Microsoft plans to develop its own AI chip, FPGA vendor Intel should worry?

CNBC reported on June 12 that Microsoft’s Azure Public Cloud Department issued at least three recruitment announcements at the end of March to solicit talents for the development of artificial intelligence (AI) chip functions. In April this year, the team opened a chip program manager and software and hardware. Coordinated Design and AI Accelerated Optimization Engineers.

Sundar Pichai, Google's chief executive, released the third generation TPU (Tensor Processor Unit, TPU3) at the 2018 Google I/O conference on May 8. The new TPU can help Google improve AI speech recognition and find photos With the specific objects in the film and understand the emotions hidden behind the text. Moor Insights & Strategy analyst Patrick Moorhead estimates that Google spends $2-3 billion on the TPU plan.

Microsoft’s Bing blog announced on March 26 that it has built deep neural networks through Intel’s Arria and Stratix field programmable logic gate array (FPGA) chips, and has developed a deep learning project called “Project Brainwave”. Accelerating platform. Bing's smart search function is to use this platform's huge computing power to perform large-scale machine reading comprehension.

Reynette K. Au, vice president of Intel's Programmable Solutions Group Marketing Group, issued a press release (Sourced) at the time. He pointed out that with Intel's FPGA chip, Microsoft Bing's smart search can understand the user's intentions and provide answers, instead of providing a Strings need manual processing of web links.

According to a report released by the Nikkei Asian Review in November last year, the analysis by Nikkei and Elsevier revealed that the top quoting number of AI-related research reports during the period 2012-2016 (note: generally considered as a research quality index) is Microsoft.