Google reveals its new Axion CPU line, specifically customized for cloud and data servers

Within the framework of the Google Cloud 2024 conference taking place today, Google officially announced Axion, the first Arm-based processor chip line designed specifically for data centers and clouds.

In a blog post, Google stated that Axion is the latest in a 'long-term plan' by the company to design its own custom processors. Back in 2016, Google introduced the first generation Tensor processor chip line also produced for data centers and brought quite a lot of success. Google's Pixel smartphone line since 2021 has used a custom System on Chip (SoC), also branded under the Tensor brand.

Google claims the new Axion processors, using Arm's Neoverse V2 platform, will deliver up to 30% better performance than previous Arm-based CPUs made for the data center, while also allowing up to 50% better performance than existing x86 chips on the market. Additionally, the Axion series' energy savings are also expected to be 60% more efficient than similar x86-based virtual machines. Google already uses Axion in services like BigTable and Google Earth Engine, and will expand to more platforms in the future.

Axion-based servers will be available to Google Cloud customers around the end of 2024. In the meantime, interested businesses can register on Google's website to learn and experience the benefits that virtual machines using Axion chips bring.

The launch of the Axion processor line could put Google in competition with Amazon, which is currently the leader in supplying Arm-based CPUs to data centers. Amazon released the Graviton processor in 2018, then rolled out a second and third version over the next two years. All have made great contributions to Amazon's cloud platform business - Amazon Web Services (AWS).

Another giant, NVIDIA, also released the first Arm-based CPU for data centers in 2021 called Grace. Followed by a series of other large and small companies present in the market. This will be a new competitive field that is not simple for Google.