Artificial intelligence (AI) is accelerating cloud-enabled transformation, but what if its true power lies at the network edge? Edge AI workloads, where processing happens on local devices, enable real-time decision-making, reduced latency, and enhanced privacy.
From smart factories predicting equipment failure to autonomous vehicles analyzing road hazards, this is the exciting world of Edge AI. Explore the opportunities and challenges of running AI workloads at the edge, and discover how it's shaping the future of IT apps.
Edge AI Market Development
Reduced Instruction Set Computing (RISC)-V processor architectures are starting to address Edge AI workloads, and this trend is set to continue throughout the decade.
According to the latest worldwide market study by ABI Research, while RISC-V's penetration into AI workloads is only just beginning, growth will be steady throughout the decade, pushing RISC-V chip shipments in Edge AI (excluding TinyML) to 129 million by 2030.
"The flexibility of the architecture to address specific workloads, as well as the scalability, increases its appeal," said Paul Schell, industry analyst at ABI Research.
RISC-V International has diligently promoted and nurtured the ecosystem, and the RISE project now seeks to develop the software side through collaboration between industry leaders including Google, MediaTek, and Intel.
Leading startups, like Axelera AI and Tenstorrent, show RISC-V's potential to address more demanding AI inferencing workloads, like computer vision in automotive and security applications.
Legacy semiconductor players also want to develop processors using the Instruction Set Architecture (ISA). Making it open source has enabled more vendors to compete and spurred its proliferation in China, a country looking to gain semiconductor self-sufficiency due to U.S. sanctions limiting access to market-leading AI accelerators.
A key driver of shipment numbers is Edge AI gateways, the bulk of which comprises systems connecting and performing inference on sensors in the home. Another driver is robotics, most of which will be in consumer products not needed for mission-critical uses.
Manufacturers will seek the value and flexibility offered by RISC-V processors to drive down prices of consumer gateways and other devices as they address an increasing number of AI workloads, like natural language processing and computer vision.
RISC-V processors are entering the mainstream, and the community eagerly awaits the addition of matrix extensions, a key component of AI workloads, to the open-source ISA.
Google's full support for Android on RISC-V processors and the participation of major players like Intel and NXP Semiconductors in RISC-V International's working groups point to a long-term commitment by the wider industry.
Outlook for Edge AI Applications Growth
According to the ABI analyst assessment, the ecosystem lacks the governance of other popular open-source projects like Android, which could lead to fragmentation and hinder adoption in embedded systems.
"Nonetheless, OEMs and hardware vendors should follow the architecture's progress given its increasing market share, which displaces some legacy architectures like those developed by Arm," Schell concludes.
That said, I believe the outlook for Edge AI is bright. With diverse industries demanding more applications, from predictive maintenance in manufacturing to personalized healthcare, the Edge Computing paradigm is primed for exponential growth.
The combination of powerful yet efficient AI chips, robust 5G wireless networks, and evolving security protocols will fuel this growth, unlocking the true potential of intelligent IoT devices and connected systems.