An era for optical networks

JAKE SALTZMAN, EDITOR JAKE.SALTZMAN@PHOTONICS.COM

The stories that surround AI models trained on imprecise or incomplete language seem to follow a similar pattern. The model may show some initial degree of aptitude for the target task. Then over time, bias becomes apparent.

The silver lining is twofold. First, there is unanimous understanding that a more complete training leads to a more dependable model. And, in most cases, any of the would-be damage that a poorly or incompletely trained AI model might cause is identified well before the model sees use in the real world.

But there is no question that words matter when it comes to AI.

This notion occupies another context. Words such as “strain” and “burden” — insofar as they relate to the effect that computationally intensive/AI workloads are poised to have on current data infrastructure — are now used with ubiquity. Though these words are accurate in their meaning, they only encapsulate the problem. What they omit from the dialogue is the opportunity.

The opportunity, in the AI era, is for optimized, robust, and readily deployable optical networks.

This issue of Photonics Spectra isolates this message. It spotlights the innovations in networking, interconnects, transmissivity, and processing that are commanding a deep look from all reaches of the industry value chain. The implications are particularly consequential for optical (and quantum optical) computing. But it will be necessary and insightful to also track the storylines for upstream disciplines such as materials science, as well as for telecom, precision timekeeping, and other applications.

Such a deep dive into networking also marks a shift in scope from a focus on components to one on systems. Optical connectivity solutions provider NewPhotonics articulates this nuanced evolution in its article “In the All-Data Revolution, Optical Solutions Advance Beyond PICs” (Photonics Spectra, July 2024). The company wrote, “The broader optical realm is charting a more immediate and sustainable path. In its aim to replace electrical dependence, its most obvious targets are the latency and power imperatives facing data connectivity for compute and transmission.”

Fast forward three-quarters of a year, and this message is front and center as OFC 2025 dawns.

One more event is important to mention; however, it is unrelated to OFC and optical networking: the 30th anniversary of the quantum cascade laser. Oliver Graydon, chief editor of Nature Photonics, flagged this celebratory occasion for the industry in the journal last month. Now, three decades after its debut, it is reasonable to wonder whether those behind its creation — Federico Capasso, Jérôme Faist, and Alfred Cho, among them — may earn a Nobel Prize for their efforts.

>