Intel's long and expensive effort to become a contract chipmaker has won an important endorsement from one of the world's largest technology companies.
Microsoft and Intel announced an expanded collaboration under a long-term agreement to advance parts of Microsoft's Azure cloud infrastructure, in a release dated November 17 and published on Wednesday, November 18. Microsoft is the first company to announce that it will use Intel Foundry's EMIB advanced packaging technology for its Azure first-party silicon, the custom chips Microsoft designs for its own data centres.
The agreement also covers multi-year supply of Intel Xeon server processors, including next-generation chips code-named Diamond Rapids, and collaboration on memory efficiency, security and acceleration technologies.
What EMIB is
EMIB, short for Embedded Multi-die Interconnect Bridge, is Intel's technology for connecting multiple chips, or chiplets, within a single package. As chips become more complex, especially AI accelerators, manufacturers increasingly build them from several smaller chiplets connected together, rather than one large piece of silicon.
Advanced packaging has become a critical bottleneck in the AI chip supply chain. Taiwan Semiconductor Manufacturing Co.'s CoWoS packaging technology, used for Nvidia's AI chips and many others, has been in high demand, limiting how many AI chips can be produced. Alternative packaging capacity, such as Intel's EMIB, could help ease that constraint.
Bernstein estimates that standard EMIB packaging costs in the low hundreds of dollars per chip, compared with roughly $900 to $1,000 for CoWoS on a high-end AI accelerator, according to Tech Times.
What the release did and did not say
The announcement did not name the specific Azure chip that will use EMIB, its manufacturing process, its shipping date or the value of the agreement, Tech Times noted. Microsoft has said previously that it planned to make a chip on Intel's 18A manufacturing process, and there have been reports that a future version of its Maia AI accelerator could use Intel technology, but Microsoft has not confirmed those details.
Tech Times said the deal most likely adds a second packaging source for an unspecified chip, and that using EMIB packaging does not by itself show where the chip's silicon wafers will be made.
Xeon supply and performance
The agreement covers multi-year supply capacity of Intel Xeon processors, including Diamond Rapids, which Tech Times said is due in 2027 on Intel's 18A-P process with up to 256 cores.
The companies will also work on Xeon processors with CXL memory technology to improve memory efficiency across Azure, on Intel QuickAssist Technology for acceleration and on confidential computing. Intel reported, in vendor-run tests, up to 2.3 times greater performance and memory flexibility for some Azure virtual machines using CXL memory, and up to 2.5 times faster SQL database backups using QuickAssist. Tech Times cautioned that both comparisons used mismatched systems.
What the executives said
"Our customers need infrastructure built around the workloads that matter to them," said Saurabh Dighe, corporate vice president of Azure System and Architecture at Microsoft. "Together, we are connecting silicon and systems engineering to move the industry forward."
"Microsoft and Intel share a clear view of where AI infrastructure is headed," said Naga Chandrasekaran, executive vice president, chief technology and operations officer and general manager of Intel Foundry.




