India, April 23 -- The artificial intelligence (AI) race is no longer just about models. It is now deeply about infrastructure, cost, and control. Over the past few years, as enterprises rushed to build and deploy AI systems, one company quietly dominated the backbone of this revolution: Nvidia. Its graphics processing units (GPUs) became the default engine powering everything from large language models to enterprise AI workloads.

Google's latest move to redesign its AI chips signals a deeper strategic shift. By separating chips for training and inference in its eighth-generation tensor processing units (TPUs), the company is not just launching new hardware. It is rethinking how AI workloads should be built, scaled, and optimised for the...