Google is making its own custom Arm-based CPU to support its AI work in data centers and introducing a more powerful version of its Tensor Processing Units (TPU) AI chips.
Google’s new Arm-based CPU, dubbed Axion, will be used to support Google’s AI workloads before it rolls out to business customers of Google Cloud “later this year.” The Axion chips are already powering YouTube ads, the Google Earth Engine, and other Google services.
Reuters reports that the Axion Arm-based CPU will also offer 30 percent better performance than “general-purpose Arm chips” and 50 percent more than Intel’s existing processors.
Google is also updating its TPU AI chips that are used as alternatives to Nvidia’s GPUs for AI acceleration tasks.
Google’s announcement of an Arm-based CPU comes months after Microsoft revealed its own custom silicon chips designed for its cloud infrastructure.
Microsoft has built its own custom AI chip to train large language models and a custom Arm-based CPU for cloud and AI workloads.
Amazon has also offered Arm-based servers for years through its own custom CPU, with the latest workloads able to use Graviton3 servers on AWS.
Google, like Microsoft and Amazon before it, can now reduce its reliance on partners like Intel and Nvidia, while also competing with them on custom chips to power AI and cloud workloads.
Tensor Processing Units (TPU) are AI chips developed by Google that are being enhanced with more power. The company is also developing its own Arm-based CPU specifically for its data center AI work. Google’s upcoming Arm-based CPU, called Axion, will be utilized to support the company’s AI workloads prior to being made available to Google Cloud business customers “later this year.”. “.
Google Earth Engine, YouTube advertisements, and other Google services are already powered by Axion chips. According to a statement provided to Reuters by Mark Lohmeyer, vice president and general manager of compute and machine learning infrastructure at Google Cloud, “we’re making it easy for customers to bring their existing workloads to Arm.”. “Axion has open foundations, but users of Arm anywhere can quickly integrate Axion without needing to rewrite or re-architect their apps.”. “.
Customers will be able to use Google’s Axion CPU in cloud services such as Cloud Batch, Dataproc, Dataflow, Google Compute Engine, and Google Kubernetes Engine, among others, according to Google. According to Reuters, the Axion Arm-based CPU will also perform 50% better than Intel’s current processors and 30% better than “general-purpose Arm chips.”.
Additionally, Google is making updates to its TPU AI chips, which are GPU substitutes for Nvidia in AI acceleration applications. According to Lohmeyer, “TPU v5p is a next-generation accelerator that is specifically designed to train some of the largest and most demanding generative AI models.”. 8,960 chips make up a single TPU v5p pod, more than twice as many as a TPU v4 pod.
Following Microsoft’s announcement of its own custom silicon chips intended for its cloud infrastructure, Google has announced the release of an Arm-based CPU. Microsoft has developed a bespoke Arm-based CPU for cloud and AI workloads, as well as its own custom AI chip for training large language models. With its own proprietary CPU, Amazon has also provided Arm-based servers for years. The most recent workloads can utilize Graviton3 servers on AWS.
Google will make these chips available for cloud services that companies can rent and use, rather than selling them to consumers. The executive in charge of Google’s internal chip operations, Amin Vahdat, told The Wall Street Journal that “becoming a great hardware company is very different from becoming a great cloud company or a great organizer of the world’s information.”.
Similar to Microsoft and Amazon in the past, Google can now depend less on partners like Intel and Nvidia while still competing with them for custom chips that support workloads in the cloud and artificial intelligence.
