The market for edge AI computing devices is poised to enter a period of expansion, according to sources at IC design houses.
As the edge AI sector continues to develop, figuring out how all types of ICs can be used to solve simple AI algorithms is a key goal for semiconductor makers.
Industry players said that customers and IC design houses are still actively exploring the possibilities of AI. Over the next few years, players expect that the standard configuration for many IC products will have to either include algorithms or reserve space for algorithms as AI applications popularize.
AI computing functions are gradually moving toward the edge to reduce the burden of AI computing and delay caused by redundant messaging. As such, the successful proliferation of AI applications requires the joint cooperation and development of all companies, particularly for technical capabilities on the edge, said industry players.
Major European IDM NXP Semiconductors mentioned the vast potential of AI applications but added that the real key is whether they are practical and can be used to solve users' problems. While there are many areas where AI functionality can be applied, the execution is still very complicated.
Industry players pointed out that requirements for different applications and customers vary drastically. Whether in-house algorithms or third-party algorithms are used, companies need to build chips that are able to fully support AI functionality.
Taiwan-based IC design houses have observed that AI capabilities are being extended to various chip products. In the past, most industry players believed AI functionality would only be incorporated in core processors and SoCs; however, more and more customers want AI algorithms to be included in peripheral chips for applications such as fingerprint recognition, display resolution upgrades, and audio quality improvements.
IC design players pointed out that Taiwan's AI chip development has grown out of generative AI, expanding from only a few players to a national movement.
