Key relationships in depth
ARM Holdings & Apple
Arm Holdings licenses the instruction set architecture (ISA) that Apple's custom chips are built on. Every A-series and M-series chip uses the Arm architecture, which Apple customizes heavily through its own microarchitecture design. Apple has been an Arm licensee since the early 1990s and was one of the original co-founders of Arm as a spinout from Acorn Computers in 1990.
Why it matters: Apple holds an Arm architectural license, the most permissive tier, which allows it to design its own cores rather than using Arm's standard Cortex designs. This license is what enables Apple to build chips like the M4, which dramatically outperform standard Arm implementations. The license fee is a small per-chip royalty relative to Apple's total silicon spend.
Microsoft uses Arm's Neoverse architecture for its Azure Cobalt 100 processor, a custom data center CPU deployed in Azure's Cobalt virtual machines. Azure Cobalt 100 is built on 128-core Arm Neoverse N2 and delivers 40% better performance per watt than comparable x86 VMs. Microsoft also distributes Arm-based Windows on Snapdragon PCs through its Surface and OEM partnerships.
Why it matters: Microsoft's commitment to Arm in Azure reduces its Intel dependence and lowers power consumption per compute unit at scale. With $80B in FY2025 data center capex, even a 10% improvement in compute efficiency translates to billions in operational savings. Additionally, Microsoft's Windows on Arm expansion through Snapdragon X Elite PCs gives Arm a foothold in the $250B PC market.
ARM Holdings & NVIDIA
NVIDIA licenses Arm's CPU architecture for its Grace CPU, which pairs with NVIDIA GPUs in the Grace Hopper Superchip. NVIDIA Grace is Arm Neoverse-based and powers the GH200 and GB200 AI supercomputers that are now deployed in Tier-1 AI data centers globally. NVIDIA's own data center CPU strategy runs entirely on Arm architecture.
Why it matters: The Grace Hopper Superchip places CPU and GPU on the same module with NVLink high-bandwidth interconnect, delivering 3-7x more bandwidth than a PCIe-connected system. This architecture is essential for LLM inference workloads that require fast CPU-to-GPU data transfer. NVIDIA's choice of Arm over x86 validates Arm's server CPU roadmap and delivers licensing royalties on every Grace chip sold.
Google uses Arm's Neoverse architecture for its custom Axion processor, a data center CPU deployed in Google Cloud's C4A instances. Google Axion delivers 50% better performance than comparable x86-based VMs according to Google and is used for general-purpose compute workloads on GCP. Google also uses Arm-based chips in its Pixel smartphones via custom Tensor chips.
Why it matters: Google's decision to build Axion on Arm architecture reflects the industry-wide shift from x86 to Arm in cloud data centers. Arm's energy efficiency advantage (up to 60% better performance per watt than x86) directly reduces Google's data center electricity costs. For Google's $75B capex year, efficient computing architecture translates to billions in operating savings.
Relationships are compiled by VexMap from company filings, announcements and reporting, and last reviewed September 2026. Spotted something wrong? Tell us. Not investment advice.