Google is currently hosting its Cloud Next 2024 event, and some major announcements are coming from the event. Now, the tech giant has launched its brand new Google Axion processor. With this processor, Google is trying to get into the Arm-based CPUs designed explicitly for data centres space.
Amin Vahdat, VP/GM, of Machine Learning, Systems, and Cloud AI, said, “Today, we are thrilled to announce the latest incarnation of this work: Google Axion Processor, our first custom Arm-based CPUs designed for the data center. Axion delivers industry-leading performance and energy efficiency and will be available to Google Cloud customers later this year.”
Also read: Google Gemini Code Assist: Here’s how you can check your code with this AI tool
The Axion has been built using Arm’s Neoverse V2 CPU. When compared to the competition, Google says that Axion will provide better performance. It revealed that Axion provided 30% improved performance over the fastest general-purpose Arm-based tools in cloud computing. When compared to the latest x86-based virtual machines (VMs), it gave a whopping 50% boost.
Google also said that Axion is 60% more power-efficient than its x86-based counterparts. The tech giant has already fused Axion with services such as BigTable and Google Earth Engine. Eventually, it will spread across its ecosystem.
Also read: Gemini in Google Messages: Chat with AI to draft messages, brainstorm ideas & more
This venture into data centre CPUs might look new for Google, but it has been working on it for many years now. The primary focus of the company was to develop it for consumer products. This furthers Google’s previous attempts, such as that with its Arm-based Tensor chip, which was initially incorporated in Pixel smartphones.
With this development, Google comes in direct competition with Amazon, which has led the market with the Amazon Web Services (AWS) Graviton processors since 2018, and NVIDIA with its Grace CPU and Ampere. Even Google’s “Tensor Processing Unit” (TPU), designed for data centres, has been in use internally since 2015, with public availability since 2018.
