Ubitium, a semiconductor startup constructing a “common processor” for machines that understand, assume and act in the actual world, has introduced $14 million in new seed funding.
The spherical was led by Ibex Traders, with participation from OpenOcean, present buyers together with KBC Focus Fund, Runa Capital and Inflection, and Florian Seibel, co-founder and co-CEO of European defence expertise firm Quantum Programs, investing personally.
It brings Ubitium’s whole seed funding to $18 million because the firm was based in 2024.
Ubitium’s thesis is straightforward: autonomous machines want three sorts of computing to work collectively to allow them to understand, assume and act in the actual world. The corporate is constructing one common processor to switch the stack of specialized chips typically stitched collectively to do this work at present.
The funding follows the return of Ubitium’s first silicon, taped out a yr after the corporate was based, which demonstrated that sign processing, AI inference and real-time management can all run on one processor structure.

A brand new processor for a brand new period of computing
Bodily AI is synthetic intelligence constructed into machines that do bodily work, akin to drones, robots, automobiles and industrial gear. Basic-purpose robotics alone may turn out to be a $370 billion market by 2040 below McKinsey’s base-case forecast.
An autonomous truck, a warehouse robotic and an inspection drone do very completely different jobs, and the identical three issues: understand, assume and act, constantly, typically concurrently, on battery energy, in actual time. The steadiness between them modifications because the machine operates.
Right this moment builders divide that work throughout a digital sign processor (DSP) for sensing, a neural processing unit (NPU) for AI and a CPU for management.
Connecting them provides knowledge transfers, coordination and engineering complexity, and bodily AI strains that association to breaking level. Nobody processor was designed for the calls for of the complicated bodily world.
In contrast to processors constructed from separate specialised logic, Ubitium has reimagined the processor as one array of programmable parts for sign processing, AI and real-time management. Sources are allotted at runtime throughout these duties, so the identical silicon adapts to the machine’s altering workload.
Hyun Shin Cho, CEO and co-founder of Ubitium, says: “Bodily AI machines deliver a complete new set of computing challenges, and the inference isn’t the toughest one.
“Sensing the actual world and controlling the machine in it are simply as laborious, and at present these nonetheless run on separate chips. We constructed one bodily AI processor that does all three, proved it in silicon in a yr, and this funding turns it right into a product.”
Ubitium CTO and co-founder Martin Vorbach has pioneered reconfigurable computing for greater than three many years. Now its second has arrived: bodily AI is a continually shifting workload, and {hardware} that reconfigures with it’s the pure match.
Ubitium holds 17 granted patents overlaying its core expertise, with extra pending, whereas its crew brings over 250 profession patents in computing, over 350 peer-reviewed publications and silicon expertise from Intel, Nvidia, Apple and Texas Devices.
With unified reminiscence and normal RISC-V instruments, the Ubitium structure offers a typical processor and toolchain instead of a number of specialised processors.
By lowering knowledge transfers between chips and simplifying software program growth, Ubitium goals to make clever machines extra economical to construct, with a computing basis producers can reuse throughout merchandise.
Silicon tech
Ubitium’s first chip, the UB108 with 256 processing parts, manufactured on Samsung’s 8nm course of, has demonstrated sign processing, AI inference and real-time management operating on the identical reconfigurable array.
Ubitium plans to tape out its first business processor, the UB410 with 4,096 processing parts, and ship analysis models in 2027.
Preliminary clients are constructing bodily AI’s sensing layer: high-resolution radar and audio programs that deliver demanding sign processing and AI onto one processor, in markets the place at present’s embedded chips already promote in quantity. Autonomous programs, together with drones, are the goal functions for the whole understand, assume and act loop.
Investing in Ubitium
Florian Seibel, co-founder and co-CEO of Quantum Programs, says: “An autonomous machine working in the actual world should hold working when communications are unavailable. It has to course of its environment, make selections and management itself on board, inside strict limits on energy and weight.
“Right this moment meaning integrating a stack of chips that weren’t designed to work collectively, and each certainly one of them prices weight, energy and integration time we don’t have. Ubitium is constructing the one processor that does that entire job. That’s why I invested.”
Jeffrey Peters, associate at Ibex Traders, says: “The tempo of AI is simply growing, far exceeding the power of {hardware} to maintain up and handle.
“Ubitium solves this with a common reconfigurable structure that may deploy the most recent advances on a single piece of silicon with a single set of growth instruments. Ubitium is actually {hardware} on the tempo of software program.”
Tom Henriksson, normal associate at OpenOcean, says: “AI is transferring from our screens into the bodily world, however tomorrow’s clever machines can’t run on a patchwork of repurposed chips.
“Bodily AI requires a ground-up {hardware} revolution. Ubitium’s crew has spent many years constructing precisely that structure, and with the expertise now in silicon they’ve created the foundational processor for this new period. We’re proud to again them.”
Rudi Severijns, funding director at KBC Focus Fund, says: “After we first invested, Ubitium’s common processor was a daring wager. First silicon in a single yr turned that wager into proof.
“We’re proud to maintain backing the crew, and we stay up for working with Ibex Traders and OpenOcean as Ubitium brings one processor for bodily AI to market.”
Jonatan Luther-Bergquist, normal associate, Inflection, says: “Since attending to know Hyun Shin and Martin at first of 2024 and being the primary test in, they’ve delivered astonishing outcomes.
“All the things they wager on relating to the market of bodily AI programs has come true, and so they’re uniquely positioned to ship compute and intelligence to an enormous a part of the market: from drones to humanoids. We’re thrilled to hitch new and present buyers on this spherical.”
