUs in partnership with Google DeepMind and incorporating feedback from external organizations.Google's broader eighth-generation TPU portfolio now includes four systems: the TPU 8t for large model training, the TPU 8i for inference tasks, and the newly introduced TPU 8tx and TPU 8i Max. The company has stated that its custom chips are gaining commercial traction, with nearly 100 customers utilizing TPUs, including frontier labs, trading firms, retailers, and universities. Some customers have also begun installing TPUs directly in their own data centers.