StartupsFeatureFuture Circle

How a Glasgow Startup Is Running Artificial Intelligence on Light

LumiAIres builds photonic neuromorphic chips in Glasgow that cut AI power consumption by 90 percent using light instead of electricity.

Future Circle member story13 November 2026 · 2 min read
A woman in a pink suit exploring a colorful and modern laboratory environment.
Photo: ThisIsEngineering

The rapid expansion of artificial intelligence has created a pressing infrastructure challenge, as conventional silicon processors demand unsustainable amounts of power and generate excessive heat. Standard electronic chips rely on billions of operations to emulate neural networks, creating major memory bottlenecks and heavy energy requirements. In Glasgow, hardware startup LumiAIres Ltd is addressing this sustainability crisis by developing a new class of computing hardware. As the company prepares to showcase its progress at Web Summit 2026, it is advancing a physical medium that allows intelligence to run far more efficiently using optical light instead of standard electricity.

The technology behind LumiAIres is the result of years of dedicated academic and industrial research into how light interacts with physical systems. Co-founder Giulia Marcucci spent nine years working across university labs and technical startup environments, investigating ways to construct functional neural networks out of optical signals. Realising that these scientific principles had a clear path to commercialisation, she joined forces with co-founder Jeremy Delhommeau, a manufacturing engineer with fifteen years of experience in high-precision machining and industrial systems. Together, they established a practical product direction designed to translate complex physical research into scalable commercial processors.

Rather than emulating biological neurons through complex digital software running on standard silicon gates, LumiAIres allows neural networks to emerge directly from the physics of light. The platform performs optical computation directly in memory, replacing traditional neural spikes with nonlinear optical interference. Photons travel much faster than electrons, allowing the hardware to perform massively parallel calculations while eliminating costly data movement between memory and processing units. Consequently, the company's neuromorphic photonic processors cut energy consumption by more than 90 percent compared to traditional hardware while achieving ultra-low sub-millisecond latency.

Thermal management presents a major cost and engineering barrier for traditional data centres, but light-based computing operates with near zero heat generation. Consuming approximately one watt of power, the compact photonic processors surpass the energy efficiency of biological intelligence while requiring minimal cooling infrastructure. The hardware is inherently tolerant to electromagnetic noise and radiation, making it durable under extreme physical conditions. These properties allow the portable, solar-powered chips to function reliably in constrained edge environments, ranging from remote industrial installations to Earth orbit.

Beyond low power consumption, the LumiAIres architecture is designed to deliver sovereign and secure AI processing by keeping data strictly local. Because the chips operate locally without relying on continuous cloud connectivity, sensitive information remains protected from external exposure. Built around a reliable European supply chain, the technology offers data centres and edge providers a path to scale their artificial intelligence workloads without hitting energy or infrastructure limits. As LumiAIres continues to develop its photonic platform, the company aims to bridge classical computing with future quantum information processing.

Find out more at lumiaires.ai

Share this story

More in Startups

View more

Events for you

See all
Share