Normal Computing UK, an AI and hardware company, was selected as one of 12 teams awarded funding from the Advanced Research + Invention Agency (ARIA) Scaling Compute Programme. The programme, backed by £50M in funding, aims to reduce AI hardware costs by 1000x while diversifying the semiconductor supply chain.
Normal Computing’s hardware initiative, led by Chief Scientist Dr. Patrick Coles, formerly from Los Alamos National Laboratory, will bring expertise in noise-based computing and thermodynamics to develop physics-based computing chips for matrix inversion and explore applications in training large-scale AI models to transform AI hardware efficiency. Normal Computing’s trademark “thermodynamic computing” approach harnesses noise as a resource rather than fighting against it, aligning with ARIA’s vision to challenge conventional computing paradigms – and to support breakthrough R&D for the UK and beyond.
“The inefficiencies of digital hardware for AI are widely known – one ChatGPT session requires 150 times more power than all-encompassing brain processing,” said Dr. Patrick Coles, Chief Scientist at Normal Computing. “Through ARIA’s Scaling Compute programme, we’re pushing towards the fundamental limits of computational efficiency by allowing physical dynamics, like thermal equilibration, to do computations for us.”
Normal’s approach to de-risking novel AI chip architectures – as well as solving for global talent shortages and brittle silicon infrastructure – depends heavily on AI itself. “We are unique in that AI is helping to design and manufacture our AI chips. The industry struggles to tackle the ‘AI energy crisis’ because of the ‘silicon complexity crisis’,” said Faris Sbahi, CEO at Normal Computing. “Even for the simplest kinds of physical architectures, like memory, complexity is now at the PhD level, so to speak. We trained the first AI to genuinely understand formal chip logic in order to help de-risk chips for our several commercial partners and now with ARIA. It’s analogous to DeepMind’s AlphaGeometry, but for hardware instead of mathematics, and this work is led by former AI leads from Meta and Google Brain.”
Normal Computing’s ARIA R&D Creators combine expertise from quantum computing and thermodynamics, probabilistic machine learning, and semiconductor design. Key team members include Dr. Gavin Crooks, known for the Crooks fluctuation theorem in thermodynamics, and silicon engineering experts Zachary Belateche and Vincent Cheung – who recently exited their last chip startup, Radical Semiconductor – as well as senior technical staff from Graphcore and Broadcom.
The programme spans three core workstreams: (1) software simulation development, (2) advancing networking and interconnect capabilities, and (3) exploring new computational primitives. Normal Computing’s work will contribute to the latter, developing technologies that could open new vectors of progress for computing, with specific relevance for modern AI algorithms.
“If successful, this programme will unlock a new technological lever for next-generation AI hardware, alleviate dependence on leading-edge chip manufacturing, and open up new avenues to scale AI hardware,” said Suraj Bramhavar, ARIA’s Programme Director for Scaling Compute.
The initiative reflects ARIA’s commitment to catalyzing scientific progress and creating new communities and industries. By bringing together diverse expertise from academia, non-profit R&D organizations, startups, and large multinational companies, ARIA aims to accelerate the journey from innovative ideas to real-world applications.