Counterpoint: More than 31 million prescription hearing aids will be sold globally by 2030, nearly half of which will support end-side AICounterpoint：到 2030 年全球处方助听器销量将超 3100 万台，近半支持端侧 AI

Counterpoint's Global On-Device AI Wearables Forecast projects global prescription hearing aid sales to top 31 million units by 2030, a 4% CAGR from 2025 to 2030, with nearly 50% of hearing aids carrying on-device AI by then. Hearing aids keep improving in size, connectivity and battery life, but the core problem persists: users struggle to follow speech in noisy places. Traditional aids rely on digital signal processing — microphones capture sound, a processor analyzes frequency and volume, and preset algorithms decide what to amplify — which handles steady noise like air conditioners but fails in complex scenes such as a crowded restaurant full of overlapping conversations. AI takes a different approach: instead of fixed rules, deep neural networks trained on millions of voice and noise samples identify the speaker in real time and suppress unrelated noise — separating voices, highlighting the target speaker, improving spatial awareness and continuously adapting. AI is especially good at distinguishing confusable sounds like s/f/t/sh/th, and numbers like fifteen vs fifty, so it improves real speech understanding in restaurants, meetings and busy streets rather than just raising the volume. In clinical speech-in-noise tests, normal-hearing listeners need roughly a 2 dB signal-to-noise ratio (SNR) to catch 50% of key words, rising to about 8 dB for mild-to-moderate hearing loss; per Phonak, fixed directional microphones add 6.4 dB over omnidirectional ones, and AI noise reduction raises the gain to 9 dB by default and 10.2 dB at maximum. Machine learning also makes aids more automatic and personalized, cutting listening effort and fatigue — but always-on neural networks cost battery and space: Phonak's Audéo Infinio Ultra Sphere lasts 56 hours in normal use yet only about 10 hours with AI constantly on. Latency is the hard constraint: processing must finish within about 10 milliseconds, so AI has to run locally on the aid rather than in the phone or cloud — the core value of on-device AI. The race for next-generation hearing aids is to run AI silently in the background, delivering clearer voices, lower listening effort and more personalization without draining the battery or growing the device.