Ajay Joshi Wins NSF Award to Build Light-Powered Chips for Next-Gen AI

by Brendan Galvin, CISE Staff

Artificial intelligence now powers countless everyday tools—from chatbots to image generators—but running these systems consumes enormous computing power and energy. To address this, Hariri Institute Faculty Affiliate Ajay Joshi (ENG) is developing a new type of computer chip that combines electronics and photonics (light-based technology) to make AI faster and more energy-efficient. Supported by the National Science Foundation’s Heterogeneously Integrated Electronic Photonic AI Accelerators (HIEPAA)  award, his goal is to redesign AI hardware so it can handle massive generative AI workloads more sustainably.

Joshi’s innovation lies in integrating many small electronic and photonic chips directly onto a single 12-inch wafer, creating a unified “superchip.” This design allows AI algorithms to run within one compact system instead of being spread across multiple boards, reducing power waste and boosting efficiency. While photonics performs computations quickly and with low energy use, it’s less precise than electronics—so Joshi’s team developed techniques to process data in smaller chunks for both accuracy and speed. This breakthrough could significantly lower the energy costs of data centers run by major tech companies like OpenAI, Google, and Microsoft.

Learn more in this CISE story.