According to a new report from French analyst house Yole Développement, AI acceleration will soon become an expected feature in smartphone application processors. However, the analysts also identified the technology’s impact on the smart home ecosystem as a fast-growing market.
The report, “Artificial Intelligence Computing for Consumer,” estimates the market for AI computing hardware for consumer applications will reach USD $15.6 billion by 2024. This represents a CAGR of 37.5% from today’s market, which was USD $2.3 billion in 2018.
AI accelerators, whether they are named neural processing units, neural engines or deep learning accelerators, aim to parallelise the type of calculations required by neural networks without the need for a power-hungry GPU (graphics processing unit). Incorporating this capability into consumer devices reduces their reliance on the cloud, improving latency and preserving data privacy. Today, this trend is happening faster than ever, especially in the imaging segment.
“In recent years, we are seeing the emergence of edge computing [for AI],” said Technology & Market Analyst, Computing & Software at Yole, in a statement. “This trend involves placing the calculation at system level. However, the constraints are strong: always-on [power] consumption and performance, in particular.”
To support this transition, consumer devices require powerful, dedicated hardware to run AI algorithms. This market is split between standalone AI accelerator chips and AI accelerator IP.
“For each type of computing hardware, the actors are different and specialised,” said John Lorenz, Technology & Market Analyst, Computing & Software at Yole, in a statement.
The market for AI accelerators as standalone co-processor chips in consumer devices is estimated to be worth USD $3.8 billion in 2024. AI accelerator IP will also be incorporated into SoCs such as application processors for smartphones. This market is projected to reach USD $11.8 billion by 2024.
The total market for AI accelerator hardware for consumer devices, worth a total of USD $2.3 billion in 2018, is today almost solely driven by AI image processing IP for SoCs. By 2024, standalone vision processor chips, and standalone and embedded audio processors, will each make up a significant share of the market. Of these other segments, standalone audio processors will grow the fastest, reaching 14% of the market by 2024.
Sally Ward-Foxton is a Correspondent for EETimes and EETimes Europe