CAMBRIDGE, UK — The architectural balance of power across global cloud computing infrastructure has permanently tilted. According to quarterly server socket census data released by Mercury Research, processors based on the Arm instruction set architecture now account for 41.2% of all active compute cores deployed inside tier-one hyperscale datacenters.
The milestone caps a decade-long engineering crusade that began as an experiment in power efficiency and has evolved into the dominant architectural blueprint for modern cloud computing.
Why Hyperscalers Build Custom Arm Silicon
Rather than purchasing standard off-the-shelf x86 server processors from traditional vendors, hyperscale cloud operators have leveraged Arm’s flexible IP licensing model to architect custom, application-specific system-on-chips:
- Amazon Web Services deployed millions of custom Graviton4 cores, citing up to 40% better price-performance for containerized workloads.
- Microsoft Azure expanded deployments of its Cobalt 100 processors across its core Microsoft 365 and SQL database infrastructure.
- Google Cloud migrated substantial YouTube and Google Ads processing pipelines onto custom Axion processors.
“In a datacenter constrained by strict electrical power caps, every watt saved by an efficient Arm core translates directly into more AI acceleration capacity,” remarked market analyst Marcus Thompson. “The x86 duopoly no longer dictates the roadmap of modern cloud infrastructure.”