Ubitium Raises $14M to Build a 'Universal Processor' for Robots, Drones and Physical AI

The AI chips that matter for the next decade may not live in data centers. German semiconductor startup Ubitium has raised $14 million in seed funding to commercialize what it calls a universal processor for physical AI — a single chip that senses the world, runs AI models and controls movement in real time, three jobs that today are split across three different kinds of silicon.

The round was led by Ibex Investors, with participation from OpenOcean and existing backers KBC Focus Fund, Runa Capital and Inflection. Florian Seibel, co-founder and co-CEO of European defense technology company Quantum Systems, also invested personally.

The raise brings Ubitium's total seed funding to $18 million since it was founded in 2024 by chief executive Hyun Shin Cho and chief technology officer Martin Vorbach.

Autonomous machines — robots, drones, vehicles, industrial equipment — must continuously process sensor data, make decisions with AI models and drive motors or actuators, all at once.

Traditionally those workloads run on separate chips: digital signal processors for sensor data, neural processing units for AI inference and CPUs for control. Every handoff between them costs time, energy and hardware complexity.

Ubitium's answer is a reconfigurable processor architecture whose programmable processing elements dynamically allocate compute between sensing, AI inference and real-time control depending on the machine's workload.

It uses unified memory and standard RISC-V development tools, so manufacturers can target it with familiar software stacks instead of learning a proprietary toolchain.

Unlike many chip startups raising on architecture diagrams, Ubitium says it already has working silicon.

Its UB108 prototype — 256 processing elements manufactured on Samsung's 8-nanometer process — has demonstrated signal processing, AI inference and real-time control running on the same reconfigurable array. The company taped it out about a year after its founding.

The new funding will advance its first commercial processor, the UB410, which packs 4,096 processing elements.

Ubitium plans to finalize the chip design and ship evaluation units in 2027, with initial applications focused on high-resolution radar and audio systems, and longer-term targets in autonomous systems including drones.

The bet is that as AI moves from screens into machines, the market for compute shifts with it.

Foundation-model training rewards maximum floating-point throughput, but robots reward latency, sensor integration, energy efficiency and predictable control — qualities general-purpose GPUs were never designed around.

Ubitium holds 17 granted patents, and says its team carries more than 250 career patents and 350 peer-reviewed publications, with silicon experience from Intel, Nvidia, Apple and Texas Instruments.

Competition is not standing still: venture money is pouring into every layer of the physical AI stack, from world models trained on gameplay footage to dedicated robot brains.

Ubitium's differentiator is consolidation — fewer chips, simpler boards, lower power draw. With this round, it has the capital to turn a proven prototype into a product.