Despite Covid-19 Slowdown, Edge AI Chipset Market Estimated to Close in at US$30 Billion by 2026

2020 was a challenging year for edge Artificial Intelligence (AI) vendors. Both market demand and deployment has slowed due to Covid-19 lockdowns and safety measures. Compared to the cloud AI chipset market that experienced 68% year-on-year growth in 2020, the edge AI chipset market only managed to grow by 1%. Nonetheless, the edge AI market is expected to bounce back stronger. Global tech market advisory firm ABI Research forecasts the edge AI chipset market will grow to US$28 billion in 2026, with a CAGR of 28.4% between 2021 and 2026.

“The demand for edge AI is not going away anytime soon. Edge AI devices can process raw data locally, reducing the reliance on constant cloud connectivity. Consumers appreciate the enhanced user experience brought by low latency and data privacy. At the same time, more and more enterprises are looking for ways to make sense of valuable asset data. They recognize the importance of edge AI in key applications, such as predictive maintenance, defect inspection, and surveillance,” said Lian Jye Su, Principal Analyst at ABI Research. “Anticipating the growing needs of AI processing at the edge, even public cloud vendors like AWS, Microsoft, and Google are introducing hardware and software solutions and forming industrial alliances and partnerships that target edge AI development and deployment.”

The post-Covid-19 recovery can also be seen in recent revenue growth and funding activities of edge AI chipset vendors. Although the automotive market suffered some setback in 2020, Intel’s Mobileye reported total revenue of US$967 million, a historical high for the Advanced Driver-Assistance Systems (ADAS) vendor. Horizon Robotics and ECARX, two automotive-focused Chinese edge AI chipset startups, have raised US$750 million and US$200 million respectively in 2021, indicating expectations for strong future performance.

Another key trend in edge AI is Tiny Machine Learning or TinyML. The ability to embed a small machine learning model in ultra-low-power devices have opened up new possibilities, enabling smart connected sensors and IoT devices to make decisions and take action based on soundwaves, temperature, pressure, vibration, and other time-series data sources. Traditional microcontroller (MCU) vendors like NXP, ST Microelectronics, and Renesas are partnering with the AI software and service provider ecosystem, to assist edge AI developers that do not have embedded system design expertise to deploy TinyML solutions. Other vendors are introducing ultra-low-power chipsets or proprietary machine learning models that are highly efficient in power consumption and memory footprint.

Not surprisingly, the ability to create developer-friendly software and platforms, and to create the best ecosystem with third-party vendors will be essential to accelerate the adoption of edge AI. “These vendors offer edge machine learning operation (MLOps) platforms that facilitate the entire development and deployment process, starting from data collection and processing to model training, optimization, and monitoring. Many introduce advanced machine learning model compression and quantization techniques that enable large deep learning models to shrink in size while maintaining their accuracy and performance. This frees machine learning models from resource-rich devices, as they can now be deployed across a wide range of devices,” Su concludes.

These findings are from ABI Research’s Artificial Intelligence and Machine Learning market data report. This report is part of the company’s Artificial Intelligence and Machine Learning research service, which includes research, data, and ABI Insights. Market Data spreadsheets are composed of deep data, market share analysis, and highly segmented, service-specific forecasts to provide detailed insight where opportunities lie.