The Boston company that decided a camera lens should be printed like a microchip - and then printed 300 million of them.
Above: Polar ID, the polarization meta-optic that fits under a phone screen. It is, technically, a lens. It is also, technically, a semiconductor. The optics industry is still deciding how to feel about that.
Somewhere on a 300mm wafer in a semiconductor fab, a camera lens is being printed - not ground, not polished, not assembled from a stack of curved glass, but etched into silicon at the scale of a few hundred nanometers. The features are smaller than the wavelength of the light they bend. To the naked eye the finished part looks like a flat gray chip. It is one of the most consequential optical components made this decade, and almost nobody who relies on it knows it exists.
That chip is a Metalenz meta-optic. The company that makes it is small - around 41 people - and it does not sell phones, sensors, or gadgets. It sells the idea that the lens, an object essentially unchanged in concept since the 1600s, can be reinvented as a printed surface. The pitch sounds academic until you learn the count: more than 300 million of these have already shipped inside everyday devices. Most arrived without a press release loud enough for anyone to notice.
This is the strange position Metalenz occupies: a deep-tech firm whose product is invisible by design, sitting one layer beneath the brands you actually recognize. Understanding why that matters means starting with the problem - the one that has annoyed optical engineers for four hundred years.
Here is the inconvenient truth that lens makers have lived with forever: a curved piece of glass bends light because of its shape, and shape takes up room. To correct the blurs and color fringes that a single lens introduces, you stack more lenses. The camera module in a modern phone is a tiny tower of them, each one fighting the errors of the last. It works. It is also bulky, expensive to align, and stubbornly resistant to getting any thinner.
For decades the optics industry treated this as a law of nature rather than a problem to solve. You wanted a smaller camera? You accepted a worse one. You wanted secure 3D face sensing? You accepted a bulky module with a projector that scatters thousands of infrared dots. The tradeoff was assumed, not questioned - which is usually the most expensive kind of assumption.
The opening for change came from physics, not product design. In the lab of Harvard professor Federico Capasso, researchers had been building metasurfaces: flat arrays of nanoscale pillars that steer light not by curvature but by the precise geometry of each tiny structure. Tune the pillars and you tune the light. In principle, a single flat surface could do the work of an entire lens stack. In principle.
In 2016, Robert Devlin - then finishing a PhD in the Capasso lab - made a wager that academics rarely make: that this lab curiosity could become a high-volume manufactured product. He co-founded Metalenz with Capasso and secured an exclusive worldwide license to the foundational metasurface IP developed at Harvard. More than 20 patents came with it.
The bet was not really about whether metasurfaces worked - the papers had settled that. The bet was about manufacturing. A clever physics demo and a part you can ship 300 million of are separated by a chasm that swallows most deep-tech startups whole. Devlin's insight was that because a metasurface is patterned like a chip, it could be made in the semiconductor fabs that already exist - the same 300mm lines that print processors. No new factories. No exotic supply chain. Just optics, finally speaking the language of silicon.
Co-Founder & CEO. Carried the metasurface out of the Capasso lab and bet his career that it could be manufactured at the scale of microchips, not microscopes.
Co-Founder & scientific anchor. The Harvard physicist whose lab produced the foundational meta-optics research - and, decades earlier, co-invented the quantum cascade laser.
Devlin and Capasso launch Metalenz with an exclusive license to Harvard's metasurface patent portfolio.
Demonstration wafers prove the optics can be patterned with standard semiconductor processes.
Backed by 3M, Intel Capital, Applied Ventures, M Ventures and TDK Ventures.
With STMicroelectronics, meta-optics ship in consumer Time-of-Flight modules - a world first. $30M Series B follows, led by Neotribe.
Samsung's ISOCELL Vizion near-infrared sensor is confirmed as the light engine for Polar ID.
UMC manufactures the meta-optic on a 40nm process with wafer-on-wafer bonding, qualified for volume.
A single image now yields lighting-independent 3D facial data, on-device.
The clearest expression of what Metalenz can do is called Polar ID. Most secure face unlock today - the kind banks trust for payments - relies on structured light: a projector throwing thousands of invisible dots to map the geometry of your face. It works, but it is bulky and pricey, which is why only premium phones bother.
Polar ID does it differently. Light reflecting off a face carries a polarization signature - a property the human eye cannot perceive but a polarization-sensitive meta-optic can read. Capture that signature in a single image, run it through machine-learning models, and you get a face print rich enough to tell a real person from a photo or a mask. No dot projector. No tower of lenses. A flat chip and a sensor.
The consequences are practical. The module is smaller, cheaper, and can sit under a phone's display rather than punching a notch through it. Its newest sibling, Polar 3D, pushes further - reconstructing lighting-independent facial shape and surface detail from one frame, the raw material for 3D-ready digital identity on ordinary hardware. And Polar ID is only the headline. The same metasurface approach already powers proximity and depth sensing inside Time-of-Flight modules, the unglamorous components that tell your phone when it's near your ear.
Deep-tech is crowded with elegant demos that never escape the lab. Metalenz's defense against that fate is a shipping count. Through its partnership with STMicroelectronics alone, more than 140 million meta-optics have shipped inside FlightSense Time-of-Flight modules since 2022. Across all of Gen 1, the figure passes 300 million. For a flat optic that the industry treated as theoretical a decade ago, that is the difference between a paper and a product line.
The partner list is its own kind of proof. STMicroelectronics manufactures the optics in its 300mm fabs and recently expanded the licensing deal. Samsung lends its ISOCELL near-infrared sensor as Polar ID's light engine. UMC, one of the world's largest foundries, brought Polar ID to mass production in late 2025 using a 40nm process and wafer-on-wafer bonding. When three semiconductor heavyweights independently bet their fab time on your optics, the technology has stopped being speculative.
Ask Metalenz what it is building toward and the answer is not "better phone cameras." It is the broader claim that optical sensing has been held back by the cost and size of conventional optics - and that removing those limits lets sensing spread to places it never reached. Smaller, cheaper, semiconductor-made meta-optics could put capable cameras and depth sensors into devices that could never justify a lens stack: cheap IoT nodes, medical instruments, robots, automotive lidar, AR glasses.
The business model is built for that reach. Metalenz does not chase consumers directly. It licenses its designs and IP to the foundries and sensor makers who already serve every major OEM, then lets manufacturing scale do the spreading. It is a quiet position - the layer beneath the brand - but it is the layer that decides what the brand can build.
By combining our metasurface innovation with UMC's manufacturing scale, Polar ID is ready to bring secure, affordable face authentication to billions of devices.
Return to that wafer in the fab. The flat gray part being printed there is, depending on who you ask, a lens that broke the rules or a semiconductor that learned a new trick. Both are right. What it represents is a shift that tends to happen exactly once per category: the moment an analog craft - grinding glass - gets absorbed into the relentless, halving-every-couple-years economics of silicon.
If that shift holds, the consequences run past face unlock. Optics that scale like chips means sensing gets cheaper on the same curve that made computing ubiquitous. The camera stops being a premium feature and becomes a default ingredient. Metalenz will not own most of those devices - it will be the invisible layer underneath, the part nobody photographs.
Which brings us back to where we started: a chip being printed on a wafer, smaller than the light it bends, shipping by the hundred million into a world that has no idea it changed. Metalenz built it that way on purpose. The best infrastructure disappears. This one is just getting started.
Figures and milestones drawn from Metalenz and partner press releases and public reporting. Shipment counts are company-stated and approximate. Where a detail could not be verified, it has been left out.
No official channel is linked from Metalenz's site, so these point to YouTube searches for Metalenz product demos and founder interviews.