A hair-thin strand of glass has just carried more data than the entire internet handles in a typical second. Researchers at Japan’s National Institute of Information and Communications Technology (NICT) transmitted 1.2 petabits per second through a single optical fiber, shattering previous records for a fiber that could actually fit inside the cables already buried under streets and oceans worldwide.
The result, published in Nature Communications in May 2025, is not just a raw speed milestone. What makes it remarkable is that the fiber measures 125 micrometers across, the universal standard diameter used in every deployed telecom network on the planet. Earlier petabit-class experiments relied on oversized, custom-built glass that could never thread through real-world ducts and connectors. This one, at least in principle, could.
19 cores where one used to be
Conventional fiber carries light through a single glass core. The NICT team packed 19 cores into the same 125-micrometer cladding, then did something counterintuitive: instead of isolating each core to prevent crosstalk, they let the cores exchange light energy deliberately.
This technique, called random strong coupling, allows photons to hop between cores in a controlled, statistical pattern. Over distance, the many possible light paths average out, which suppresses a problem called differential group delay, where some signals arrive earlier than others and blur the data. The result is a fiber that can carry far more information than 19 isolated cores could manage on their own.
Twisting the fiber during manufacturing helps govern how aggressively the cores share energy. The Nature Communications paper includes detailed measurements of twist rate, coupling strength, and transmission quality, building on an earlier proof-of-concept the team presented at the OFC 2023 conference, the field’s premier industry and research gathering.
Why 125 micrometers matters so much
Telecom operators have drawn a hard line with equipment vendors: any next-generation fiber must fit existing infrastructure. Ducts, splice machines, connectors, and protective sheaths are all built around the 125-micrometer standard. Fibers with 200-micrometer or larger cladding, common in earlier multicore experiments, bend poorly, break more easily, and require entirely new hardware to deploy.
By hitting petabit-class throughput without exceeding that diameter, the NICT team is making a pointed argument: this is not just a physics stunt. It is a design that could, with further development, slot into the physical plant that carriers have spent decades installing.
For perspective, NICT itself set a previous landmark in 2022 by pushing 1.02 petabits per second through a multicore fiber, but that fiber used a wider cladding. Achieving even higher throughput while shrinking back to the standard diameter represents a significant step forward in making the technology practical.
About that “every movie ever made” claim
The headline comparison is vivid but rough. Estimates of total global film output vary widely, but a commonly cited figure puts it in the range of 500,000 feature films. Encoded at typical streaming quality (around 2 to 4 gigabytes each), that library would total roughly 1 to 2 exabytes. At 1.2 petabits per second, transferring that volume would take closer to two or three minutes than one, depending on the assumptions.
No formal derivation of the “one minute” figure appears in the technical literature. Still, the comparison captures something real: a single petabit-per-second link dwarfs the capacity of today’s commercial backbone fibers, which typically top out in the tens of terabits per second range. The leap is roughly two orders of magnitude.
The long road from lab to ground
Several substantial hurdles stand between this laboratory result and anything a carrier could buy and bury.
Manufacturing at scale. Drawing a preform with 19 precisely positioned cores, coating it reliably, and keeping defect rates low is significantly harder than producing standard single-core fiber. No manufacturer has publicly confirmed that this design can be fabricated at industrial volumes.
Field conditions. The published measurements were taken in a controlled lab environment over a limited fiber span. Real terrestrial and submarine routes stretch thousands of kilometers through temperature swings, vibration, and imperfect splices. Over those distances, the random coupling that suppresses distortion in the lab could interact in more complex ways with noise, nonlinear optical effects, and component tolerances. Independent labs have not yet published splice-loss statistics, bend sensitivity data, or aging results for this fiber.
Cost. Even if the fiber works flawlessly, it must compete economically with a simpler alternative: lighting more wavelengths on conventional single-core fiber using dense wavelength-division multiplexing, a technology that continues to improve. Without public cost models or pilot deployments, it is unclear whether the 19-core design will pencil out beyond high-capacity niches like data center interconnects or next-generation submarine cables.
Reliability. Operators expect buried fiber to last 25 years or more. No long-term durability data exists for this architecture yet.
None of these gaps are unusual. Cutting-edge optical transmission experiments routinely spend years, sometimes a decade or more, crossing the valley between a record-setting demo and a shipping product.
What this actually changes right now
For the average internet user, nothing changes tomorrow. No provider is about to rip out existing cables and thread 19-core fiber through the conduits under your neighborhood.
What the result does change is the engineering conversation about where fiber capacity tops out. For years, researchers have warned of a looming “capacity crunch” as data traffic grows exponentially while single-core fiber approaches its theoretical Shannon limit. Multicore fiber in a standard-diameter cladding offers one of the most credible escape routes from that ceiling, and the NICT result is the strongest evidence yet that the approach can deliver petabit-scale throughput without demanding a wholesale rebuild of physical infrastructure.
As of June 2025, no carrier has announced trials of this specific 19-core fiber, and no independent replication has been published. But the peer-reviewed validation in Nature Communications, combined with the earlier OFC presentation, places the work on solid scientific footing. The next milestones to watch for: a manufacturer announcing production trials, a carrier running a field pilot, and independent labs confirming the performance claims under real-world stress. Until those arrive, this fiber is best understood as a genuinely impressive prototype pointing toward a future that is plausible but not yet guaranteed.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
