• British AI hardware company Signaloid will preview its recently-taped-out ASIC for physical AI at Bosch Connected World, from 10th–11th June 2026, in Berlin.
• The ASIC is complementary to Signaloid’s edge hardware modules which are already achieving over 37-fold speedup for algorithms used in physical AI and robotics.

CAMBRIDGE, England: British computing technology company Signaloid will preview its C0-ASIC for physical AI this week at Bosch Connected World, taking place from 10th-11th June, in Berlin. Designed for robotics, industrial automation, and probabilistic AI workloads, the ASIC is projected to deliver up to 1000× better performance-per-Watt than existing state-of-the-art approaches.

Signaloid’s distribution-extended compute hardware (UxHw®) is already available for use in physical AI/robotics as a family of hardware modules, as well as via a virtualization- and binary-translation-based solution. UxHw enables autonomous mobile robots (AMRs) to improve their navigation algorithms for safer and faster navigation in factories. It similarly enables industrial programmable logic controllers (PLCs) to achieve better predictive maintenance.

Why Physical AI and robotics needs different compute

Many of the important algorithms enabling robotics and AI today require compute-intensive GPUs or similar hardware. They often involve algorithms that must evaluate hundreds of thousands or even millions of possible scenarios each second, from estimating a robot’s position to tracking a drone in space. Because these scenarios are not equally likely, today’s processors rely on repeated computations to approximate the ideal solutions. If AI hardware could however consider all the possible scenarios when handling any single value, that could enable everything from more efficient AI datacenters to more agile robots and safer autonomous mobility.

A new kind of compute hardware

Instead of single numbers, UxHw can represent values as arbitrary non-uniform ranges (i.e., probability distributions) and performs computation directly on this digital form, without requiring significant software changes. A single execution of traditional software on a UxHw-enabled computing platform can therefore deliver what classical iteration-based methods need millions of repetitions to approximate. In competitive benchmarking against the latest high-end computing platforms, UxHw already delivers 1000-fold speedups, with further gains expected from Signaloid’s C0-ASIC.

What the ASIC will enable

“The compute workloads that Signaloid’s UxHw is designed for, are used across many aspects of computing, from physical AI and robotics, to supply-chain modeling, logistics, and quantitative finance”, says Phillip Stanley-Marbell, founder and CEO of Signaloid. Even before availability of the C0-ASIC, cloud- and FPGA-based implementations of Signaloid’s UxHw are already demonstrating speedups of over 600-fold for infrared sensor data analysis and over 37-fold for particle filter sensor fusion algorithms. The C0-ASIC will complement Signaloid’s existing hardware modules, which are available for use with a range of industrial applications including for integration with the Bosch Rexroth’s ctrlX core X2 and core X3 PLCs.

About Signaloid

Signaloid was founded by Prof. Phillip Stanley-Marbell, a former Professor of Physical Computation at the University of Cambridge and a researcher whose previous roles include Bell Labs, IBM, Apple, and MIT. Signaloid provides a computing platform that benefits computationally-challenging workloads, many of which can be reformulated in terms of algorithms that process probability distributions. Its technology is already used by more than 3,000 users worldwide and is available as cloud, on-premises and low-power edge hardware. www.signaloid.com

Fonte: Business Wire