Google alters its cloud playbook, enlisting partners to sell more AI chips

Google is widening access to its Tensor Processing Units by enlisting partners to sell the AI chips beyond Google Cloud. That marks a shift from using TPUs mainly to differentiate its own cloud platform toward treating custom silicon as a broader business, as demand for AI compute continues to outstrip supply. For builders, wider TPU distribution could create another option alongside Nvidia GPUs and other cloud accelerators, particularly for teams running workloads built around Google's AI software stack. The practical impact will depend on which partners participate, what TPU generations they offer, and whether pricing and availability are competitive. Those details have not yet been disclosed, but the strategy suggests Google wants TPUs to generate growth even when customers do not buy compute directly from Google Cloud.