Microsoft Is Designing Its Own AI Chips for the Next HoloLens
On Sunday, the software giant unveiled a new microprocessor of its own design that will go into future hardware products, such as the next version of its HoloLens augmented reality glasses, to perform artificial intelligence tasks. The AI coprocessor will analyze visual, auditory, and other sensor data locally on the device instead of relying on sending the information to a cloud-based server for review.
“This is the kind of thinking you need if you’re going to develop mixed reality devices that are themselves intelligent,” Marc Pollefeys, Microsoft’s director of science for the HoloLens, wrote in a blog post. “Mixed reality and artificial intelligence represent the future of computing, and we’re excited to be advancing this frontier.”
Get Data Sheet, Fortune’s technology newsletter.
Previously, Microsoft and others have focused on developing their own custom chips largely for use in cloud data centers. But after Apple had great success developing its own line of A-series processors for the iPhone, followed by its M-series motion processors, more companies are looking at custom chips for mobile devices as well.
The exploding use of AI programs for everything from image and speech recognition to on-the-fly language translation is also fueling the drive for custom chips as each company perfects its own apps and algorithms.
Google (GOOGL) is using its own chip, called Tensor Processing Units, for AI programs running on its servers. Apple (AAPL) is rumored to be working on an AI chip ready for future iPhones. Nvidia (NVDA), Advanced Micro Devices (AMD), and Intel (INTC) are also pitching more general-use chips for AI tasks.
Microsoft’s (MSFT) earlier AI chip efforts relied on a reprogrammable chip, called a field programmable gate array, or FPGA, made by Intel’s Altera unit.