As the race to build smarter AI intensifies, so does the demand for powerful and energy-efficient computing. Today’s data centers, built on traditional silicon chips, are hitting their limits—both in performance and sustainability. But a University of Oxford spinout may have cracked the code. Meet Lumai, a groundbreaking startup using optical processing to push AI beyond the bottlenecks of silicon.
Backed by over $10 million in fresh funding, Lumai is bringing a new class of AI acceleration hardware to market—one that uses light, not electrons, to supercharge large language models (LLMs) and other compute-heavy AI systems. And it might just change the way data centers operate forever.
Tackling AI’s Power Problem: Lumai’s $10M+ Raise Fuels U.S. Expansion
With U.S. data center electricity consumption on track to triple by 2028—potentially using up to 12% of the country’s total power—AI’s energy footprint is becoming a pressing concern. That’s where Lumai steps in.
The company recently closed a funding round exceeding $10 million, led by Constructor Capital, with participation from IP Group, PhotonVentures, Journey Ventures, LIFTT, Qubits Ventures, State Farm Ventures, and TIS Inc. This round will drive Lumai’s product development, double its team size, and fuel expansion across the U.S.
Reinventing AI Compute: The 3D Optical Processing Edge
Founded in 2022 by Oxford researchers Tim Weil, Xianxin Guo, Alex Lovsky, Thomas Barrett, and James Spall, Lumai is reimagining how AI models are trained and deployed.
Traditional silicon-based chips, even cutting-edge GPUs, struggle with the scale and energy demands of modern AI workloads. Lumai’s solution? Optical processors that compute using light. Their system replaces electrons with photons, unlocking faster speeds, lower latency, and far greater energy efficiency for both training and inference.
The startup’s 3D optical computing technology handles matrix multiplications—the core arithmetic behind neural networks—by sending light beams through a three-dimensional space. This design avoids the energy bottlenecks of silicon and even surpasses the limits of current photonics-based systems.
Designed for Speed and Sustainability: AI Compute That’s 90% More Efficient
Lumai’s optical accelerator slots directly into existing data centers, offering a PCIe-compatible form factor and low-cost optical components. But the impact is anything but incremental.
The hardware promises 50x performance gains compared to current silicon-only systems while slashing energy use by up to 90%. For cloud operators and AI labs, this could reduce inference costs to just a tenth of today’s rates—dramatically lowering both capital and total ownership costs.
According to CEO Tim Weil, “The cost of today’s LLMs is unsustainable. Future AI breakthroughs depend on a complete shift in computing. Lumai’s optical design solves the power and scalability issues that have long held back innovation in AI hardware.”
Building Momentum: Global Recognition and Strategic Partnerships
Lumai’s bold approach is already winning industry recognition. It clinched the ‘Best Overall Technology’ award at the Global OCP Future Technologies Symposium and was chosen for the inaugural Intel Ignite London cohort—a prestigious program for high-growth deep tech startups.
Co-founder Dr. Xianxin Guo joined the Royal Academy of Engineering’s Shott Accelerator 2024, while Dr. James Spall was named to the Photonics 100 list for 2025, a nod to his work in shaping the future of optical technologies.
Investors Back Lumai’s Vision for the Next-Gen Data Center
Dr. Serg Bell, founder of Constructor Capital, sees optical computing as a core piece of AI’s future: “Photons are the only viable energy source to scale intelligence beyond biological limits. Lumai’s optical system offers quantum-like leaps in performance for AI’s most critical calculations.”
Deep tech investor Dr. Lee Thornton from IP Group agrees, noting that Lumai has “cracked the code” on optical compute by delivering a low-cost, scalable solution ready for deployment.
Ewit Roos, General Partner at Photon Ventures, added, “Lumai isn’t just iterating—it’s reshaping the entire landscape of AI infrastructure. This is one of the most compelling breakthroughs we’ve seen in next-generation compute.”
As AI models grow more powerful, the need for better hardware is becoming urgent. Lumai’s optical processors could be a game-changer—bringing speed, scale, and sustainability to the heart of data centers. With a world-class team, cutting-edge tech, and strong backing, Lumai is positioned to become a foundational layer in the AI stack of tomorrow.
