Microsoft and Intel Expand Azure Collaboration in Long-Term Deal Covering EMIB Packaging and Xeon Processors
Key Facts
Microsoft and Intel announced on November 18, 2026, an expansion of their decades-long collaboration through a long-term agreement to develop specific aspects of Azure’s infrastructure, in a move aimed at meeting rising demand for cloud and AI services. Intel said the agreement expands cooperation between the companies, without disclosing its financial value. The partnership places Intel technologies across multiple areas of Azure, including manufacturing silicon Microsoft develops for internal use, providing server processor capacity, and improving memory and security. It comes alongside Microsoft’s development of its custom processors, including Maia and Cobalt, making the collaboration part of an infrastructure that combines the company’s own silicon with Intel processors. The announcement did not specify that Maia 200 or Cobalt 200 chips would use Intel’s packaging technology.
The companies said Microsoft is the first company to announce the use of Intel Foundry’s EMIB advanced packaging technology for Azure silicon that Microsoft develops for internal use. The technology connects multiple dies within a single package, while the announcement focused on its use in Azure silicon without naming specific chips. Intel confirmed that the agreement includes a multiyear commitment to provide Xeon processor capacity, including next-generation Diamond Rapids server units. Combining advanced packaging with processor supply gives the agreement a scope beyond purchasing central processing units alone. The collaboration also links Microsoft’s custom hardware design with Intel’s manufacturing, packaging and processing capabilities.
The work extends to Intel Xeon processors equipped with Compute Express Link, or CXL, technology, with the aim of improving memory efficiency and supporting innovation across Azure systems. The companies cited the performance of Azure M-series v4 virtual machines as an example of this effort, saying they achieved up to 2.3 times higher performance and memory flexibility when using CXL memory and Intel Flat Memory Mode on Xeon 6 processors. The improvement is tied to workloads requiring greater memory capacity, according to the companies. The example indicates that the collaboration is not limited to introducing new processor generations, but also covers how memory is expanded and managed within servers. The announcement did not specify the scope of the performance measurement beyond M-series v4 machines and the stated technical configuration.
In another area, Microsoft and Intel said SQL database backups on Azure became up to 2.5 times faster through the first cloud virtual machines enabled by Intel QuickAssist Technology. The result targets a specific operational task, backup, while the announcement did not provide an absolute time for the process or a comparison with another service. The companies linked the technology to improved performance and reliability, and said these machines are the first of their kind in the cloud to be enabled by QuickAssist. This adds another practical use of the collaboration alongside improved memory performance in M-series v4 machines. The companies did not specify the availability of these machines or the scope of SQL workloads covered by the result.
The agreement also addressed protection for confidential virtual machines as they move between hosts. Intel said the collaboration includes encrypting virtual machine memory state and processor registers during the transfer using Intel Trust Domain Extensions. The mechanism is intended to keep memory data and processor state encrypted when a machine moves from one host to another, according to the announcement. This adds a security dimension to the processor, memory and packaging areas announced by the companies. A Microsoft executive said the collaboration combines "advanced packaging, custom compute, memory and security capabilities to power Azure compute and continuous innovation across Azure."
The collaboration comes as Microsoft continues to roll out its own processors while also relying on processor suppliers. On October 12, 2026, the company said Maia 200 had entered production and was already serving production workloads, including Microsoft Copilot, and that Maia 300 was coming in 2027. In a separate announcement, Microsoft said Cobalt 200 virtual machines had become generally available in 14 Azure regions, with plans to expand to 23 regions by the end of 2026. The company said Cobalt 200 delivers up to 50% better performance compared with Cobalt 100. These announcements provide broader context for Azure infrastructure, but do not specifically link Maia or Cobalt chips to Intel Foundry services under the new agreement.
Microsoft’s announced plan is to expand Cobalt 200’s geographic presence, then increase its role in new compute capacity. The company said these machines would account for a significant share of new Azure Compute capacity starting in 2027. The Intel agreement confirms a multiyear supply of Xeon capacity, including Diamond Rapids, but does not specify a delivery schedule or volume for those processors. Azure’s announced plans therefore include two parallel tracks: expanding the use of Microsoft’s own processors and strengthening collaboration with Intel on processors, packaging, memory and security. The companies did not disclose the agreement’s value or figures for the capacity it will add to Azure.