Google is intensifying its effort to expand adoption of its in-house Tensor Processing Units (TPUs), taking direct aim at Nvidia’s dominance in AI infrastructure by courting “neocloud” providers that have traditionally built their businesses around Nvidia GPUs.