The global artificial intelligence (AI) software market is estimated to be worth $62 billion in 2022 according to Gartner.1 AI use cases extend from automation at factories, inventory management and even customer experience. Yet despite the impressive market size, innovation has not been fully exploited because many objects using AI, such as drones, need to communicate with a remote server, increasing latency, security issues and other risks. AI edge technology aims to bridge that divide by bringing computing to the edge.
AI computing requires large amounts of compute power and storage, making it impractical to place in the end device itself using traditional server components. Recently, researchers have created AI microchips, less than 4.5 millimeters across and weighing less than two ounces, that combine the central processing unit and random access memory.
Advancements in edge computing and microchips have helped to push AI to the device edge, opening up new opportunities for businesses to innovate in ways previously unthinkable. The AI edge not only leverages microchip technology to process data and even run applications, but it also helps reduce latency. There's also less of a need to transmit data back and forth from a central data center to process it and make the decisions.
Software can make the edge smarter
While advances in hardware are necessary to bring AI to the edge, the technology is dependent on the software which allows the emulation of an intelligent system. Emergen Research estimated the global AI edge software market size reached $585.1 million in 2020 and will grow at a compound annual growth rate of 20.1% through to 2028.
The evolution and growth of the edge and microchips are in response to diverse use cases that are, in part, enabled by other emerging and maturing technologies, such as 5G connectivity and the Internet of Things (IoT). For example:
- Fully autonomous vehicles need access to data in near-real time to guide decisions that maintain traffic flow and ensure the safety of drivers, cyclists and pedestrians.
- The edge and 5G connectivity also help enable the management and piloting of emergency and other drones.
- Smart cities are also driving the need for advanced microchip technology at the AI edge, as are manufacturing facilities as they adopt Industry 4.0 technologies and best practices to optimize production.
Hardware advancements
With simple edge use cases demanding more intelligence and local processing capabilities, software is key to enabling new hardware. For moving computation to the edge to be successful, it must address the performance and energy efficiency of AI edge capabilities if they're to be made widely accessible and cost-effective. One technology that's well suited for the task is an application-specific integrated circuit (ASIC) chip, which addresses the need for higher levels of processing power.
The benefit of using an ASIC is that it's a customized device, so the design and packaging are optimized for each use case, which makes it ideal for unique requirements (including size constraints). ASICs offer strong power efficiency for high-performance applications, a valuable feature for more complex edge computing. The flexibility of ASICs allow for the use of multiple voltages and multiple thresholds to match the performance of critical regions to their timing constraints, and therefore minimize power consumption.
An alternative to ASICs are Field-Programmable Gate Arrays (FPGAs), which are well-suited for accelerating domain-specific compute-intensive tasks because their hardware is also customizable, including high-performance and energy-efficient deep learning algorithms. Research from Arizona State University found that FPGAs offer advantages for edge computing, as they can deliver predictable performance and their hardware architecture can be adapted to provide consistent throughput. Meanwhile, Princeton researchers have also presented their chip design, which conducts computation and stores data in the same place through in-memory computing, reducing the energy and time used to exchange information with dedicated memory.
In-memory computing combined with different types of memory is drawing substantial research interest when it comes to solving AI challenges, including the development of neural networks that mimic the human brain while also being mindful of power consumption.
Microchip technology advances enable autonomy
Global Market Insights predicted the global market value for AI chipsets is expected to grow from $8 billion in 2019 to more than $70 billion by 2026, driven by a surge in IoT devices with machine-learning capabilities and the development of smart cities. Companies that are deep into the development of microchip technology are predicting that AI applications will continue to be integrated into the fabric of business success.
Not only will IoT continue to drive many opportunities, but it will be more autonomous as the AI edge becomes increasingly populated by more intelligent industrial machines, drones and robots, while services for the public become more intuitive. The edge will benefit from advances in centralized high-performance computing as microchip technology trickles down over time and AI exerts greater influence on chip design and new players enter the market.
Businesses can expect the AI edge to mature and evolve rapidly over time so that more can be done more quickly with less dependence on a central data center. As the ecosystem of software and hardware grows at the edge, so will the number of building blocks to create new opportunities for innovation.
Most of all, the AI edge not only enables new services but also allows businesses to understand how existing services are functioning and how they can be improved. By embedding more hardware-based intelligence in the home, office, manufacturing facilities and entire communities, individual lessons learned can be applied more broadly to benefit the broader population.
Learn more about how AI edge computing data analytics can help industries such as manufacturing.
The author of this content is a paid contributor for Verizon.
1 Gartner, Gartner Forecasts Worldwide Artificial Intelligence Software Market to Reach $62 Billion in 2022, Meghan Rimol, November 2021.