Google AI For Cloud: Tech Giant Partners with ARM to Create High-Performance AI Chip

Google has upped the ante in the cloud race against Amazon and Microsoft with its custom-designed AI chip, Axion. This move signifies the search giant's continued push to redefine the boundaries of computing.
At its recent Cloud Next conference in Las Vegas, Google unveiled a custom-built ARM-based server chip to make cloud computing more affordable. The chip, expected later in 2024, aims to bring cloud computing within reach of budget-conscious users.

"Axion delivers industry-leading performance and energy efficiency and will be available to Google Cloud customers later this year," Amin Vahdat, VP/GM of Machine Learning, Systems, and Cloud AI, noted in a blog post.

Google enters the ARM fray: Taking on cloud giants

Google has finally joined the cutthroat cloud infrastructure market, dominated by established players like Amazon and Microsoft. These companies have used custom chips for years. This growing market allows companies to rent computing resources in remote data centres, paying only for what they use.

Alphabet, Google's parent company, relies heavily on advertising for revenue, accounting for roughly 75 percent. Meanwhile, the cloud segment is experiencing significant growth. Cloud services, including corporate productivity applications, now represent nearly 11 percent of Alphabet's revenue and are already profitable.

Despite this growth, Google Cloud held a 7.5 percent market share in 2022, trailing behind Amazon and Microsoft's combined dominance of around 62 percent, according to Gartner estimates.

Google complained to the UK's Competition and Markets Authority (CMA) last year, alleging anti-competitive practices by Microsoft in cloud computing. Amazon joined Google in accusing Microsoft of anti-competitive licensing practices.

Market leader Amazon Web Services took the lead in ARM-based server chips for cloud computing with the launch of Graviton in 2018. "Almost all their services are already ported and optimised on the ARM ecosystem," Chirag Dekate, an analyst at technology industry researcher Gartner, told CNBC in an interview.

Companies like Datadog, Elastic, Snowflake, and Sprinklr have adopted Graviton for their cloud services. Alibaba unveiled ARM processors in 2021, followed by Microsoft in November of the same year. Last year, the Chinese e-commerce company launched the latest version of its AI, Tongyi Qianwen 2.0, to compete with Microsoft and Amazon.

Google is no stranger to ARM processors.

Google isn't a newcomer to ARM technology. The company dipped its toes into ARM technology in 2022 by offering virtual machines powered by Ampere's ARM-based chips.

Economic concerns are driving a shift towards ARM-based cloud computing. Many organisations are now porting applications to ARM machines to benefit from their cost-efficiency.

When ARM Holdings filed for an IPO last year, it cited Amazon's claim that Graviton chips deliver up to 40 percent better price performance than traditional x86 processors used by Intel and AMD.

Google uses ARM-based servers internally, including YouTube ad processing, BigTable/Spanner databases, and BigQuery analytics. A spokesperson said these workloads will migrate to Google's cloud-based ARM instances (Axion) upon launch.

Google's spokesperson confirmed that Datadog, Elastic, OpenX, and Snap are among the initial adopters of Axion chips. ARM chips could cut emissions for some cloud tasks. Google's tests show their Axion chip VMs use 60% less energy than similar x86 options, thanks to ARM's more straightforward design (standard in phones), Google cloud chief Thomas Kurian wrote in a blog post.

Axion delivers 60 percent better energy efficiency and up to 50 percent faster performance compared to standard x86 VMs, making it ideal for specific cloud workloads.