NVIDIA Bets $4 Billion on Silicon Photonics

Published on March 06, 2026 | Translated from Spanish

NVIDIA has announced a strategic investment of $4 billion in photonics companies Lumentum and Coherent, also committing to be a customer of their products. This move is not just a simple financial investment, but a key positioning for the future of its AI platforms. The company confirmed that, starting in 2026, its systems will use silicon photonics-based optical interconnects for communications between GPU clusters, a radical technological shift to power the next generation of high-performance computing.

NVIDIA logo next to a chip with luminous optical circuits, representing the investment in silicon photonics.

Visualizing the paradigm shift: from electrons to photons 🚀

Silicon photonics represents a fundamental evolution in chip and system architecture. While traditional electrical interconnects face physical limitations in bandwidth, distance, and energy consumption, this technology integrates optical components directly into the silicon substrate. Using modulators and detectors, it converts electrical signals into light pulses (photons) that travel through tiny waveguides. Visualizing this process in 3D allows understanding its complexity: silicon layers are etched with nanometric precision to create these hybrid circuits, combining electronic logic with optical transmission. This leap is crucial for connecting thousands of GPUs in AI clusters, where the bottleneck is no longer computing power, but the speed and efficiency of communication between nodes.

The future of microfabrication is hybrid âš¡

NVIDIA's bet validates that the future of semiconductor manufacturing lies in hybrid integration. It's not just about making smaller transistors, but incorporating new functionalities like optics into the chip itself or the substrate. This decision will boost the silicon photonics industry, demanding advances in lithography, material deposition, and packaging. For professionals in 3D modeling and process simulation, a fascinating field opens up: visualizing and optimizing these new architectures where light and electricity converge, defining the new standard for data centers and exascale computing.

How can NVIDIA's investment in silicon photonics accelerate the arrival of optical computing and redefine the limits of 3D microfabrication for next-generation chips?

(P.S.: integrated circuits are like exams: the more you look at them, the more lines you see)