Submarine optical cables carry the overwhelming majority of intercontinental data traffic, and the surge in AI training, cloud interconnect and video distribution is putting unprecedented pressure on this layer of the internet. Industry headlines increasingly talk about "single-wave" speed records, but the numbers behind those headlines are easy to misread. This article explains how submarine cable capacity is actually measured in 2026, what coherent optics like 800G, 1.2T and 1.6T per wavelength can realistically achieve, and how cable design and manufacturing constrain the upgrade path.
Why Submarine Cables Still Define Global Internet Capacity
Despite the visibility of low Earth orbit satellite services, satellite links remain a small fraction of intercontinental capacity. Industry sources, including the U.S. Federal Communications Commission and analyses by TeleGeography, indicate that submarine cables carry well over 95% of international data traffic, with figures commonly cited in the 95–99% range. According to TeleGeography's Submarine Cable FAQs, more than 1.5 million kilometres of submarine cable were in service globally as of early 2026, and the firm currently tracks over 600 active and planned systems on its 2026 Submarine Cable Map.
Satellite communications complement this infrastructure in remote regions and as a backup for resilience, but the bulk of the bandwidth that allows cross-border video calls, cloud workloads and AI inference traffic still travels through glass fibre on the seabed. Readers new to the topic may find a short primer in our overview of fibre optic cables in the ocean useful before going further.
What Is Submarine Cable Capacity?
Most "record-breaking capacity" stories blur three different metrics. Keeping them separate is essential for any technical or procurement decision.
Per-wavelength capacity (per channel) describes how much data a single optical channel - one wavelength of light - can carry on the cable. Modern fifth- and sixth-generation coherent transponders typically offer 800 Gb/s, 1.2 Tb/s or 1.6 Tb/s per wavelength, with the achievable rate depending strongly on distance, fibre type and the rest of the line system.
Per-fibre-pair capacity is the total throughput of a single pair of fibres (one for each direction), summed across all wavelengths multiplexed onto that pair through Dense Wavelength Division Multiplexing. Real production capacities on long transoceanic routes are typically in the high tens of Tb/s per fibre pair.
Per-system (per-cable) capacity is the total of all fibre pairs in the cable. Submarine systems usually carry between 8 and 24 fibre pairs. As TeleGeography's 2026 transport network review notes, submarine cables are practically limited to roughly 24 fibre pairs because the optical amplifiers along the route have to be powered from shore.
When a press release talks about "Pbps-class capacity," it almost always refers to a per-system figure across all fibre pairs, not what a single wavelength can carry. To learn more about how multiplexing scales fibre throughput, see our discussion of DWDM in high-capacity telecommunications.
Where Per-Wavelength Capacity Actually Stands in 2025 and 2026
Recent public deployments and field trials make the realistic envelope clear:
In March 2026, Ciena and Meta announced an 800 Gb/s single-carrier wavelength transmission across an unregenerated 16,608 km link on Meta's Bifrost cable system between the U.S. West Coast and Asia, using WaveLogic 6 Extreme coherent optics. The trial reportedly delivered a total fibre pair capacity of around 18 Tb/s. The technical details are summarised in Ciena's announcement of the Bifrost result.
Earlier, Colt achieved 1.2 Tb/s per wavelength on its Grace Hopper transatlantic cable using the same WL6e generation, and Altibox Carrier and Ciena demonstrated 1.6 Tb/s per wavelength on the NO-UK route in 2025, although on a much shorter span than full transoceanic paths.
Two implications matter for anyone reading these numbers. First, the headline single-wavelength figure scales roughly inversely with distance: 1.6 Tb/s is achievable on regional or short subsea spans, while transpacific links are still mostly in the 800 Gb/s per-wavelength regime. Second, claims of "24 Tbps per single wave" or comparable numbers do not match any publicly verifiable system in operation as of early 2026, and should be treated cautiously. The widely cited "24 Tbps" figure on cables such as PEACE refers to per-fibre-pair capacity, not per-wavelength capacity.
Why AI Is Pushing Operators to Upgrade Subsea Capacity
Hyperscale cloud and AI workloads have changed the shape of demand on submarine networks. Model training distributes data and gradients between geographically separated compute clusters; AI inference serves users across regions; and content distribution networks pre-position increasingly large media payloads. The aggregate effect is sustained, multi-year double-digit growth in international bandwidth demand.
Operators have responded along three tracks: building new high-fibre-count cables, retrofitting existing wet plant with new terminal equipment, and adopting space-division multiplexing approaches that increase fibre count per cable. The market analyst view, summarised in TeleGeography's 2026 outlook, suggests roughly 40 new submarine cables are projected to enter service in 2026, representing capital expenditure on the order of USD 6 billion. For a manufacturer-side perspective on these dynamics, see our analysis of how AI is reshaping the global optical communications market.
Can Existing Submarine Cables Be Upgraded?
Yes, but with conditions. The wet plant - the cable, repeaters and branching units on the seabed - is built for an engineering life of 25 years or more. The dry plant - the Submarine Line Terminal Equipment in the cable landing stations - has a much shorter refresh cycle, typically 5 to 7 years. By replacing the SLTE with newer coherent transponders, operators can extract more capacity from the same wet plant.
