Profile
Running AI on Light
Gasquet, California, is a town of around 900 people tucked into the Smith River National Recreation Area, so deep in the redwoods that the nearest airport is an hour away. It is not where you expect to find the origin story of a $4.4 billion company. Nicholas Harris grew up there anyway.
By the time he finished a BSc in Electrical Engineering at the University of Idaho and spent two years as an R&D engineer at Micron Technology testing DRAM and NAND circuits, Harris had mapped the terrain of conventional computing. He knew what the chips could do. More importantly, he started thinking hard about the wires between them.
A master's at the University of Washington, then a PhD at MIT's Quantum Photonics Laboratory under Professor Dirk Englund, brought him to the frontier. The lab's focus was programmable nanophotonic processors - chips that do computation with light. Harris spent years there building not just theory but physical hardware: designing control and readout systems, writing the software stack, developing calibration techniques. The research was painstaking. The payoff was a new field: programmable photonics became an internationally recognized discipline, in part because of his work.
"The next step in the evolution of computing is all about light."
- Nicholas Harris
In 2017, MIT's $100K Entrepreneurship Competition handed Lightmatter its first validation - and Harris and his co-founders their first prize money. The company launched in September of that year. Harris had a postdoctoral fellowship through the Intelligence Community running concurrently at MIT, and he kept both going through January 2018. Then it was full-time founder.
Lightmatter's earliest product was an AI accelerator chip that ran matrix operations - the core math of neural networks - using photons instead of electrons. The physics advantage is real: light moves through silicon at roughly 100 picoseconds across the chip, faster than electronic clock cycles, and with far less energy per calculation. By 2024, the company had pivoted its commercial focus to Passage, a photonic interconnect product that doesn't replace the GPUs - it replaces the electrical switches connecting them.
The shift was strategic and lucrative. Every major GPU cluster has a networking problem: move data between thousands of accelerators through layers of electrical-to-optical-to-electrical switches, and you burn power, add latency, and cap your effective compute. Harris describes the result with characteristic directness. The GPU is sitting idle 70% of the time, waiting. His analogy: "You've got this Ferrari engine, it's in Manhattan, and there's stoplights everywhere."
Passage collapses that hierarchy. Instead of six or seven layers of network switches between GPUs, Lightmatter's system gets it to two. A single GPU can connect directly to thousands of others. The system currently supports 30 terabits of throughput and can synchronize clusters of up to 1,024 GPUs in specialized racks. Harris has positioned the company as a neutral foundry in the style of TSMC - building infrastructure that serves Microsoft, Amazon, xAI, and OpenAI without publicly endorsing any of them.
"Photonics is coming way faster than people thought - people have been struggling to get it working for years, but we're there. After seven years of absolutely murderous grind."
- Nicholas Harris, 2024
The fundraising record tells the story of the market's verdict. Lightmatter reached unicorn status in December 2023 at $1.2 billion after a $155 million Series C-2. Less than a year later, in October 2024, T. Rowe Price led a $400 million Series D that pushed the valuation to $4.4 billion - a quadrupling in under twelve months. Total capital raised: over $850 million. Investors include Google Ventures and Fidelity Management.
The recognition has kept pace. MIT Technology Review named Harris one of its Innovators Under 35 in 2021, judged by a panel that included Andrew Ng. Previous winners of the same award include Larry Page, Sergey Brin, Mark Zuckerberg, and Jony Ive. In 2025, Goldman Sachs named him one of the most exceptional entrepreneurs of the year. In 2026, the Optical Society of America elected him a Fellow - one of optics' most formal honors. His research record alone would sustain a distinguished academic career: 87+ patents, 82 publications, 10,448 citations, h-index of 34.
Harris's stated view on where all this lands: "Ten years from now, interconnect is Moore's Law." Not transistor density. Not raw compute. The speed at which information moves between processors. That is where the bottleneck lives, and that is where he is building.
