Google TPU with Chip‑Level Encryption

Google is shipping a TPU variant (TPU 8tx) that performs encryption at the chip level so data can remain encrypted while the accelerator processes it, targeting regulated and public‑sector workloads. It also released a dual‑chip TPU (8i Max) specifically tuned for long‑context AI tasks like drug discovery, math and coding by combining two TPU dies for more memory and compute. — This marks a trend toward accelerator specialization for privacy and long‑context AI, shifting how cloud providers, regulators, and enterprises think about hardware, procurement, and model deployment.