Join the 155,000+ IMP followers

electronics-journal.com

Cloud AI Infrastructure with CXL Memory and Advanced Packaging

Microsoft and Intel extend their Azure collaboration, combining Xeon processors, CXL memory, QuickAssist acceleration and EMIB packaging for cloud AI workloads.

  www.intel.com
Cloud AI Infrastructure with CXL Memory and Advanced Packaging

Microsoft and Intel have expanded their long-standing collaboration under a long-term agreement covering selected areas of Azure cloud infrastructure, from server processors and memory architecture to data acceleration, confidential computing and advanced semiconductor packaging. The work targets compute-intensive cloud and AI workloads in data centres.

Roles of the Partners
Microsoft operates Azure and designs first-party silicon and server systems for its cloud. Intel supplies Xeon server processors and, through Intel Foundry, manufacturing and packaging technology. Microsoft is the first company to announce the use of Intel Foundry's EMIB (Embedded Multi-die Interconnect Bridge) advanced packaging for Azure first-party silicon. EMIB connects multiple dies within one package through small silicon bridges embedded in the substrate, an approach used to build large chips from smaller chiplets. The agreement also secures multi-year Intel Xeon supply capacity, including next-generation server CPUs codenamed Diamond Rapids, and extends cooperation across the Azure roadmap and Intel-based Azure hardware systems.

Memory Expansion, Acceleration and Confidential Computing
A core element is Compute Express Link (CXL), an interconnect standard that allows additional memory to be attached to the processor beyond the DIMM slots. In Azure M-series v4 virtual machines, CXL memory combined with Intel Flat Memory Mode in Xeon 6 processors delivers up to 2.3 times greater performance and memory flexibility for demanding workloads. In the tested configuration, a VM with 1 TB DRAM plus 1 TB CXL memory was compared with a DRAM-only VM.

Intel QuickAssist Technology (QAT) offloads compression and encryption from CPU cores to dedicated hardware. Azure offers what the companies describe as the first QAT-enabled virtual machines in the cloud, with up to 2.5 times faster SQL backups. For confidential computing, Intel Trust Domain Extensions (TDX) encrypt VM memory and CPU register state while confidential VMs are migrated between hosts, protecting sensitive data during live migration.

Deployment in Azure
The technologies are deployed in Azure virtual machine families and in Intel-based Azure hardware systems. According to both companies, the aim is to provide predictable, scalable and resilient infrastructure with more efficient memory use and stronger protection for sensitive workloads.

"Our customers need infrastructure built around the workloads that matter to them," said Saurabh Dighe, CVP, Azure System & Architecture, Microsoft.

"Microsoft and Intel share a clear view of where AI infrastructure is headed," said Naga Chandrasekaran, EVP, Chief Technology and Operations Officer and General Manager of Intel Foundry.

Expected Impact
The published figures indicate where gains are expected: memory-bound workloads benefit from larger addressable memory without moving to larger VM sizes, data protection tasks benefit from hardware offload, and confidential workloads can be migrated between hosts without exposing memory contents. Performance results are based on tests completed in November 2026 and depend on configuration.

Edited by Evgeny Churilov, Induportals Media - Adapted by AI.

www.intel.com

  Ask For More Information…

LinkedIn
Pinterest

Join the 155,000+ IMP followers