Over the past 12 months, the number of engineers that see cost or power as obstacles to increasing onboard processing in their designs has halved from 2021 levels, according to a survey of engineers. The research, conducted by Semiconductor specialist XMOS, covers key trends related to the artificial intelligence of things (AIoT) market.
The lower barrier to entry also translates into more designs with ON-device processing. The majority of engineers surveyed are actively adopting increased on-device processing into their designs. Seventy percent of respondents report that most of their planned products would require “high” or “very high” onboard-processing capabilities.
The research report, Edge of Now, XMOS’s third annual survey on the AIoT, polled over 100 engineers who design consumer, battery-powered, low-power, industrial, handheld, and rugged products across a range of industries, including consumer electronics, smart homes, automotive, smart cities, health care, and telecommunications. When asked about what specific benefits the AIoT could enable, respondents cited a wide range of potential applications, such as sensor processing, user-interface improvements, remote monitoring, communications, data integration, and predictive maintenance.
The research indicates a “significant decrease in the perceived barrier of the on-device–processing capabilities required to enable the AIoT,” said XMOS.
AIoT: Is AI the next breakthrough for the IoT?
Here are some research highlights: The report finds that 24% of respondents see cost as a barrier to increasing the on-device processing in their designs, which is down from 48% in 2021 and from 64% in 2020. There is also increased optimism when it comes to power consumption: Only 23% of engineers see power as a barrier to increasing onboard processing, which is down from 53% last year and 65% in 2020.
XMOS noted that it wasn’t a “surprise to see engineers reporting that cost and power are also becoming less of a barrier to adopting AI functionality in their designs” due to the importance of higher processing power to enable on-device AI functionality. The survey reveals that only 16% of engineers report cost as a barrier to adoption of the AIoT and only 13% see power requirements as a barrier to introducing AI into their products.
“However, these design challenges are mitigated — not eliminated,” said Mark Lippett, CEO of XMOS, in a statement. “User experiences still have a long way to go: We need to empower engineers to keep smashing through barriers to adoption, which means providing them compelling capabilities running on versatile, low-power silicon at a price that makes the mass proliferation of AIoT-enabled devices a realistic ambition.”
In addition to cost and power, engineers still face several other barriers. In terms of on-device–processing barriers, obstacles include design complexity (15%, though down from 38% in 2021), development time (14%), and system component requirements (12%). Barriers to adopting AI include lack of data for training AI systems (17%), availability of models (15%), and data security and autonomy (14%, though down from 31% in 2021 and 45% in 2020).
The good news for the AIoT market is that 64% of respondents are working on projects with sufficient onboard-processing capabilities to support on-device AI, which are expected to reach the market in the next six months. In addition, 38% said their entire product range will have suitable onboard-processing power to support AIoT functionality in the next two to three years, and 26% report that 70% of their product range will have on-device AI capability.