In 2022, a small team of photonics researchers at the Technical University of Denmark (DTU) and Chalmers University of Technology in Sweden quietly set a record that still defines the frontier of fiber-optic speed. Using a chip barely larger than a fingernail, they pushed more than one petabit of data per second through a single strand of standard optical fiber. One petabit is one million gigabits. At that rate, the entire contents of roughly 125,000 Blu-ray discs could cross the Atlantic in a single second.
The peer-reviewed paper, published in Nature Photonics, remains one of the most cited results in modern optical communications. And as of mid-2026, no rival group has matched the feat on standard, already-deployed fiber. That detail is what makes the result so consequential: it suggests the glass already buried under cities, strung across continents, and laid along ocean floors could carry orders of magnitude more data without ever being replaced.
A chip that replaces a room full of lasers
The core innovation is a device called a microcomb. Traditional long-haul fiber links use wavelength-division multiplexing (WDM), a technique that sends many parallel streams of data on slightly different colors of light through the same glass strand. Each color typically requires its own laser, its own temperature controller, and its own power supply. Scaling to hundreds of channels means hundreds of lasers filling equipment racks at both ends of the link.
The DTU-Chalmers microcomb takes a different approach. A single pump laser feeds into a tiny ring-shaped resonator etched onto a photonic chip. Inside that ring, nonlinear optical effects generate dozens of evenly spaced laser lines, each one a usable data channel. Lead researcher Asbjørn Arvad Jørgensen and colleagues encoded data onto these comb lines and multiplexed them into a single fiber, reaching an aggregate throughput that crossed the one-petabit-per-second threshold.
For scale, a typical high-end home broadband connection in the United States tops out around one to two gigabits per second. The demonstrated capacity exceeds that by a factor of roughly 500,000. Even the fastest commercial backbone links, which run at 400 gigabits per second per wavelength channel, look modest next to a petabit aggregate.
Why standard fiber matters
Speed records in optical communications are not new. Japan’s National Institute of Information and Communications Technology (NICT) has repeatedly pushed aggregate throughput even higher, reaching 22.9 petabits per second in a 2024 demonstration. But NICT’s approach relies on multicore fiber, a specialty cable that bundles many separate light-guiding cores inside a single sheath. That fiber does not exist in the ground today. Every meter of it would need to be manufactured and installed from scratch, a process that costs millions of dollars per kilometer for undersea routes.
The DTU-Chalmers experiment used standard single-mode fiber, the same type that already forms the backbone of global telecommunications. If microcomb transmitters can be commercialized, carriers could upgrade capacity at both ends of an existing link without touching the cable in between. In an industry where the most expensive line item is often the fiber itself, that distinction is worth billions.
About that “every movie ever made” claim
Headlines love the comparison, and it deserves a reality check. Estimates of the total number of feature films ever produced range from roughly 500,000 to over one million titles worldwide. The math depends heavily on compression. A tightly compressed HD stream runs about 1.5 to 3 gigabytes per film. At the low end of the catalog (500,000 titles at 1.5 GB each), you get about 750 terabytes, which a sustained one-petabit-per-second link could transmit in roughly six seconds. At the high end (one million titles at 5 GB each in higher quality), the total balloons to about five exabytes, and transmission time stretches to roughly 11 hours.
The “one minute” figure circulating in popular coverage lands somewhere in the middle, assuming a moderate catalog size and aggressive compression. It is a vivid illustration, but readers should understand it is an order-of-magnitude estimate, not a precise calculation. What is not in dispute is the sheer scale of the throughput: one petabit per second is a volume of data that dwarfs anything a single fiber strand has carried before using already-deployed glass.
What stands between the lab and your internet connection
Nearly four years after publication, the microcomb transmitter has not entered commercial network trials, and several hard engineering problems explain the gap.
Reliability under field conditions. The experiment ran on fiber sitting on a vibration-isolated optical table at a controlled temperature. Deployed fiber runs through conduits that bake in summer, freeze in winter, and vibrate near highways. Whether the microcomb can hold stable channel spacing and output power across those swings is an open question.
Bit-error-rate transparency. The published paper does not include granular, channel-by-channel bit-error-rate data across the full comb spectrum. Bit-error rate is the standard measure of whether a data link is clean enough for commercial traffic. Without those numbers, outside engineers cannot fully audit the margin of reliability at each wavelength.
Power and cost. Microcombs are expected to consume less power than racks of discrete lasers, but the study does not publish specific watt-per-bit figures or cost projections. Telecom operators making capital-expenditure decisions need those numbers before they will commit to new transmitter architectures.
Standards and integration. Moving from a lab prototype to a qualified telecom component typically takes years of temperature cycling, humidity testing, mechanical shock testing, and interoperability certification with existing routers, amplifiers, and network-management systems. No public timeline from the DTU or Chalmers teams has surfaced for when a commercial-grade device might be ready.
Where the technology fits in a crowded race
The microcomb result is one entry in a broader contest to keep global bandwidth ahead of demand. Internet traffic has been compounding at roughly 25 to 30 percent per year, driven by streaming video, cloud computing, AI model training, and emerging applications like volumetric video and real-time remote rendering. At that growth rate, backbone capacity that feels comfortable today could become a bottleneck within a decade.
Multiple technologies are competing to close that gap. Multicore fiber, as demonstrated by NICT, offers raw capacity gains but requires new cable. Hollow-core fiber, which guides light through air rather than glass, promises lower latency and potentially higher power handling, but manufacturing it at scale remains difficult. Space-division multiplexing, which uses multiple spatial modes within a single fiber, is another active research front.
The microcomb approach stands out because it targets the transmitter, not the fiber. If the cable already in the ground can carry a petabit, the bottleneck shifts to the equipment at each end. A chip-scale device that generates hundreds of channels from a single resonator could, in principle, be mass-produced using the same photonic foundry processes that already fabricate silicon photonics transceivers for data centers.
That “in principle” carries weight. Photonic chip fabrication has advanced rapidly over the past decade, with companies like Intel, Cisco, and Broadcom shipping millions of optical transceivers annually. Integrating a microcomb source into that supply chain is a plausible next step, but it is not a guaranteed one. The history of photonics is littered with elegant lab demonstrations that stalled on the road to manufacturing.
A record that still sets the pace
As of June 2026, the DTU-Chalmers microcomb experiment remains the benchmark for single-fiber, standard-glass throughput. No subsequent peer-reviewed study has surpassed one petabit per second on the same type of cable that already connects the world. The result proved that the physical capacity of ordinary fiber is far greater than what current commercial systems extract from it, and that a chip small enough to sit on a thumbnail can unlock that capacity.
What it did not prove is that the technology is ready for the messy, expensive, risk-averse world of global telecommunications. The next chapter belongs to engineers working on packaging, thermal management, error correction, and the dozens of other unsexy problems that separate a Nature paper from a product datasheet. If they succeed, the fiber under your street may already be capable of carrying more data than anyone imagined when it was first buried. The glass is not the limit. The light source is.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
