Alphabet’s Google is expanding its lineup of AI chips as it intensifies its efforts to catch up with Nvidia in a market worth hundreds of billions of dollars. Google has released two new versions of its Tensor Processing Unit (TPU), custom chips designed for training and running AI software. These new chips are capable of meeting confidential computing requirements, allowing models to access a wider range of information when processing queries—an advance that aligns with the industry's shift toward AI agents and the various security challenges they pose. The launch signals an acceleration in the pace of iteration within Google's chip division: previously, the business launched a new AI processor every two years, but this year it released two at once. The 8tx and 8i Max variants, set to be officially announced on Monday, are additional subdivision versions built on these two chips. Amin Vahdat, head of Google's chips, said the four TPU products each suit different needs, "which aligns with Google's core belief: the future will move toward specialization, and the room for gains from specialization remains enormous."