Google Axion is its first data centre CPU based on Arm

At Google Cloud Next 2024, Google introduced its new Axion processor, an Arm-based CPU specifically designed for data centres using Arm’s Neoverse V2 CPU. This is Google’s first foray into Arm-based data centre processors, and the company claims Axion offers a 30 percent improvement in performance compared to its fastest general-purpose Arm-based tools in the cloud. Additionally, Axion is said to outperform the latest x86-based VMs by 50 percent and is 60 percent more energy-efficient. Axion is already in use for services like BigTable and Google Earth Engine, with plans to expand its use further.

Read: MultiChoice is now 40% owned by Canal+

The launch of Axion places Google in direct competition with Amazon, which has been leading the field with its Arm-based CPUs for data centres, known as Graviton. Amazon Web Services (AWS) released the Graviton processor in 2018, with subsequent iterations over the next few years. Similarly, NVIDIA introduced its first Arm-based CPU for data centres, Grace, in 2021, while companies like Ampere have also made significant progress in this area.

Google has a history of developing its processors, primarily for consumer products. Its original Arm-based Tensor chip debuted in the Pixel 6 and 6 Pro smartphones in late 2021, with updated versions powering subsequent Pixel phones. Before that, Google developed the Tensor Processing Unit (TPU) for its data centres, which have been used internally since 2015, publicly announced in 2016, and available to third parties since 2018.

Arm-based processors offer lower costs and higher energy efficiency. Google’s announcement follows Arm CEO Rene Haas’s warning about the energy consumption of AI models like ChatGPT, which he described as “insatiable” regarding electricity usage. Haas emphasized the need for increased efficiency to maintain the pace of AI advancements, as AI data centres could consume up to 20-25 percent of US power requirements by the end of the decade. This contrasts with the current consumption of around four percent or less, underscoring the importance of sustainable development in AI technology.