Photonic Computing
Accelerating Tensor Operations at the Speed of Light
Improving power and performance for the future of AI
AI has hit a physical power wall. This talk looks at the hardware shift being built to break through it — silicon photonic processors that compute with light instead of electrons.
When electrons run out of headroom, use light
As AI models scale toward trillion-parameter architectures, traditional electronic tensor processors are hitting a physical power wall. The energy required for massive matrix-vector multiplication — the core operation of modern AI — has become the primary constraint on both performance and sustainability.
This discussion explores silicon photonic tensor processors, where light is used not just for high-speed interconnect but as the medium of computation itself. By calculating through the interference and modulation of light, these chips target orders-of-magnitude gains in energy efficiency and latency over today's GPUs and TPUs.
Integrating photonics into the silicon stack is a pivotal transition for the industry. Beyond AI, the architecture offers a scalable roadmap for supercomputer simulation, real-time high-bandwidth multivariable control systems, and the specialized demands of quantum computing.
The Power Wall
Why energy per matrix-vector multiplication — not clock speed — is now the limit on how far AI models can scale.
Computing With Light
How interference and modulation perform tensor operations on-chip, and where the efficiency and latency gains come from.
Beyond AI
What photonics in the silicon stack means for HPC simulation, high-bandwidth control systems, and quantum computing.
Walt Tucker
MIT Class of '85, Course VIII · Managing Partner & Principal Investigator, Omega Photonics Systems
Walt Tucker
Walt Tucker is managing partner and principal investigator at Omega Photonics Systems in Orlando, FL. He works with CREOL at the University of Central Florida to commercialize a silicon photonic processor designed to significantly improve the power and performance of AI — in the data center and at the edge.
Engineers, researchers, and technology leaders
A technical talk with an accessible on-ramp. Come with curiosity about where AI hardware is heading — and stay for the conversation afterward.
Three communities, one evening
MIT Alumni of RTP
MIT Regional Alumni Network of Research Triangle Park — connecting MIT alumni across the Triangle through technical programming and community.
PanIIT RTP
Connect, Socialize, Support, and Grow — the IIT alumni community of the Research Triangle.
SAM IT Solutions
Durham-based technology partner and host venue for the evening, bringing practitioners together around emerging technology.
Seats are limited
Tickets are $10 and include light appetizers and drinks. Registration is handled by the MIT Alumni Network of RTP.
Questions? [email protected]