We’re designing a family of chips for the problems the industry keeps trying to solve with more power, more cores and more chips. They’re built on one shared foundation, from a processor small enough to fit inside a robot’s elbow to a rack in a data centre.
Smaller transistors and bigger chips carried the industry for decades. That’s running out. Data centres now need power plants of their own, and every new demand gets its own workaround chip. We think the next leap won’t come from shrinking the old design. It will come from changing what the chip is built to compute.
A faster chip for the same math.
New math, in silicon.
Each chip we’re designing starts from a real problem in AI, networking, robotics, energy, science or medicine. Underneath, they all share the same foundation.
Today’s AI chips burn enormous amounts of energy, give different answers to the same question and can’t tell when they’re wrong.
We’re designing AI chips that give the same answer twice, check their own work and run on a fraction of the power.
Networks spend billions of dollars in chips and huge amounts of electricity re-checking the same data at every stop.
We’re designing chips that check every message as it moves, at full speed, with nothing bolted on.
Robots still send every reading back to one central computer, which is too slow for a slip or a bump.
We’re designing reflex processors that answer within a millisecond and are small enough to fit inside a joint.
Most electric motors run on a fixed magnetic field, which wastes energy whenever the job changes.
We’re designing controllers that shape a motor’s magnetic field thousands of times a second.
Research that starts from new mathematics still has to run on chips designed for something else.
We’re designing processors built for the mathematics our research runs on.
Every new demand on today’s chips gets its own special-purpose design, built from scratch and incompatible with the rest.
We’re building all of our chips on one shared foundation, so a breakthrough in one carries over to the rest.
We’re not designing a chip. We’re designing a family that runs from a sensor the size of a fingernail to a data centre. Chips take years to build, so we’re planning in decades.
A chip the size of a fingernail that knows what it’s measuring and can prove it.
A processor inside a robot’s limb that reacts in a millisecond, on its own.
Controllers that run the motor, the battery and the machine around them as one system.
AI and networking chips that do more with less power, rack after rack.
Every world-class hub started somewhere. Silicon Valley was orchards once. It took a few people who decided not to wait for permission. Silicon is made from sand, and we’re building in the desert.
Why not us.
Why not here.
Why not now.
AI / Networks / Robots / Vehicles / Science & medicine / Anywhere computing meets the real world
The next foundation of computing is wide open. If you’d rather design new chips for new math than shrink the old ones again, we should talk.
People who decide what a chip should compute before anyone draws a transistor.
Introduce yourself →Designers who build it right and verification engineers who prove it.
Introduce yourself →Where the chip meets sensors, motors and the physical world.
Introduce yourself →Designers who control energy precisely, thousands of times a second.
Introduce yourself →Software people who make new hardware easy to use.
Introduce yourself →Partnerships, product and operations people who want to bring new chips to market.
Introduce yourself →Partners · Research · Early access
For investors, manufacturers and research partners, and anyone who needs a chip that doesn’t exist yet.