For 30 years hearing aids have been very good at one thing: making everything louder. Matthew de Jonge's company built a hearing aid that decides, in real time, which sounds matter and turns the rest down.
The industry's own scientists told him not to try. Their reasoning was that a model big enough to separate speech from noise could never run on a battery that hides behind an ear.
"And they were kind of right in so far as if you limited yourself to the components that power today's hearing aids, it is completely impractical."
De Jonge helped build the handheld ultrasound machine at Butterfly Network, which is how he knew a small team could design its own silicon; one of his co-founders was an early designer on Google's Tensor Processing Unit, and the company raised $163 million at a $740 million valuation from Founders Fund, Thrive Capital and Valor Equity Partners.
The full interview is covered here so you can skip it. 27 minutes of audio, 15 minutes of reading.
Here are the 10 takeaways that matter.
👤 Guest: Matthew de Jonge, CEO of Fortell, the AI hearing aid company he co-founded, previously part of the team that built Butterfly Network's handheld ultrasound machine
🎙️ Host: Rebecca Bellan, Senior Reporter at TechCrunch, who tried the hearing aids at the company's Manhattan clinic before this interview
📰 Published: 16 September 2026 on YouTube (Equity, TechCrunch's startups podcast)
🔴 YouTube | ⏱️ 27 min | ✅ Time saved: 12 min
Key Takeaways
Five companies make 97% of the world's hearing aids, and Medicare covers none of them
De Jonge calls the non-coverage outrageous, on the argument that a hearing aid is medicine
Hearing aids solved loudness 30 years ago and never solved selectivity
Age-related hearing loss also damages the brain's own sound filtering, so louder makes it worse
Fortell's AI splits incoming sound into two tracks and amplifies only the important one
The filter works on what a sound is and where it comes from, using four microphones on the head
The company designed its own AI chip because no existing hearing-aid component could run the model
De Jonge says Apple's AirPods are clinically valid audiometers that can test hearing at home
The device costs $6,800 and comes bundled with five years of audiology care and free software upgrades
The constraint on growth is not manufacturing or clinics but finding audiologists
Remote fitting is blocked in some states by what he calls anti-telemedicine rules
1. Why He Built a Hearing Aid
Bellan opened the episode on AI wearables and then turned to what she called the unsexy part of the market, with Apple improving hearing support in AirPods and glasses companies including Meta and the startup Legato putting hearing aids into frames.
De Jonge's route into medical devices went through a hedge fund. "And studied mechanical engineering in college. Spent a long time wandering through the desert figuring out what I wanted to be when I grew up. I did a long stint at a hedge fund. Discovered there a real passion for artificial intelligence and eventually I decided that what I wanted to do was work on high stakes medical devices."
The training ground was Butterfly Network. "Got my education in how to do high-tech medical devices at a startup company called Butterfly Network. It's now a public company."
What that company gave him was nerve as much as skill: "We built a handheld ultrasound machine and that gave me the audacity and access to capital and talent and confidence that you need to take on something like building a better hearing aid."
Fortell is about six years old and launched about nine months ago, with its first audiology clinic on 61st Street and Park Avenue in Manhattan.
The personal reason is four grandparents. "This has affected my other co-founders in varying degrees, but all four of my grandparents lost their hearing and just drifted away."
The question he could not drop, about 20 years ago, was a comparison with eyesight: "Why do I have to beg my grandparents to put on their hearing aids, but no one has ever needed to ask me to put on my glasses?"
"When I put on my glasses, I can see again. I can see just as well as a healthy seeing person with my glasses on." The hearing aids did not do that outside quiet, simple rooms.
He watched four people he loved follow the same path, which he described as a "cascade of hearing loss and cognitive decline."
On whether the technology would work: "And we didn't know if it was going to work but it did work."
2. Louder Is Not Clearer
Bellan put the user complaint first: everything arrives at the same volume, so a plane overhead and the person next to you are indistinguishable, and a dinner table with five people talking becomes sensory overload.
