Microsoft picks new chip maker for AI co-processors on Azure, cutting Intel’s business in half – GeekWire

Victor Peng, CEO, Xilinx (Xilinx Photo)

Xilinx is reportedly the beneficiary of a decision by Microsoft Azure’s infrastructure wizards to add another chip supplier to its ranks, as it looks to serve more customers interested in machine learning.

Bloomberg reported Tuesday that Microsoft and Xilinx have reached an agreement where Xilinx chips will account for half of the co-processors currently used on Azure servers to handle machine-learning workloads, one of the strongest areas of growth in the cloud. To date, that was the exclusive domain of Intel’s Altera division, which the chip maker acquired in 2015 to offer both its general-purpose processors as well as reconfigurable chips like Xilinx’s co-processors to cloud computing vendors.

Flexible chips that can be configured to run machine-learning services have become hot commodities over the past several years. Xilinx was actually the first to bring the concept of FPGAs (field-programmable gate arrays) to the market in the 1980s, and it has been butting heads with Altera ever since, but for the most part the market for such chips was relatively limited.

That’s changing quickly as cloud vendors pour resources into research and development around machine learning, and as their customers start to experiment with services that could change the way they design and build their own businesses. The market for these chips has also changed as building and running your own data centers has fallen out of fashion, consolidating much of the hardware market around the major cloud vendors in the U.S. and China.

Earlier this year, Xilinx CEO Victor Peng announced plans to focus on data center customers, saying “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,” in a statement.