Lumotive.

Here is the thing about seeing in three dimensions: it is mostly a problem of aiming. LiDAR works by firing a laser at the world and timing how long the light takes to bounce back. Do that across a whole scene and you get a depth map - the thing a robot or a car uses to not run into a wall. The catch is the aiming. For twenty years, the standard way to point the laser was to physically move something: a spinning mirror, a gimbal, a little motor whirring away inside a puck bolted to a roof. It works. It also wears out, costs money, and rattles.

Lumotive's whole premise is that you should not have to move anything at all. The company, founded in 2018 and based in Redmond, Washington, makes a semiconductor called a Light Control Metasurface - LCM, because everything in this industry eventually becomes a three-letter acronym. It is a CMOS chip with thousands of tiny, tunable optical resonators etched onto its surface. Apply the right pattern of voltages and the surface bends an incoming laser in whatever direction you want. Change the voltages and the beam points somewhere else. No mirror. No motor. Microseconds, not milliseconds.

The metamaterials backstory

This is not a garage story. The underlying physics came out of metamaterials research - engineered surfaces that manipulate light in ways ordinary materials cannot - with roots at Duke University and at Intellectual Ventures, the invention firm that also happens to be where the Bill Gates money enters the picture. Lumotive was spun out to commercialize it, co-founded by William Colleran and CTO Gleb Akselrod. Gates himself has offered the kind of quote that founders frame: the technology, he said, holds "great promise for products that help mobilize and automate the world." That is a nice sentence. It is also, notably, an endorsement of a chip, not a car.

Which brings us to the most interesting decision Lumotive ever made.

The pivot: sell the chip, not the box

The obvious business - the one every LiDAR startup of the late 2010s chased - is to build a finished LiDAR unit and sell it to carmakers. Lumotive, for a while, looked like it might do exactly that. Instead it did something quieter and smarter: it decided to sell the chip inside, and let everyone else build the box.

This is a genuinely different bet. A finished-LiDAR company lives or dies on one enormous market - automotive - that keeps insisting it is two years from being ready. A chip company sells to everyone building LiDAR at once: industrial-automation firms, robotics companies, security and surveillance vendors, smart-infrastructure people counting cars at intersections. The market is smaller per customer but there are far more customers, and you are not betting the company on a single design win. CEO Dr. Sam Heidari frames it as a shift in where the value lives: chip-scale LiDAR, he says, "creates a paradigm shift whereby application-specific software drives value rather than hardware alone." Translated: the beam is now programmable, so the interesting engineering moves into software, and Lumotive sells the thing that makes it programmable.

What it actually sells

The flagship is the LM10, which Lumotive calls the world's first commercially available optical beamforming semiconductor - a claim that won it a CES 2024 Innovation Award. The specs are the kind of thing that make sensor engineers lean forward: up to 160 degrees of native field of view (past 180 with expansion optics), compatibility with both 905 and 940 nanometer lasers, support for both flavors of time-of-flight sensing, and, again, no moving parts. Around the chip, Lumotive sells reference designs and development kits - the M30, the NM120, the TX10 - so a customer can go from "interesting chip" to "shipping product" without reinventing the optics from scratch.

The result is a growing list of partners who put the chip in their own boxes. Hokuyo, a Japanese industrial-sensing firm, is building solid-state industrial LiDAR around it. Korea's Namuga launched its Stella-2 solid-state LiDAR powered by Lumotive. Saudi Arabia's E-Photonics built software-defined LiDAR for intelligent transport systems. Sony, ams OSRAM, Lattice Semiconductor and Seoul Robotics all sit somewhere in the ecosystem, supplying components or perception software. The company says it has around 30 customers in the pipeline.

The second market nobody expected

Now the part that turns a nice sensor company into a genuinely large opportunity. The same chip that scans a room can, in principle, route light between processors. AI data centers have a heat and latency problem: shoving ever more data between chips electrically is expensive in watts. Optical switching - moving photons directly between processors - sidesteps a lot of that. A programmable 2D beamformer that can redirect thousands of light paths in real time is, conveniently, exactly what Lumotive already makes. The company plans to launch its first purpose-built optical switches for data-center interconnect by the end of 2026.

The timing is its own small joke. When Lumotive raised $45 million in early 2025, the Wall Street Journal framed it as "bucking the AI trend" - a hardware company raising money for optics while everyone else chased models. And then the AI data centers turned out to need optics. Contrarian bets tend to look wrong right up until the market walks over to you.

The raise itself was oversubscribed to roughly $59 million, valued the company around $200 million, and drew a global cast: Gates Frontier and MetaVC Partners from the beginning, plus Swisscom Ventures, Hokuyo, TSVC, Oman's ITHCA and others. In 2025 the company grew headcount about 50 percent, opened offices in Oman and Taiwan, and got named the Global Semiconductor Alliance's Startup to Watch. None of that is a finished story. But for a company whose entire product is a rectangle of silicon that redirects a laser, it is a lot of the world deciding the rectangle matters.