Reinventing the Physical Layer: The Path to the Million-XPU AI Data Center
ABSTRACT
We introduce Lightmatter’s approach to reinventing the physical layer of networking, utilizing the Passage photonic interconnect platform to enable 100x more bandwidth and create "single-brain" data centers capable of linking 1,000,000 XPUs. By moving from traditional pluggables to 3D photonics, we demonstrate energy efficiencies of 2 pJ/bit with edgeless I/O. Key advancements in silicon photonics, 3D packaging, and high-density laser integration enable this transformation. Finally, we present simulation data showing how this optical interconnect architecture significantly reduces training time for massive Mixture of Experts (MoE) models compared to traditional electrical scale-up architectures.
BIOGRAPHY
Nick Harris is the founder and CEO of Lightmatter, a pioneering photonic-computing company that is redefining AI infrastructure.
An MIT-trained engineer and scientist, Nick has authored over 70 publications in top-tier journals, including Nature, Nature Photonics, and Nature Physics, and holds over 30 patents on revolutionary photonic technologies. His seminal work on quantum and classical information processing has seeded new fields in photonic AI interconnects, processor design, and quantum computing. He has been elected to the prestigious Optica 2026 Fellows Class and was recognized by the MIT Technology Review’s Innovators Under 35 (TR35) award (2021), an honor whose previous winners include Google founders Larry Page and Sergey Brin.
Under his leadership, Lightmatter has rapidly become the industry benchmark for ultra-fast photonics for connecting AI supercomputers. He received his doctorate in Electrical Engineering and Computer Science from the Massachusetts Institute of Technology, where he was also a National Science Foundation Graduate Research Fellow and an Intelligence Community Postdoctoral Fellow.