How much more depends on several factors:
Fibre type and condition. Cables built with G.652.D fibre support coherent upgrades but have higher attenuation and tighter Shannon-limit constraints than those built with low-loss G.654.E or pure-silica-core fibre. Newer transoceanic cables increasingly use G.654.E fibre, which is optimised for long-haul, high-power coherent transmission.
Repeater and amplifier performance. Existing repeaters along the route limit the spectrum that can be used. C-band-only systems cannot be expanded into the L-band without replacing or supplementing the amplifiers, which on the seabed is generally not feasible.
Spectrum plan and channel spacing. Higher per-wavelength rates often need wider channel spacing, which can reduce the number of channels that fit into the available spectrum, partially offsetting the gain.
Operating margin. Older cables operating close to their Shannon limit have less headroom to increase modulation order without raising the bit error rate.
The honest framing is that terminal-equipment refreshes can multiply usable capacity by a factor of two to several times on a given cable, at a small fraction of the cost of laying a new system. They cannot, however, indefinitely substitute for new build, and the achievable gain varies cable by cable.
What This Means for Submarine Cable Design and Manufacturing
From a manufacturer's perspective, the AI-driven capacity push is reshaping requirements at the cable-build stage rather than only at the terminal-equipment stage. Several design choices matter more than they did a decade ago.
Fibre selection. Long unrepeatered or transoceanic spans favour G.654.E single-mode fibre for its larger effective area and lower attenuation. Choosing the right fibre at design time effectively sets a ceiling on the cable's lifetime capacity.
Fibre count and space-division multiplexing. Modern submarine systems are moving toward 16 to 24 fibre pairs, leveraging space-division multiplexing to scale capacity even when the per-fibre-pair Shannon limit is approached. This implies more compact fibre packaging and tighter requirements on cabling structure.
Mechanical protection. Cables in shallow water, on continental shelves and in fishing zones face mechanical risks that the deep-sea sections do not. Armouring layers, water-blocking compounds and the outer sheath must be matched to deployment depth and seabed conditions. Our guide to fibre optic cable structure from core to sheath outlines these layers in detail.
Power delivery to repeaters. Because submarine optical amplifiers are powered from shore, repeater design and the cable's power conductor are tightly coupled to the maximum number of fibre pairs the system can support.
Manufacturing and testing. Submarine fibre optic cables are subject to demanding factory acceptance testing, including pressure, tensile, water-blocking and optical performance tests. Hengtong's underwater fibre optic cable product family and broader fibre optic cable manufacturing processes illustrate the engineering depth involved.
Sustainability considerations are also becoming part of buyer requirements. Industry discussion on this topic is summarised in our piece on sustainable subsea cables and global connectivity.
FAQ
Q: Is "Single-Wave 24 Tbps" A Real Submarine Cable Specification?
A: Not as a per-wavelength figure on any publicly verifiable system in operation as of early 2026. Where 24 Tbps appears in cable documentation, such as on the PEACE Mediterranean segment, it generally refers to per-fibre-pair design capacity. Verified per-wavelength capacities on long transoceanic routes are currently in the 800 Gb/s to 1.2 Tb/s range, with 1.6 Tb/s per wavelength demonstrated on shorter spans.
Q: How Is Submarine Cable Capacity Actually Scaled?
A: Through three combined techniques: higher-order modulation and faster baud rates per wavelength (coherent optics), wavelength-division multiplexing to fit more channels per fibre pair, and space-division multiplexing to add more fibre pairs per cable. Recent gains come mostly from the second and third levers, since per-wavelength capacity is approaching the Shannon limit of installed fibre.
Q: Can Old Submarine Cables Really Be Upgraded By Changing Only The Terminal Equipment?
A: In many cases yes, but the gain depends on the original fibre type, repeater bandwidth and operating margin. Cables built in the past 10 to 15 years with G.654.E fibre and C+L band repeaters tend to upgrade well; older C-band-only systems gain less.
Q: How Long Do Submarine Cables Last?
A: The standard engineering design life is 25 years, although cables are often retired earlier when they become economically obsolete relative to newer systems with higher capacity per dollar.
Q: Why Is The Per-Cable Fibre Pair Count So Limited?
A: Because amplifiers along the route must be powered from shore, and the voltage and current that can be delivered through the cable's metallic conductor place a practical limit on the number of amplifier chains. Most modern submarine cables carry between 8 and 24 fibre pairs.
Summary
Submarine cable capacity is being upgraded at every layer - coherent optics, wavelength-division multiplexing, fibre count and cable design - to keep pace with AI, cloud and content-distribution traffic. Anyone reading the headlines should keep three things in mind. The "single-wave" figure typically lies in the 800 Gb/s to 1.6 Tb/s range, not higher. The cable, repeaters and fibre type set hard limits on how much terminal-equipment upgrades can deliver. And from a manufacturing standpoint, fibre selection, mechanical protection and rigorous testing remain decisive for whether a cable can carry tomorrow's traffic safely for its full design life.
For specification details, fibre options or project-specific submarine cable design questions, contact our engineering team through the Hengtong contact page.