Google doesn't pay the Nvidia tax. Its new TPUs explain why.

Google previewed its eighth-generation TPUs in two versions: TPU 8t for frontier model training and TPU 8i for low-latency inference.

The new chips are built to better match AI workloads, with major upgrades in scale, bandwidth, memory, and networking.

Google is using custom silicon to cut AI infrastructure costs and optimize hardware separately for training and inference.

The move also underscores Google’s push to challenge Nvidia’s dominance in frontier AI compute.