AI at the Edge Reality Bolstered by Advances in Silicon and More

By | Managed Services News

Oct 14

Omdia research identifies multiple factors shaping AI at the edge.

For artificial intelligence (AI) at the edge to become a reality, there are some core requirements. Among the most vital are high-end processors capable of doing the job.

Emerging products are poised to make AI at the edge – which is key to application success across vertical markets – a reality in the near future, according to recent research from Omdia.

Connecting the Dots: AI at the Edge” examines the factors that will shape AI at the edge. The research also looks at short-term and long-term practical uses cases for AI at the edge across five vertical markets.

Omdia's Bill Morelli

Omdia’s Bill Morelli

“A lot of what we’ve seen with AI advancements over the past few years – even more when you look at IBM Watson – were driven by the cloud. These were real-world applications, but we were figuring out how to use AI effectively — and what the boundaries and the capabilities of the technology were,” Bill Morelli, vice president for enterprise at Omdia, told Channel Futures. “What we’ve grown to as the market has evolved is a point where we see the potential for AI to be used more effectively, if we move it out of cloud.”

All kinds of compute workloads have been pushed to the cloud over the last decade in the as-a-service model. What Omdia sees now is the need and potential to do more of this at the edge with better results.

Edge computing is part of a distributed, or decentralized, computing architecture where data processing happens closest to the device. The proliferation of IoT devices and the need for data collection and analysis is a key driver of AI. The increasing need for processing speed for IoT data-intensive applications is driving AI at the edge.

From Cloud to Edge

Overcoming network latency and driving speed is a must to meet the needs of IoT. That’s why Omdia’s research suggests a shift from cloud-based AI solutions to AI at the edge. 5G is also an enabler here.

Here are the five factors Omdia says will shape AI at the edge.

  1. New and emerging edge use cases. While smartphones may have been the tip of the spear for AI-at-the-edge devices, emerging use cases from enterprise and industrial will be what drive future growth.
  2. Diverse, cost-effective silicon solutions. AI has been a key driver for silicon growth in the high-end processor market for the past several years, but what will enable AI to proliferate at the edge will be a diverse portfolio of processors that can address a range of use cases. [Worldwide shipments of AI-optimized processors for edge systems was forecast at 340.1 million in 2019. By 2023, IDC research estimates the market will for unit shipments will reach 1.5 billion, for a CAGR of 64.9%.]
  3. New compute ecosystem from cloud to edge. Distributed computing technologies such as Kubernetes have made hybrid and distributed cloud infrastructures more manageable and operationally agile. The server market is already seeing the impact of this trend.
  4. CSPs use AI at the edge to optimize for AI at the edge. In order to offer compelling AI services at the edge, CSPs are also adopting AI at the edge to achieve low latency for new services, improve the customer experience and reduce costs.
  5. Broad range of industry verticals with specific requirements. Video surveillance is just one area where on-device AI is enabling a fundamental shift in the industry. Other industries, from fast food to oil and gas, are also seeing the benefits of deploying on-device AI.

Vertical Markets

The Omdia report looks at the short-term and long-term outlook for AI at the edge across vertical markets. The five vertical markets are manufacturing, health care, smart buildings, utilities and video surveillance.

Manufacturing

  • Short-term outlook (2020): With limited edge expertise and edge products new to market, applications (bar vision) will continue with threshold analytics rather than “true AI” machine learning.
  • Long-term outlook (2021): Edge compute and control devices from IT companies will shift strategic relationships from …

About the Author

>