In Berkeley, California, there is a company using light to take a swing at the limits of computing. Opticore, founded in 2024 by Dr. Zaijun Chen of UC Berkeley, Dr. Ryan Hamerly of NTT Research, and Dr. Mengjie Yu of UC Berkeley, is betting that photons can do what electrons no longer can. Their optical processing units use waveguides instead of wires, chasing 100x lower energy consumption and 25x greater computational density than the GPUs dominating racks today. This is not vaporware. These are chips fabricated with standard foundry processes on nodes older than 28nm, designed to slip into existing manufacturing capacity, ready to fuel the multi-trillion-dollar race in AI infrastructure, HPC, and hyperscale data centers.
The financing tells the story of momentum. Opticore just closed a $7.5 million seed round on September 10, 2025, co-led by Origin Ventures and Jetha Global, with Neotribe Ventures, Thunderbolt Ventures, Sagax Capital, and Bioeconomy adding fuel. This round stacks on the $5 million pre-seed in December 2024 from Sagax Capital and Jetha Global, putting total capital raised at $14.5 million. Investors like Prashant Shukla of Origin Ventures, Karan Danthi of Jetha Global, Alex Turnbull of Sagax Capital, and Dorjee Sun of Bioeconomy are not backing theory, they are backing working prototypes that already demonstrate 100 TOPS per watt compute efficiency, validated in Physical Review X (2019), Nature Photonics (2023), and Science Advances (2025).
Here is the strategic edge: Opticore is not chasing bleeding-edge silicon nodes. It is betting on proven, accessible manufacturing where global capacity exists but demand does not. While others fight for the scraps of 3nm, Opticore is lining up fabs with room to scale. Their architecture processes billions of parameters on photonic integrated circuits, with time-multiplexed computing that enables dynamic weights for both training and inference. In short, it is a system designed not just to outperform but to outlast.
The founders are academics with the execution mindset of operators. Dr. Zaijun Chen, Ph.D. from Max-Planck Institute of Quantum Optics. Dr. Ryan Hamerly, Ph.D. from Stanford, former MIT fellow, whose theoretical work laid the foundation for the company’s design. Dr. Mengjie Yu, Ph.D. from Cornell, sharpened at Harvard, pushing the edge of photonic integration. Around them, advisors like MIT’s Dirk Englund and InnoLight’s Rang-Chen Yu bring a mix of deep research and market experience.
Opticore plans to demonstrate its first scaled system in 2026 after tapeouts in 2025. Partnerships with HPC system integrators like HPE and Dell and cloud providers who live and die by efficiency are already in focus. With validated science, capitalized momentum, and hardware ready to scale, the message is clear. The next leap in AI acceleration may not be electric, it may be optic.
