Fullpower‑AI wants your bed to understand you.

Here is a fact that sounds made up but is not: the man who invented the camera phone did it because his wife was in labor. It was June 1997, Philippe Kahn was in a delivery room, and he wanted to send a picture of his newborn daughter to a few hundred people at once. So he wired a digital camera to a cell phone, hacked together an uplink to a server at home, and did the thing that a billion people now do without thinking. That is the origin story of the selfie, and it is also, in a roundabout way, the origin story of Fullpower-AI.

Because the interesting thing about Kahn is not that he had one good idea. It is that he keeps having them, in the same shape, over and over. Take a sensor that already exists. Notice that it is trapped inside a device doing something boring. Free it. Turn it into data. The camera in the phone becomes a camera phone. The accelerometer in the watch becomes a step counter. And - this is the current chapter - the sensor under the mattress becomes a sleep lab.

Fullpower-AI, founded in 2005 with his wife and longtime co-founder Sonia Lee, is a company that sells the idea that sleep is measurable, and that you should not have to strap anything to your wrist to measure it. This is a genuinely contrarian bet in a market that has spent a decade convincing people to buy rings and bands and watches. Fullpower-AI's pitch is roughly: you already own a bed. What if the bed did the work?

A platform pretending to be a mattress

The way to understand Fullpower-AI is that it is not really a consumer company, even though its technology touches consumers constantly. It is a platform company that licenses its guts to other people's products. Its flagship is Sleeptracker-AI, a contactless biosensing system that sits in the bed, watches how you breathe and move and stir all night, and turns that into sleep stages, disturbances, and - increasingly - clinical signal. There is no gadget to charge, no strap to remember. You go to sleep; the math happens underneath you.

If you have ever slept on a high-end Tempur-Pedic smartbed and watched it quietly adjust when someone snored, you have used Fullpower-AI without knowing it. More than 300,000 of those beds run its technology. That is the tell of a good platform business: the customer loves the product and has no idea who built the part that makes it smart.

The lineage of that platform runs deep. Before beds, the same core sensor-fusion IP - branded MotionX - lived inside Nike's running app, Jawbone's UP band, and a joint venture that put smart modules into luxury Swiss watches from the likes of Alpina and Frederique Constant. The brand on the box kept changing. The thing underneath did not. This is the quiet genius and the quiet risk of Fullpower-AI's model: it owns the layer that everyone else rents, which is a wonderful place to be right up until a partner decides to build its own.

From "how long" to "how well"

The company's actual intellectual argument - the thing Kahn will talk your ear off about - is that the entire consumer sleep-tracking industry has been measuring the wrong thing. Everyone reports duration, because duration is easy. You slept seven hours and twelve minutes. Congratulations, that number tells you almost nothing. Fullpower-AI's whole 135-patent thesis is that quality is the metric that matters, that it is different for every person, and that you need real sensor data plus deep learning to get at it. Duration is the vanity metric. Quality is the one that would actually change your behavior, if you could see it.

Whether they have fully cracked that is a scientific question, and to their credit Fullpower-AI has behaved like a company that thinks it is one. It cites 70-plus research studies and 10-plus publications, holds ISO 27001 certification, and is HIPAA and GDPR compliant - the boring letters that are, in health technology, the actual product. A mattress company can only be trusted with medical-grade data if someone has done the unglamorous compliance work. Fullpower-AI did the unglamorous compliance work.

The Large Action Model in your bedroom

The newest piece is KOA, which Fullpower-AI describes as a "Large Action Model." The industry spent 2023 and 2024 building Large Language Models that generate words. KOA is meant to generate actions - autonomous AI agents that take a stream of biosensing data and do something with it, whether that is coaching, adjusting a bed, or flagging a change for a clinician. It is a subtle but important bet: that the value in health AI moves from models that talk to models that act. If they are right, the mattress is not the product. The mattress is a data-collection endpoint for something much larger.

The business is licensing, not gadgets

It helps to be clear-eyed about how Fullpower-AI actually makes money, because it is not the way most sleep-tech companies do. There is no flagship Fullpower device you buy at Best Buy. Instead the company sells a platform - firmware, cloud, models, and the patents that protect them - to manufacturers, researchers, and care providers, and gets paid through licensing and embedded technology. Reported annual revenue sits around $18.5 million, which is modest until you remember it is coming from a 63-person shop that mostly collects checks from other people's product lines. That is a high-margin place to stand, assuming the partners keep renewing.

The competitive map is crowded but the lanes are distinct. Wearables - Oura's ring, Apple's watch, the Fitbits of the world - ask you to put something on. Smart-bed rivals like Sleep Number's SleepIQ and the venture-heavy Eight Sleep compete on the mattress itself. Contactless sensing outfits like Withings, Google's Nest sleep sensing, and Emfit fight over the nightstand and the bed frame. Fullpower-AI's answer to all of them is the same: it does not want to sell you a device at all. It wants to be the software layer that the device makers license, which is a bet that the intelligence, not the hardware, is where the durable value lives.

The customer that became the investor

In August 2025, Tempur Sealy - already Fullpower-AI's biggest partner - did the thing that partners do when they cannot afford to lose you: it signed a 10-year global licensing deal and put in $25 million of Series C money, valuing the company at roughly $160 million. When your largest customer writes you an equity check and locks in a decade, you have stopped being a vendor and become infrastructure. It is also, if you are keeping score, a fairly elegant way for a bedding giant to make sure a competitor never gets access to the brains inside its beds.

What is a little wild about all of this is the scale relative to the size. Fullpower-AI is a company of roughly 63 people in Santa Cruz. It runs sleep intelligence in more than 60 countries, on AWS infrastructure, across a device base it says exceeds 1.5 million units, sitting on what it calls the world's largest sleep dataset. You do not need to be big to be foundational. You need to own something specific and defensible and genuinely useful, and then wait. Fullpower-AI owns sleep, and it has been waiting patiently, compounding data, for twenty years.

The honest uncertainty is what happens next. A dataset is a moat until a partner or a hyperscaler decides to dig around it. A platform is a great business until the brands you power decide they want the margin. But for now, the camera-phone guy has done the same trick a fourth time. He found a sensor doing something boring, freed it, and turned it into data. The sensor this time is your bed. You will probably never see the company's name. You will just sleep on it.