Nvidia shows optics-on-interposer prototype, moving optical engines onto the chip interposer

Nvidia has presented next-generation co-packaged optics (CPO) research at SEMICON West 2026 in San Francisco: an optics-on-interposer (OOI) prototype and dense wavelength-division multiplexing (DWDM) technology. The research was presented on October 13 by Nikola Nedovic, a principal researcher at Nvidia, under the theme “Optical Integration for the Next Generation of AI.”

A different place for the optical engine

Existing co-packaged optics place optical engines on the package substrate. OOI moves them directly onto the interposer that carries the computing chips. Signals then travel through the interposer's short, fine-pitch interconnects — similar to the architecture linking graphics processors and high-bandwidth memory — instead of passing through the organic substrate. According to The Elec, removing the organic substrate from the electrical signal path could raise bandwidth density and cut power consumption.

256 Gb/s on a single fibre

The DWDM demonstration sends separate data streams on multiple wavelengths of light down one optical fibre, separating them at the receiver. The demo used eight wavelengths at 32 Gb/s each, for 256 Gb/s per fibre.

What comes next

Both technologies remain in the research stage. Nedovic said Nvidia is proposing and evaluating them for future products while addressing the challenges that prevent commercialization, including device and circuit design trade-offs, thermal management and packaging. No production timeline or named customer has been announced, and further technical details are expected at future GTC events. For now, OOI is a research direction for AI-server interconnects, not a product on sale.