Researchers have pushed a single strand of glass fiber, thinner than a human hair, to carry data at rates that dwarf what entire bundles of copper cable can manage. A peer-reviewed experiment achieved 44.2 Tb/s over 75 km of standard optical fiber using one integrated chip source, while a separate effort recorded 120 Tb/s on a single-mode fiber by widening the usable light spectrum with a new amplifier design. Together, these results show that existing fiber already in the ground can handle far more traffic than operators currently send through it, a finding with direct consequences for internet providers, cloud companies, and subsea cable operators racing to keep up with surging global bandwidth demand.
Why single-fiber capacity gains change the cost equation for operators
The practical tension behind these results is straightforward: laying new fiber or submarine cable is expensive and slow. Permits, trenching, and ocean-floor installation can take years and cost billions of dollars per route. If engineers can instead squeeze dramatically more data through fiber that is already installed, operators avoid those capital outlays and accelerate upgrades.
The 44.2 Tb/s result, published in the journal Nature Communications, used a micro-comb chip that generates dozens of evenly spaced wavelengths from a single laser source. Each wavelength carries its own data channel, a technique called wavelength-division multiplexing, or WDM. Because the chip replaces racks of individual lasers, it shrinks the transmitter hardware while boosting total throughput on standard fiber over a 75 km link. That distance is long enough to be relevant for metro rings connecting data centers and exchange points around large cities, suggesting the approach could slot into existing network topologies.
The 120 Tb/s experiment took a different approach. Instead of reinventing the light source, a team at University College London developed a hybrid amplifier that combines Raman amplification with erbium-doped fiber amplification across a continuous 91 nm bandwidth window. Standard telecom amplifiers cover roughly 35 nm. By nearly tripling that usable window, the amplifier lets signals occupy far more of the light spectrum without unacceptable noise penalties. The result, detailed in a preprint, reached 120 Tb/s on one single-mode fiber, a capacity figure that exceeds the total traffic on many national backbone links.
If hybrid Raman-EDFA bandwidth continues scaling at the observed 91 nm increment per generation, single-fiber line rates could exceed 200 Tb/s on installed standard fiber within a few years, independent of new cable builds. That projection rests on a big assumption: that noise figures and signal quality hold up as the amplification window widens further and that transceiver electronics can keep pace. But even a partial step toward that target would reshape how operators plan network upgrades, shifting spending from civil works to electronics and amplifier swaps.
For subsea cable consortia, the economics are particularly stark. A transoceanic system might consist of just a few fiber pairs, each carrying multiple WDM channels. Pushing per-fiber capacity higher delays the need for entirely new wet plant and lets owners monetize existing routes longer. Terrestrial operators face similar incentives in dense urban corridors where digging up streets is politically and logistically difficult. In both cases, squeezing more out of existing glass can be more attractive than adding new strands.
Chip-scale sources and wideband amplifiers as complementary breakthroughs
These two results attack the same bottleneck from opposite ends of the optical link. The chip-based micro-comb demonstrated in the Nature Communications paper solves a transmitter problem. Traditional WDM systems need one laser per channel, each locked to a precise frequency. A micro-comb generates all those frequencies from a single device, cutting power consumption and simplifying the hardware. The 44.2 Tb/s line rate over 75 km of standard fiber proved the concept works at distances relevant to metro and regional networks, not just lab benches, and showed that the comb lines were stable enough for advanced modulation formats.
The UCL amplifier, documented in the university’s institutional repository, solves a mid-link problem. Optical signals weaken as they travel, and amplifiers must boost them without adding too much noise. By combining distributed Raman gain with lumped erbium-doped fiber gain across both the C and L bands, the 91 nm hybrid amplifier keeps noise low enough for high-order modulation formats that pack more bits into each wavelength. That is what allowed the 120 Tb/s total on a single fiber. The design specifically targets long-haul and subsea transmission, where repeaters are spaced tens of kilometers apart and reliability requirements are stringent.
Taken together, a compact chip source feeding into a wideband amplifier chain could, in principle, combine the advantages of both approaches. One device generates dozens of channels; the other ensures those channels survive long-haul distances with acceptable signal-to-noise ratios. No published experiment has merged the two in a single end-to-end demonstration, but the physics are compatible, and the incentive is clear: operators want maximum capacity per fiber strand with minimum new infrastructure.
Integrating these innovations into commercial systems would require more than just swapping components. Network designers would need to address thermal management on dense transceiver cards, ensure that micro-comb sources remain stable over temperature and aging, and validate that wideband amplifiers behave predictably under real traffic patterns rather than carefully shaped laboratory signals. Still, both results point toward a future in which the limiting factor in backbone capacity is less about glass in the ground and more about the sophistication of the terminal equipment at each end.
Missing data on cost, reliability, and copper comparisons
Several gaps in the evidence deserve attention. Neither the Nature Communications paper nor the UCL amplifier record includes cost-per-bit or energy-per-bit figures. Those numbers matter because a technology that doubles capacity but triples power draw may not pencil out for commercial deployment, especially in submarine systems where powering repeaters across thousands of kilometers is already a major constraint. Without clear measurements, it is difficult for operators to compare these approaches to incremental upgrades of existing equipment.
Reliability is another open question. Micro-comb chips and hybrid Raman-EDFA chains introduce additional complexity compared with conventional laser arrays and single-band amplifiers. Long-term behavior under temperature cycling, vibration, and real-world fault conditions is not yet documented in the available research. Operators planning multi-decade submarine investments tend to favor conservative, well-understood technologies; convincing them to adopt newer designs will require extensive field trials and qualification.
The headline comparison to copper wires also lacks a precise anchor in these studies. Copper twisted-pair cables used in older telephone and broadband networks top out at single-digit gigabits per second over short distances, so the qualitative gap is enormous. But the primary papers do not include a measured copper-bundle benchmark at equivalent distances, which means the “thousands of copper wires” framing is directional rather than experimentally validated. It illustrates the general magnitude of improvement but should not be read as a rigorously derived ratio.
There are also practical limits beyond the fiber itself. Router line cards, switching fabrics, and data-center interconnect architectures must all evolve to ingest and process 100 Tb/s-class links. Simply upgrading the optical layer without addressing the electronic bottlenecks risks creating choke points elsewhere in the network. In that sense, the optical advances are necessary but not sufficient for end-to-end capacity gains.
Still, the core message from these experiments is hard to ignore. Standard single-mode fiber, deployed worldwide over the past several decades, has far more headroom than today’s commercial systems exploit. By refining light sources and amplifiers, researchers have shown that individual strands can carry tens to hundreds of terabits per second over distances relevant to both metro and transoceanic links. As traffic growth continues and spectrum becomes crowded in wireless domains, the ability to unlock that latent capacity in buried glass will be central to keeping the internet’s physical backbone ahead of demand.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
