A warehouse in El Segundo. A sensor that sees in thousands of colors.
Step inside Matter Intelligence's headquarters and the first thing you notice is what's missing. There are no marketing posters of rockets. No retro NASA stickers. There's a sensor on a bench - matte black, smaller than a microwave - patiently staring at a piece of mineral the way a sommelier stares at a glass of wine. It is, in a literal sense, looking for the molecules. The room is quiet, except for a fan and the very specific hum that hardware engineers make when they're about to disagree about tolerances.
This is the working theory of Matter Intelligence in one frame: hardware first, hype later. The 33-person team isn't building a chatbot. They're building the optical front-end of a planet-scale data system - a satellite-bound sensor that captures light across thousands of spectral bands, then hands the readout to a foundation model that knows, on the basis of physics rather than vibes, what it's looking at.
Most cameras give you red, green, blue. The human eye is happy with that. Machine vision has built an entire industry pretending three channels of light are enough to understand the physical world. Matter's bet is that they aren't, and that the next leap in AI won't come from another billion lines of internet text. It will come from light most cameras throw away.
From JPL labs to orbital startups.
Matter Intelligence was co-founded by three people whose résumés read like an aerospace history syllabus. Vishnu Sridhar, the CEO, led the team that built SuperCam - the laser-firing spectrometer head riding around Mars on the Perseverance rover. He also served as a flight director on the Opportunity rover and helped develop REASON, the ice-penetrating radar bound for Europa Clipper. Co-founder Thomas Chrien spent years at JPL building spectral instruments, and Nathan Stein came from Caltech's planetary science orbit. They have, collectively, sent gear to two planets and one moon.
The leap from JPL to startup is rarely subtle. Most government-trained instrument scientists land at large defense primes. Matter went a different direction, raising $12 million in seed funding from Lowercarbon Capital, Toyota Ventures, Pear, Mark Cuban, and E2MC - a coalition that quietly signals the company's three audiences: climate, mobility, and space.
A small team, an oversized claim.
Figures per Matter Intelligence public statements, Oct 2024 - present.
Three products, one stack.
Ultraspectral Sensor
The flagship hardware. Captures thousands of bands from ultraviolet through thermal infrared - the spectral equivalent of swapping a flip phone for a spectrometer.
Earth-1 Satellite
Matter's first orbital platform, designed for sub-meter hyperspectral and thermal imaging. The pitch: more usable data per pass than any commercial sensor flying today.
Large World Model
A foundation model trained on spectral physics, not Reddit. Takes the sensor's raw light and returns answers about what something is made of, not just what it looks like.
Aerospace + Robotics
The same sensor stack is being pitched for drones, aircraft, and robots - anywhere a machine needs to handle, inspect, or identify a physical object.
Pipelines. Pastures. Pit mines. Plumes.
Hyperspectral imaging at sub-meter resolution is genuinely useful, but only if you have a problem that needs molecular truth from above. A traditional satellite image can show you that a field is green. A hyperspectral image can tell you whether that green is corn, soybean, or weeds, and whether the plants are nitrogen-stressed. The same physics applies to methane leaks, illegal tailings ponds, counterfeit minerals, and hidden infrastructure.
Find the leak
Pipeline operators want methane plumes flagged before regulators do. Matter's thermal + spectral combo is built for it.
Read the crop
Nutrient stress, water stress, disease - all of it leaves a spectral fingerprint before it's visible to a drone camera.
See the unseen
Concealed assets, surface composition changes, infrastructure that wasn't there last week. Hyperspectral is the genre.
Map the deposit
Mineral exploration from orbit isn't new - doing it at sub-meter resolution with thousands of bands is.
Why data, not models, is the moat now.
The dominant narrative in AI is that models are the moat. Sridhar is making the opposite bet: that the bottleneck for physical-world AI isn't architecture, it's the absence of dense, spectrally-rich training data of the actual world. Most geospatial AI models today are trained on the equivalent of three-channel security camera footage. Matter is trying to feed them MRI scans.
The choice has consequences. Building sensors is expensive. Launching satellites is expensive. Calibrating spectral data is, by all accounts, a thankless and humbling exercise. But if Matter is right, the AI that runs farms, factories, and forests in 2030 will have learned to see from a sensor that didn't exist in 2023.
Three people who've already shipped to Mars.
Vishnu Sridhar
Led SuperCam on Perseverance. Flight director on Opportunity. MBA, Harvard Business School. The translator between physics and pitch decks.
Thomas Chrien
Veteran JPL instrument engineer. Has spent a career calibrating sensors that have to work the first time, in space.
Nathan Stein
Former Caltech planetary scientist. Brings the science side - if the sensor sees something strange, he's the one who knows whether it's a mineral or a glitch.
From stealth to sub-meter.
Quietly founded
Three JPL and Caltech alumni leave their day jobs to start working on the sensor.
Emerges from stealth
Raises $12M seed led by Lowercarbon Capital, with Toyota Ventures, Pear, Mark Cuban, and E2MC. Announces the Earth-1 program.
Aerial demo
Stated plan: demonstrate the spaceflight-ready ultraspectral sensor on an aerial platform before year's end.
Earth-1 in orbit
The first satellite launch and the start of a planned constellation. Launch date - per the company - to be announced.
The seed table.
Lowercarbon Capital
Lead. Climate-focused fund that backs the unglamorous infrastructure of measurement.
Toyota Ventures
Strategic interest in sensing for physical systems and autonomy.
Pear VC
Early-stage generalist with a track record in deep tech.
Mark Cuban
The angel slot. Cuban tends to back companies he can explain in one sentence.
E2MC Ventures
Space-focused fund. The category specialist on the cap table.
If you'd rather hear it from them.
Back to the bench.
Return to the warehouse in El Segundo. The sensor is still on the bench. The mineral is still there. But now imagine the same sensor, miniaturized, sitting in low Earth orbit, looking down at a wheat field outside Lubbock, or a methane plume drifting over the Permian, or a stretch of coastline where someone is - quietly - dumping something they shouldn't. The signal is the same. The light is the same. What changes is who gets to see it.
That's the small idea hiding inside Matter Intelligence's big one. Every machine that touches the physical world - from a robot in a warehouse to a tractor in Kansas to a fighter jet over a contested coast - is making decisions on too little information. Matter wants to ship the missing channels. And then, somewhere between bench and orbit, hand them to an AI model that's been waiting for better data its whole short life.
It's an idea that took 33 people, three founders with three planets of credit between them, and exactly one very patient sensor on a bench. The light, as it turns out, was always there. We just hadn't built the right eyes to see it.