With ARM’s Edge AI chips, AI doesn’t need the cloud anymore. Two new ARM chip designs, ARM Cortex-M55 and Ethos-U55, can run AI processing on the edge – no cloud needed. Specs-wise, ARM Cortex-M55 is the latest model in ARM’s Cortex-M line of CPUs. Compared to previous Cortex-M generations, it should represent a 15 times improvement in machine learning performance and a 5 times improvements in digital signal processing. And when you need a little bit more power, Ethos-U55 NPU comes to the rescue. Indeed, it has 32 times better performances compared to Cortex-M55, representing a staggering 480 times improvement in processing power over Cortex-M CPUs.
As for any other ARM chip, ARM’s Edge AI chips won’t be manufactured by ARM. Also worth mentioning, they are not designed for phones and tablets. Rather, they are meant to be used in IoT/edge computing scenarios, such as AI-powered cameras and temperature detection systems.
Edge AI is one of the most powerful trend in 2020. No doubt that many IT giants other than ARM look at Edge AI with keen interest. Recently, Apple bought Wyze, a Seattle-based company that manufactures a camera with people detection on the edge. Google, on the other hand, launched Coral, a hardware and software platform to unlock AI without the cloud. Edge AI is truly a booming market. Deloitte estimates sales for edge AI chips and computers for 750 million in 2020 alone, rising to 1.5 billion by 2024.
Edge AI has many advantages over cloud-based AI in terms of privacy, security, and speed. Regarding privacy and security, Edge AI doesn’t keep your data in remote servers, where they can be hacked or stolen. But, also important, Edge AI comes in handy when you need real-time data processing and cannot afford to rely on cloud-based infrastructures and the slowness it implies.