Recently, Microsoft dropped Intel to exclusively power its Azure services, in favor of another chip maker, Xilinx. Microsoft's addition of another chip supplier would replace up to half of its Intel processors currently powering Azure. As noted in a report by GeekWire, Microsoft made the move "to serve more customers interested in machine learning" and help entice more companies to move away from running their data centers and moving more services to Azure.
Xilinx CEO Victor Peng announced plans earlier this year to focus exclusively on data center customers. In a statement, Peng said:
“Data center is an area of rapid technology adoption where customers can quickly take advantage of the orders of magnitude performance and performance per-watt improvement that Xilinx technology enables in applications like artificial intelligence (AI) inference, video and image processing, and genomics."
Microsoft and Xilinx made an agreement that Xilinx chips will be used for half of the co-processors that are currently used on Azure servers for machine-learning workloads. Previously, Intel's Altera division was exclusively used by Microsoft for Azure. Increasingly, flexible chips that can be customized for machine-learning are becoming more valuable for cloud services.
In the 1980's, Xilinx was the first chip maker that proposed the concept of FPGAs (field-programmable gate arrays). FPGAs are becoming increasingly popular as more companies stop running their own data centers and move their data to cloud services such as Azure.