Cornell Tech Updates Robot AI With Light: An Optical Chip in Place of Power-Hungry Circuits

Cornell Tech researchers unveiled an optical receiver in June 2026 that updates an AI model's parameters directly in chip memory, using flashes of light in the form of QR-like matrices and no power-hungry analog circuits. The work was presented at the IEEE/JSAP Symposium on VLSI Technology & Circuits.

How light rewrites chip memory

The receiver directly alters its own memory using photocurrents generated by a beam aimed at it. In a demonstration, postdoc Yifan He placed the receiver's lens nearly a meter from a red LED, and a connected monitor instantly displayed an array of squares resembling a QR code. Unlike a QR code, which hides a simple link, this optical code carries neural network parameters.

In the system, DRAM sits next to the transmitter, while the receiver is built into the processor's SRAM. The transmitter shines light onto an array of SRAM cells that have had photodiodes added: light striking each photodiode generates a current and flips the binary values in memory. Today's optical receivers lose light's advantage to power-hungry analog circuits, whereas Cornell Tech's technology receives digital matrices directly and stays fully digital.

• Authors: postdoc Yifan He and associate professor Jae-sun Seo, Cornell Tech (New York)
• Where presented: IEEE/JSAP Symposium on VLSI Technology & Circuits, June 2026
• Transmitter prototype: a static 14×14-bit matrix projected through a metal mask
• Goal: millions of matrix updates per second, gigabit-per-second transfer
• Independent expert: Dennis Sylvester, IEEE Fellow, University of Michigan

Why the chip is still far from a product

Photosensitive cells are larger than ordinary SRAM cells, so such a chip fits less memory, and that trade-off could eat up the entire energy-efficiency gain. That is the warning from Dennis Sylvester of the University of Michigan, who was not involved in the work. The Cornell Tech team is already working to shrink the cells through transistor optimization and CMOS scaling.

The transmitter prototype shown in the lab is only a proof of concept: it emits a static 14×14-bit matrix through a metal mask. Real-world applications will require an optical transmitter that can change the matrix millions of times per second and transfer gigabits per second; the authors are developing one together with optics research groups.

"People are designing all kinds of AI chips," says Jae-sun Seo, associate professor of electrical and computer engineering at Cornell Tech.

In his words, the electrical connections between DRAM and the processor create cost and efficiency problems at scale: "This is one of the major bottlenecks."

Why robots and edge AI care

Seo and He are aiming at robotics and other edge applications. One example is warehouses and factories full of AI robots: optical data transfer would save time and energy when updating the models in each robot. Microrobots, whose small size severely limits their memory, could also benefit, although they would need an even more compact design.

The independent expert rates the technology's potential highly, saying the solution has "massive commercial implications."

"Edge AI is a big growth area, and in three, four, five years we'll be talking about it about as much as we talk about data centers, as intelligence increasingly migrates into our devices," —

Dennis Sylvester, IEEE Fellow, University of Michigan.

What this means

Delivering data optically straight into chip memory is an attempt to remove one of the main bottlenecks in AI hardware: the costly, power-hungry exchange between DRAM and the processor. The technology is far from a product, but it targets the growing edge AI market, where every watt and millisecond counts.

Frequently asked questions

How does light change a chip's memory?

Photodiodes are built into the SRAM cells. When light from the transmitter hits them, it generates a current that flips the binary values in memory, so model parameters are updated directly, without analog converters.

Why isn't the technology ready for sale yet?

Photosensitive cells are larger than standard SRAM cells, so less memory fits on the chip. According to Dennis Sylvester of the University of Michigan, this trade-off could cancel out the energy-efficiency gain until the cells are shrunk.

Where can it be used?

The authors cite warehouses and factories with AI robots, as well as microrobots with limited memory. Updating models optically would save time and energy compared with sending data over metal wires.