Lumai ™ indicates that by utilizing light for AI inference, it is possible to reduce data center power consumption by 90% and increase processing speed by 50 times.
Oxford, UK, November 3, 2026 / PRNewswire / -- The UK's public research and development funding agency Advanced Research and Invention Agency ( ARIA ) will deploy 3D optical inference technology from Lumai for the first time. Lumai is an optical computing company that is building the infrastructure for the next generation of AI.
As the importance of energy usage in data centers continues to rise on both political and industrial agendas, ARIA will integrate the Lumai Iris™ inference servers into its Scaling Inference Lab, marking it as the first optical inference platform within this laboratory. The laboratory is operated by the non-profit organization CommonAI CIC, which focuses on collaborative AI development. This deployment will enable researchers to assess the value of optical computing in a commercial data center environment, as well as its potential to reduce data center power consumption by 90%. The Lumai Iris servers may also reduce the total cost of ownership to about one-tenth of that of traditional GPU architectures.
The Iris technology of Lumai is capable of running production-level large language models (LLMs). Its working principle involves performing AI calculations within a beam in three-dimensional space, rather than relying solely on transistors in silicon chips. Since numbers are encoded as light intensity and passive optical components such as lenses are used for mathematical operations, the calculation process itself consumes almost no energy. Moreover, as performance improves, energy consumption hardly increases.
"Electricity, rather than demand, is now limiting the speed of AI deployment," said Dr. Xianxin Guo, co-founder and CEO of Lumai. "Operating multi-node systems in commercial data centers is precisely where new computing architectures prove whether they are truly ready for practical implementation. This collaboration with ARIA Scaling Inference Lab has verified our roadmap on a real system in a commercial environment."
ARIA is an independent funding organization supported by the UK government, which promotes advanced AI and other transformative technologies. Established under parliamentary legislation, it is sponsored by the UK Department for Business, Innovation, Science, and Trade. The organization is dedicated to supporting scientists and engineers in advancing ambitious ideas that may lead to significant scientific and technological breakthroughs, especially those projects that can benefit society as a whole. Its Scaling Inference Lab aims to conduct stress tests on the next generation of computing architectures in a real business environment, using standard power supply, cooling, and rack integration.
ARIA Chief Science and Technology Officer David Lewis stated: "Among the platforms we have evaluated, Lumai's technology is the most mature optical computing platform, and it has the potential to make AI推理 more energy-efficient, as well as reduce reliance on the limited global hardware supply chain. Scaling Inference Lab provides a means to publicly and independently test this potential on a rack-scale basis."
The path to more efficient reasoning
As the demand for AI inference accelerates, creating efficient data centers becomes crucial. Traditional AI inference methods operate electronically on silicon and are very power-consuming in large-scale scenarios. If we rely solely on these traditional accelerators to meet the computing needs of even a leading AI laboratory, infrastructure expenditures are estimated to reach trillions of dollars over the next five years. In addition, the additional energy demand will exceed 1,000 gigawatts. Without changing the physical approach to AI computing, these energy and cost requirements will make achieving this goal impossible.
Sovereignty AI requires smarter computing.
For the UK, this challenge is also related to sovereignty. Building the necessary domestic AI computing power in the UK will depend on how to obtain more computing power with limited electricity and grid capacity, as well as on bringing the technology developed in the UK into real data centers.
CommonAI CEO Gavin Ferris stated: "As the UK builds its sovereign AI computing power, it must be done under existing power conditions. We need to find and validate new, innovative AI computing methods. Scaling Inference Lab is designed for this purpose – to test optical computing innovations like Lumai in parallel with today's systems on real workloads."
Addressing compute-intensive prefill workloads
Large model inference is divided into two stages: The prefill stage is limited by computation, while the decode stage is limited by memory bandwidth. Although dedicated inference hardware is increasingly focusing on the decode stage, the approach for Lumai is different. It targets the prefill stage, which is intensive in computation and matrix operations, before the generation of the first token. Lumai Iris can perform millions of matrix operations within a single optical cycle.
Iris servers are not a substitute for silicon chips, but they are particularly effective when used in conjunction with silicon chips to process decode. Compared to systems that use only silicon, the energy efficiency of each inference is increased by at least 16 times, thereby addressing the current limitations of power consumption and cost constraints in AI deployments.
ARIA This deployment is built upon Iris Nova which was launched in April 2026. Iris Nova is the first product of Lumai that is available for evaluation, and it is also the first optical AI accelerator designed specifically for data center-scale inference.
About Lumai
Lumai is an optical computing company that utilizes the unique capabilities of light to advance high-energy efficiency AI inference. Its 3D optical computing architecture can accelerate compute-intensive inference workloads, breaking through the energy consumption limitations of traditional computing and helping data centers increase their AI capacity within their power budgets. Iris Nova is the first server in this series and is currently available for evaluation. Lumai was spun off from Oxford University's world-leading optical research in 2021, with a mission to unleash sustainable intelligence on a global scale. The work of Lumai has won the 2025 Falling Walls Award Scientific Breakthrough Award and the OCP Future Technologies Symposium Best Overall Technology Award, and the company was also a member of Intel Ignite's first London project. Visit Lumai: www.lumai.ai.
About ARIA
Advanced Research and Invention Agency ( ARIA ) is a research and development funding organization in the UK, aimed at bringing about scientific and technological breakthroughs that benefit everyone. By empowering scientists to explore the boundaries of what is possible, ARIA promotes ambitious projects, catalyzes progress, and creates new communities and industries. The organization brings together scientists and engineers with different expertise to break down traditional barriers and pave new paths for technological advancement. https :// aria.org.uk
About Scaling Inference Lab
Scaling Inference Lab is funded by ARIA with 50 million pounds and is implemented by CommonAI CIC. Its aim is to address a key bottleneck faced by startups and research teams developing innovative AI hardware: the lack of an open platform to showcase their technologies in today's AI hardware market, which is dominated by large, closed systems.
