Researchers in China report 3.2 petabit-per-second fiber speed record
Researchers in China report a real-time transmission of 3.2 petabits per second over 24-core optical fiber, beating their earlier 2.5 petabit result.
According to TechRadar, researchers in China say they have sent data at 3.2 petabits per second through a 24-core optical fiber, a rate the outlet describes as a new real-time speed record. The outlet's headline puts the figure in everyday terms, comparing it to more than 20,000 4K movies moving in a single second. The comparison describes a theoretical capacity, not a typical speed customers would see.
The new result surpasses the team's previous record of 2.5 petabits per second, which was set earlier this year using similar 24-core fiber technology. TechRadar reports that the gain came from combining two methods. The first is spatial division multiplexing, which sends separate data streams through different cores inside one strand of fiber. The second is an approach called S+C+L transmission, which uses three wavelength bands to carry information at the same time.
The report presents the achievement as a laboratory demonstration. It does not describe a commercial product or a timeline for deployment. Some specifics of the test setup, including the distance covered and how long the link was sustained, are not clear from the coverage, so the figure should be read as the researchers' claim as reported by TechRadar.
Background
Optical fiber carries most of the world's internet traffic, from undersea cables linking continents to the connections inside large data centers. Engineers have spent decades increasing how much information a single strand can hold, and the main tools are easy to describe even though the physics is demanding. One approach is wavelength division multiplexing, which sends many signals down one fiber by assigning each its own color of light. Another is to use wider bands of the light spectrum beyond those common in standard equipment, which adds more usable channels.
A third approach treats one cable as several parallel paths. Multi-core fiber packs multiple light-guiding cores into a single cladding, so each core can carry its own stream. This brings its own difficulties. Signals weaken over distance, so amplifiers are needed, and different wavelength bands respond differently to amplification. Cores placed close together can interfere with one another, a problem known as crosstalk, which requires careful design and signal processing.
Units can also be confusing. A petabit is one thousand terabits, or a quadrillion bits. Laboratory speed records are typically set under controlled conditions and often describe raw throughput rather than the experience of end users. Such experiments have a long history in optical research. They show what physics allows, and that knowledge can later guide equipment makers and standards bodies as technologies move toward practical use.
Why it matters
Headline speed numbers matter less for their own sake than for what they reveal about the limits of today's networks. Demand from video streaming, cloud computing and the training of large AI models keeps growing. Being able to move more data through the same physical infrastructure can reduce the need to lay new cables, which is costly and slow. Greater capacity per fiber can also lower the energy and space required for each bit transmitted, a concern for operators whose facilities already draw heavily on power grids.
The result also reflects a competitive field. Many research groups around the world push similar frontiers, and records tend to be surpassed quickly. For researchers, the work suggests that multi-core and multi-band designs can still be extended. For network operators and vendors, it is a reminder that moving from a lab result to a deployed service usually takes years and involves trade-offs in reliability, compatibility with existing systems and manufacturing cost.
Readers should also separate backbone capacity from personal experience. A faster core network does not automatically mean faster home internet, because the final stretch to a house or phone depends on different equipment and economics. Improvements at the center can still ripple outward over time, shaping what becomes affordable and possible at the edges.
What to watch
Watch whether the team or other groups publish fuller technical details, including the fiber length, the amplification approach and how long the link held its performance. Independent verification would strengthen confidence in the claim. It will also be worth tracking whether equipment makers or telecom carriers express interest in multi-core fiber for commercial routes, and whether standards groups begin discussing how such systems would interoperate.
Keep an eye on the gap between laboratory and field conditions. Real cables are buried, bent and exposed to temperature changes, all of which can affect performance. Finally, watch for the next record. The step from 2.5 to 3.2 petabits per second suggests that rival teams may respond, and future results could arrive in new formats that are harder to compare directly.
Viewpoints
Commentary from our columns: three angles on this story.
The Skeptic open questions and weak points
Records are exciting, but they invite hard questions. The headline figure is a peak rate achieved in a research setting, and the coverage does not make clear how long the link ran, over what distance or under which conditions. Those details matter. A petabit-per-second number can look enormous on its own while revealing little about error rates, reliability or whether the setup could survive years of ordinary use.
The movie comparison is also a rhetorical device. Translating bits into films makes the scale feel tangible, yet the conversion depends on assumptions about file size and compression that vary widely. Readers should be wary of letting a vivid comparison stand in for a careful explanation.
None of this diminishes the work. Pushing physical limits is how the field learns. But the sensible response is patience. Wait for independent replication, detailed peer review and evidence from real deployments. Until then, treat the number as a promising claim rather than a settled fact about the future of the internet.
The Builder the engineer's view
From an engineer's standpoint, the interesting part is the stack of techniques rather than the headline total. Combining multiple cores with several wavelength bands means every channel must be managed closely. Crosstalk between cores, amplifier behavior across bands and the cost of advanced transceivers all become design constraints that multiply as systems grow more complex.
The progression from an earlier 2.5 petabit result to 3.2 suggests incremental engineering gains rather than a single radical leap. That is typically how these systems mature. Better amplifiers, tighter control of signal impairments and improved digital processing each contribute a slice of improvement, and those slices eventually reach production equipment.
The practical question is manufacturability. Multi-core fiber must be spliced, terminated and maintained with tools that many field crews do not use routinely. Until connectors, test gear and documentation catch up, lab results remain a roadmap rather than a toolkit. I would watch closely whether the researchers release enough detail for others to reproduce the configuration.
The Optimist the hopeful reading
This is the kind of result that quietly moves the world forward. Each increase in what a single strand of glass can carry expands what networks can support, from large scientific datasets to services that depend on near-instant transfer across long distances. Records like this are signposts showing that the physical capacity of our infrastructure still has room to grow.
There is also reason for hope in the approach itself. Using more cores and more wavelength bands means getting greater value from fiber already in the ground, rather than digging up everything anew. Upgrades that work within existing routes can be cheaper and less disruptive, which could eventually help communities that have long waited for better connections.
Progress of this kind is often collaborative in spirit even when it is competitive in practice. Researchers publish, others build on their work and the frontier moves outward. If the team and its peers share details openly, the benefit could reach far beyond one laboratory. This record is a milestone, and milestones are worth celebrating.