De Jonge said the industry has been optimizing one variable since it went digital. "For 30 years, since hearing aids went from analog to digital, hearing aids have been really really good at making sounds louder."
The variable has a name. "A hearing aid's primary job is to give you audibility."
The physical loss it corrects is sensitivity, mostly at the top of the range: "Because your ears, when you have hearing loss, your ears are less sensitive to quiet sounds. Typically, you see more loss in high frequencies than you do in the mids and lows."
The failure is that the amplification is untargeted. "But the problem, as you described well, is that these hearing aids are almost indiscriminate in the amplification that they create."
The part that makes it worse than it sounds is that age-related hearing loss damages two things, not one. "But it's actually even worse for someone with a sensory neural hearing loss with an age related hearing loss because in addition to having a reduced sensitivity to quiet sounds, they also as a consequence of aging have an impaired signal processing facility."
A young listener does the filtering for free: "So as a young healthy hearing person, we could go to a noisy restaurant and mostly we could just kind of tune in to the sounds we care about." An older listener cannot, and then the hearing aid raises the volume of everything. "Their signal processing facilities are already impaired. Now this is like a perfect storm."
His summary of the 30-year gap: "And so what we've seen over the last 30 years is this fundamental mismatch between what the technology can do, which is make sounds louder" — digital signal processing is good at volume and bad at "differentiating important from unimportant."
3. Two Tracks of Sound
The company's answer is to put a model between the microphone and the amplifier, and to sort the sound before amplifying any of it.
"And so what we do as a company is we built this very very large powerful AI that in real time is trained to call balls and strikes on sounds in real time and separate incoming sound into two tracks independently. Important stuff and unimportant stuff."
Only one track gets the conventional treatment. "Make the important stuff louder. While attenuating the background noise, even if that background noise is other people talking."
What makes it different from older noise reduction is the basis of the decision: "So, we've trained this AI to understand sound at a semantic level and filter it on the basis of what it is, what it means to the user instead of just the frequencies that comprise it."
The claim he ends on is that the two problems are addressed together — the ear and the brain: "So, you can make sounds louder and clearer at the same time."
Bellan described testing it herself on a Manhattan street, where the noise was overwhelming until he pressed a button in the app and the person in front of her became audible — and the effect moved as she turned her head.
4. A Custom AI Chip
The physical constraints are the reason the industry considered this impossible, and the reason the company built its own silicon.
"These hearing aids are they're tiny. They have to hide entirely behind your ear and they have to last all day." Fortell's run more than 24 hours on a charge.
"But to pull off this magic trick requires an unbelievable amount of compute power and sensors."
Experienced scientists and engineers in the industry told him not to bother, and he conceded they were right about the existing parts: "And they were kind of right in so far as if you limited yourself to the components that power today's hearing aids, it is completely impractical."
The way out was to stop being limited to them. "So part of our approach was that we knew from our previous company that we could design our own custom chips."
The credential behind that: "And so we built this little AI processor. My co-founder, one of my co-founders, our CTO, was a early designer on the Google Tensor Processing chip."
The development order was deliberate: prove the science on unlimited hardware first. He described telling the signal-processing side to use a rack of servers if they needed one, establish that the important-versus-unimportant split works, and only then build a chip to compress that computation into an in-ear device.
The second novelty is how much the chip is fed. "And to your point, yes, like in addition to this unbelievable amount of compute relative to any other hearing aid that's come before, we also did some novel things about just how much data that AI chip is consuming from the world."
Multiple microphones add direction to identity. The inputs tell the system not only what a sound is but "where the sounds are coming from," inferred from subtle differences between the signals arriving at all four microphones on the wearer's head.
"So we do this filtering at the intersection of both the what and the where." In the noisiest rooms, he said, the user can effectively tell the model to keep only the people they are looking at.
5. The $163M Round
Bellan noted that hearing aids read as medical technology rather than venture-backed technology, and cited a Forbes report from late July: $163 million raised from investors including Founders Fund, Thrive Capital and Valor Equity Partners, at a $740 million valuation.
De Jonge confirmed the shape and not the decimals: "Yeah I don't there might be like you know a decimal point that's off but yeah that's by and large that's about 160 million about 700 million last round."
