A few months back, I was able to uncover the fact that Microsoft had canceled its upcoming Hololens 2 release as the company shifted its attention to a truly next generation release. Today, the company revealed that they are building a new, custom fabricated chip, for the next generation device that will bring AI from the cloud down to the local device.
The second generation of the company’s Holographic Processing Unit (HPU) will include an AI coprocessor that will include support for Deep Neural Networks. The reason for the need for a new type of HPU is that offloading intensive tasks to the cloud results in an unacceptable amount of latency when working with objects in real-time which means to solve this problem, Microsoft needs to bring ‘AI’ processing directly to the HoloLens.
Microsoft has taken a keen interest in designing its own chips, the company has previously architected its own silicon with Kinect, field programmable gate arrays (FPGA) and of course, the first HPU. And considering the company is investing heavily in AI, it’s not all that surprising to see them building out new silicon that is designed specifically for this task as its use-case reaches far beyond the Hololens headset.
Microsoft isnt saying when the next generation Hololens will be available for everyone but insiders have told me to expect 2018 at the earliest with a more likely goal of 2019 for broad availability.