Google, a bit like all companies in the sector, has thrown itself headlong into the race for Artificial Intelligence, not only from a software point of view thanks to its LLMs, but also from a hardware point of view, in fact all the large companies have started a race also in the development of solutions to process AI software, so Google has followed in the wake of NVIDIA, AMD, Intel and even Microsoft and here he launched a very important project called Axionwhich consists of a processor capable of offering performance for the AI of its data centers 50% higher performance compared to common x86 solutions.
According to estimates calculated by Google, the new Axion platform would guarantee superior performance by 30% compared to “general-purpose ARM chips” and by 50% compared to “current generation x86 chips”.
Google bet
Google seems to be very confident about the adoption of these new chips, although these are only related solutions to data centers and their AI functionsin detail which CPUs will be used in those centers that manage and offer Google’s AI services in 8960 chip pods, thus guaranteeing a clear improvement in service performance which should even be visible to the end user.
Mark Lohmeyer of Google explained how, among other things, all the software platforms that natively support ARM can already be compatible since Axion is a sort of direct derivation of the lattertherefore there is no need to rewrite the various applications to be able to take advantage of these new processors, a sign that the aim is also to spread Axion beyond the walls of Google.
This is an important attempt by Google since it will shake an important monopoly held by Nvidia which holds 83% of chip production for data centers, an absurd value that is not easy to scratch.
