For posterity wanted to add this interesting post [1] by TSMC about their work with Oticon (one of the established hearing aid manufacturers). It's a few years old (2021), so instead of 28nm, more recent models probably use even 5nm.
"Compared to the previous generation of Oticon Opn S™ which is powered by TSMC’s 65nm low-power technology and contains 64 million transistors, the newest model features one of the most advanced chips with AI sound processing and analysis capabilities. The migration to TSMC’s 28nm opened the door for Oticon to utilize DNN, which is trained with 12 million real-life sound scenes. This new feature allows hearing aids to mimic how the brain would hear sounds and aims to respond the same way as the brain. As more functions are added, more compute performance and improved power efficiency will be required to support future generations of hearing solutions."
"Compared to the previous generation of Oticon Opn S™ which is powered by TSMC’s 65nm low-power technology and contains 64 million transistors, the newest model features one of the most advanced chips with AI sound processing and analysis capabilities. The migration to TSMC’s 28nm opened the door for Oticon to utilize DNN, which is trained with 12 million real-life sound scenes. This new feature allows hearing aids to mimic how the brain would hear sounds and aims to respond the same way as the brain. As more functions are added, more compute performance and improved power efficiency will be required to support future generations of hearing solutions."
[1] https://esg.tsmc.com/en-US/articles/106