When we talk about the hardware for operating AI systems, the name Nvidia quickly comes up. In fact, the graphics chip manufacturer undoubtedly plays a central role in this area. What is often overlooked, however, is how important Google has become in this area. Google’s own models are no longer trained and operated on the self-developed Tensor Processing Units (TPUs). Many other AI companies also use TPUs – including Anthropic, Meta, Apple and many startups in this area. Last but not least, this is because the Google chips are considered particularly efficient to use.