Bellan raised Antonio Gracias, who co-led the Series B, noting his history on Tesla's board and his role with Doge. De Jonge took only the Tesla half, saying he knew of Gracias from a book about the creation of Tesla and "his like unbelievable support for hard tech."
The pitch had two parts, and the first is market size. "And hearing loss is so prevalent. It affects an unbelievable number of people."
"The prevalence goes up as people get older and almost nobody is happy with their hearing aids."
There is no alternative product on the horizon. "There's no cure for hearing loss. There's no cure on the horizon." The best available answer is an audiologist and a hearing aid.
His framing of the investor's question, given the technical risk, was whether "is the pearl worth the dive?"
The second part of the pitch was not a pitch. "I think the other thing was more like we have to remember like these investors are people and they have families" — and because the condition is so common, the investors had watched the same decline at home. "Yeah like we that happened to my dad that happened to my grandpa that happened to my mom."
Asked whether the chip itself was part of the attraction, he pointed to track record rather than a second revenue line. "Well, I you know I think that we had this unusual credibility building chips as a small team from our previous experiences."
The Josh Kushner story is a coincidence of end markets. "So, Josh Kushner when we first met him, Josh was invested us at the most risky time when we were just a couple of guys and an idea, but he had he was intimately aware of what we had done at Butterfly." The ultrasound machine was built for emergency medicine physicians — "And it just so happens that his father-in-law is an emergency medicine doctor."
He named three elements that made the concept fundable: a potentially very large market, a motive he described as being "out of revenge" for what happened to their grandparents, and a track record of custom AI and custom chips in medical imaging.
6. Medicine, Not Wearables
Bellan asked where a hearing aid sits in the AI wearables conversation, and how the company is priced — as a wearable, a health services company or an AI company.
De Jonge answered with the mission rather than the category. "The reason we started this company is to help people with disabilities live life in the ways that they want as themselves for as long as possible."
"We really see ourselves as almost like we're making medicine. We're here to solve high stakes problems."
His complaint about where the industry's talent goes: "And when we looked out into the ecosystem of companies, we saw a lot of attention, a lot of smart engineers, a lot of capital being funneled towards consumer use cases" with limited benefit to the consumer at the end of it. His example of the thin case is asking a pair of glasses to describe a painting.
The recruiting problem he names is unglamorous work. "How many of the smartest engineers you know want to go work at Medtronic? And yet like those products are really important."
He did not rule out glasses or headphones as applications of the technology, and said licensing was possible, but that it is not something he would expect the company to pursue first.
"But the reason we exist is to help people in profound high stakes moments." "And so we're going to stay focused on curing disability" rather than entertainment, where he said large companies are already doing good work.
7. What Apple Got Right
Asked whether a large player had approached the company, de Jonge declined — "I probably shouldn't comment on that" — and then praised the competitor at length.
"I think that what Apple's done with AirPods is a profound improvement in access." For a couple of hundred dollars, he said, a buyer gets amplification broadly comparable to "a mid to high-end hearing aid" — though not, he added, to his own product.
The limit of that comparison is what AirPods are good at: audibility, making the world loud enough to hear.
The part he says is under-reported is the measurement, not the amplification. "And then the other thing that's not been reported on nearly as much, but AirPods are also clinically valid audiometers."
"That they're are capable of testing your hearing in a clinically rigorous way from home."
Why that matters is friction and privacy at scale: "The idea that now the millions and millions of AirPods out there can allow people with very low friction in the privacy of their home to test their hearing in a way that's clinically rigorous that's valid."
"That's awesome and we like just completely applaud the work that Apple's doing there."
8. Five Firms, 97% Share
Bellan asked about the structure of the industry and whether insurance covers any of it.
"Yeah, it's true that there are five major companies that make all the hearing aids." Between them, he said, they "make 97% of the products."
His analogy is the supermarket cereal aisle, where a wall of brands resolves into two or three manufacturers, and he said the same dynamic runs through hearing aids.
