On August 24, 2026, Swiss hearing-care giant Sonova began shipping its next-generation Phonak hearing aid platform, called Eon, in the United States, with rollouts following in Germany, France, the UK and Australia in September. The headline feature is a new chip, HYPERSONIC, built to run a deep neural network inside a device small enough to sit behind an ear — continuously separating speech from background noise in real time rather than simply amplifying every sound around the wearer.
Solving the Problem Hearing Aids Have Never Solved
Understanding speech in a noisy room — a crowded restaurant, a train platform, a family dinner — has long been the single biggest complaint among hearing aid users, more so than volume or fit. Traditional devices rely on rule-based sound classification: detect “noise,” apply a preset filter. Eon’s Spheric Speech Clarity 3.0 system instead uses a trained neural network to isolate speech from surrounding noise as it happens, and, according to Sonova, users fitted with it were twice as likely to understand every word spoken to them from any direction compared with hearing aids lacking the feature. Josipa Kern, Sonova’s global head of Phonak, said the platform “sets a new benchmark for intelligent hearing. It brings together the capabilities that matter most in daily life: understanding speech, adapting automatically and staying connected with confidence.”
The Engineering Problem Was Power, Not Intelligence
Running a neural network is computationally expensive, and hearing aids have almost none of the battery or thermal headroom of a phone. Sonova’s chief R&D officer, Anders Rosengren, described HYPERSONIC as the fix: it draws 37% less power than the chip in the prior Phonak Sphere generation while still executing real-time AI inference, which let engineers shrink the new Audéo EON Sphere model by 25% in volume and 20% in weight compared with its predecessor, while stretching battery life to roughly 38 hours per charge. A companion feature, WindBlock 2.0, cuts wind noise by up to 50 decibels and improves the signal-to-noise ratio by as much as 30 decibels — aimed at wearers who previously avoided cycling, sailing or just walking on a blustery day with hearing aids in. The lineup includes the flagship Audéo EON Sphere, a step-down Audéo EON R, and a CROS EON R model for people with hearing loss in only one ear.
A Market Under Pressure
The stakes for getting this right are large and growing. The World Health Organization estimates hearing loss affects 430 million people globally today, a figure Sonova cites as likely to reach 700 million — one in ten people — by 2050, with unaddressed hearing loss costing an estimated $980 billion a year in health care, lost productivity and social isolation. Sonova, which operates in more than 100 countries with over 18,000 employees, reported CHF 3.6 billion in sales for its 2025/26 fiscal year. Its rival ecosystem includes Starkey, GN ReSound and Widex, all of which have pushed their own AI sound-processing chips over the past two years, meaning Eon’s real test isn’t a lab benchmark but whether it holds up against competitors racing toward the same speech-in-noise problem.
Real-World Testing, and Real-World Skepticism
Sonova has leaned on field testing to make its case beyond the clinical trial data: audiologist Dr. Amanda Klekowski von Koppenfels wore Eon during an Arctic expedition and reported that “hearing aids aren’t limiting, they are empowering … I could hear so clearly that, at times, other members of the expedition without hearing loss were asking me to repeat what had been said.” That kind of anecdote makes for a strong marketing moment, but audiologists outside the company caution that clinical trial gains in speech understanding — measured in controlled settings with specific noise profiles — don’t always translate one-to-one into everyday use, where background chatter, reverberation and a wearer’s own hearing loss profile vary enormously. Cost is also a real barrier: premium AI-driven hearing aids typically run several thousand dollars per pair and are fitted through licensed audiologists rather than sold over the counter, keeping them out of reach for many of the people the WHO statistics describe.
What Happens From Here
Sonova’s bet is that shrinking the chip while adding real-time AI will pull hesitant users — particularly younger adults with early-stage hearing loss who balk at bulky devices — into treatment sooner, since untreated hearing loss is itself linked to higher dementia and depression risk in longitudinal studies. The company is watching adoption data from the U.S. rollout closely, and competitors are expected to respond with their own neural-network chips within the next product cycle. The broader question is whether AI-driven hearing aids can do for hearing loss what continuous glucose monitors did for diabetes awareness: normalize a category of medical device that people previously associated with aging and stigma, by making the underlying technology feel less like a hearing aid and more like a pair of very small, very capable computers.
Photo: wal_172619 / PIXABAY via Pixabay
