As early as 2014, Microsoft first announced that it was exploring the possibility of using FPGAs in data centers to speed up some of Bing's tasks. Later in 2016, Microsoft announced that they had extended the use of FPGA in Azure and Office 365. Each group uses FPGA according to their needs. For example, Azure uses FPGA for faster network connectivity, while Office 365 uses it for machine intelligence activities.
Microsoft currently has the world's largest custom development field programmable gate array (FPGA) deployment, covering 15 countries and five continents. These FPGAs can be used to accelerate AI to near real-time, making Azure the foundation of the world's first AI supercomputer. Earlier this year at the Build Developers Conference, Microsoft announced the launch of Project Brainwave, bringing powerful FPGA capabilities to Azure cloud customers.
So far, the FPGA used by Microsoft has been supplied by a company called Altera, which was later acquired by Intel in 2015. According to Bloomberg's latest report, Xilinx has now won orders for Microsoft data center to provide FPGA. Microsoft will use Xilinx chips in more than half of its servers instead of Intel chips.
When asked about the story, Microsoft responded that there was no change in the procurement of existing infrastructure and products.