What he objects to is the combination: the products are expensive and, on his account, have not improved at the pace buyers expected.
On insurance the answer is short. "So, it's a little complicated, but the short answer is they're not covered by insurance. Medicare does not cover hearing aids at all. And I think that's like outrageous to be honest."
"Like the hearing aids are medicine and I think that they should be covered by insurance companies just like medicine."
9. $6,800 and Free Updates
Bellan raised the Wired piece on the company, whose framing was that customers paid $6,800 for the devices, and asked how a new entrant breaks into a market this concentrated.
The price is the price, and he did not soften it. The product is $6,800 — "That's a lot of money no matter who you are."
He positions it below the incumbents' top-specification products while including five years of audiology care, a three-year warranty and extensive diagnostics in the price.
The unusual part of the model is that improvements are free. He said the company ships software upgrades continuously and does not charge again each time the device gets better.
The example is a release the company calls AI 2.0, which he described as a major improvement to the core experience of the hearing aid, delivered as "a free over-the-air update" to existing patients.
He placed the product at the high end of a market that runs all the way down to over-the-counter devices at AirPod prices, but "less expensive than the top spec hearing aids" from the five incumbents.
10. Scaling on Audiologists
The company sells through its own clinics — Manhattan, Greenwich and Palm Beach at the time of the interview — which Bellan pointed out is not how venture-backed companies usually scale.
The bottleneck is not the factory. "So the limiting factor on our scaleup is not actually like how quickly we can manufacture the devices." Nor, he said, is it opening clinics.
It is finding and training audiologists he would be "happy to send our own parents to see." The company interviews a large number every week.
What he screens for is "the horsepower, the empathy, the intellectual curiosity" and the aptitude to learn the technology — because the AI adds a variable experienced audiologists have never had to set, such as how much signal-to-noise improvement to target for a given user, and how that choice interacts with the physical ear tip.
New clinics are planned for San Francisco, Palo Alto, Boston, Miami and Los Angeles, opening as fast as the company can staff them.
The second route is remote fitting, where state law allows it. "Other thing that we have started doing in the states where we can do it legally is we've started fitting people fully remotely and this is an amazing solution particularly for experienced hearing aid wearers."
The physical fit is derived from a photograph. "If you send us a picture of your hearing aids from that picture we can tell okay how long a wire do you need?" The rest is a recent audiogram plus a video appointment with an audiologist licensed in the patient's state.
What slows it down is regulation he describes as deliberate. State-by-state rules, which he called anti-telemedicine laws, were put in place by the incumbents "to reduce the likelihood of competition" — so each expansion requires a careful legal review.
Bonus Insights
Bellan raised a Bloomberg piece that called Meta's glasses "perve spectacles." De Jonge said that is about their cameras rather than their hearing, and both moved on from the surveillance question.
Bellan noted that Legato, a hearing-technology startup, came out of stealth recently with $12 million and AI hearing glasses of its own.
For people researching the product, de Jonge said he reads Reddit himself. "So Reddit has a very active community of people who wear hearing aids or interested in hearing aids."
De Jonge's bottom line is that the hearing aid industry spent 30 years perfecting amplification and never addressed selection, and that the way in is a model big enough to judge sound by meaning and direction — which required a custom chip, because nothing already inside a hearing aid could run it.
Products, Companies & Tools Mentioned
Fortell (His company: an AI hearing aid at $6,800, bundled with five years of audiology care, sold through its own clinics)
Butterfly Network (Where he and his co-founders built a handheld ultrasound machine, and the source of their credibility with investors)
Apple (AirPods, which he says deliver real amplification cheaply and are also clinically valid audiometers)
Google (Where his CTO was an early designer on the Tensor Processing chip)
Meta (Named by Bellan as one of the glasses makers adding hearing features, and the subject of the Bloomberg line)
Medtronic (His example of important medical products that struggle to attract the best engineers)
Reddit (Where he personally follows the hearing aid community and answers questions)
Watch the full episode:
If this was worth your time, send it to someone closer to the industry than you are.
Get the latest market chatter as it happens: