Ubitium GmbH, a German semiconductor startup developing a universal processor for machines that perceive, think and act in the real world, today announced it has raised $14 million in additional seed funding led by Ibex Investors.

OpenOcean participated in the round alongside existing investors including KBC Focus Fund, Runa Capital and Inflection. Florian Seibel, co-founder and co-CEO of European defense technology company Quantum Systems, also joined the round. The new funding brings the total seed investment to $18 million since the company was founded in 2024.

Ubitium builds custom silicon for autonomous machines, otherwise known as physical AI, that do physical work, such as drones, robots, industrial equipment and self-driving vehicles. The company is building a single universal processor to replace today’s more specialized chips.

Autonomous AI machines rely heavily on three different things: signal processing for sensing, artificial intelligence inference and orchestration. These three foundational principles are generally stitched together across circuitry or silicon with wiring, allowing a machine to perceive, think, and act in the real world.

Today’s developers divide that work across a signal processor, a neural processing unit (or a graphics processing unit) and a central processing unit for control. The interconnection generates additional latency, data transfer time, coordination and engineering complexity.

Ubitium built a single processor using RISC-V technology and separate specialized logic as one array of programmable elements for signal processing, AI and real-time control.

“Physical AI machines bring a whole new set of computing challenges, and the inference isn’t the hardest one,” said co-founder and Chief Executive Hyun Shin Cho. “Sensing the real world and controlling the machine in it are just as hard, and today those still run on separate chips. We built one physical AI processor that does all three, proved it in silicon in a year, and this funding turns it into a product.”

The company built on RISC-V because it is a free, open-source architecture based on the reduced instruction set principle. It is open and royalty-free, unlike proprietary frameworks such as those from Intel Corp. and Arm Ltd. or, more importantly, CUDA, the standard that underpins Nvidia Corp.’s chipset.

Nvidia currently dominates the physical AI market by a wide margin, acting as the gravitational center by providing the hardware and software stacks. Estimates for the global market itself vary in scale, but it is projected to surpass $430 billion by 2030 and approach $1.6 trillion by 2040, according to a ResearchAndMarkets.com research report.

Ubitium’s first chip, the UB108 with 256 processing elements, manufactured on Samsung Electronics Co. Ltd.’s eighth-nanometer process, proved that the three processes could work together on a single silicon die. The company plans to roll out its first commercial processor, the UB410 with 4,096 processing elements, and deliver units for evaluation in 2027.

Image: Ubitium