How Undersea Internet Cables Work: Repeaters, Power and Repair

An undersea internet cable is a garden hose sized bundle of glass fiber pairs, wrapped around a copper power conductor and steel, that carries light between two landing stations. Every 70 km or so a sealed optical repeater boosts the signal, powered by a constant current of about one ampere pushed down the copper from shore at thousands of volts. Cable ships lay it, plough it into the seabed in shallow water, and come back with grapnels to haul it up and splice it on deck when it breaks.

Below, the system is taken apart the way an engineer would: the cable itself, repeaters and their power, fiber pair counts, survey and laying, the landing, repair, US licensing and monitoring. Fault statistics live in two companion posts, how often undersea internet cables break and undersea cable cuts and internet resilience.

Key takeaways

• A modern submarine cable is a system, not just a cable. ITU-T Recommendation G.977 splits it into a land portion (terminal station to beach joint) and a submarine portion (the cable plus repeaters, branching units and equalizers on the seabed).
• Deep water cable is thin, about 17 to 21 mm across, because nothing much threatens it below about 1,500 m. Closer to shore, steel wire armour takes it to 40 to 50 mm, and it is usually buried.
• Repeaters are erbium doped fiber amplifiers in pressure housings. Spacing of about 70 km is typical; Meta says a 7,000 km cable "typically needs about a hundred repeaters".
• Power comes from shore. Power feeding equipment (PFE) in each landing station drives a regulated constant current (NEC describes 1.1 A) through the copper conductor, with the sea as the return path. NEC estimated about 11 kV for a 9,000 km, four fiber pair system; Meta says its newest design stays within PFE rated up to 18 kV.
• Space division multiplexing (SDM) adds fiber pairs instead of pushing each pair harder, and shares pump lasers across pairs. Fiber pair counts have gone from 8 (Marea) to 12 (Dunant), 16 (2Africa, Amitié) and 24 (Anjana), and Meta's planned Petal cable uses 2-core fiber in 24 pairs.
• Repairs are slow, careful work. A ship locates the fault, grapples the cable, cuts and recovers both ends, and joins in a spare section. Each fibre joint takes 10 to 24 hours according to the ICPC and UNEP-WCMC report.
• In the United States, a landing needs a licence under the Cable Landing License Act of 1921. New FCC rules adopted on 7 August 2025 took effect on 26 November 2025, adding foreign adversary restrictions and a cybersecurity and physical security risk management plan certification.

The cable system at a glance

It helps to fix the vocabulary first, because the industry uses a few terms that are not obvious.

ITU-T G.977, the standard for optically amplified submarine cable systems (current edition 01/2015, in force), defines the system between two kinds of points. Point A is the system interface in the terminal station, where the submarine system hands traffic to terrestrial networks. Point B is the beach joint or landing point, where the marine cable meets the land cable. Everything between the two B points is the submarine portion. Everything between A and B on each end is the land portion.

Engineers also talk about "wet plant" (anything in the sea) and "dry plant" (the equipment in the station). The wet plant is designed to run for decades without anyone touching it. The dry plant is where most of the upgrades happen.

The cable contains one or more optical fibre pairs, and G.977 is explicit about why they come in pairs: one fiber carries each direction. When a press release says a cable has 16 fiber pairs, it means 32 strands of glass, 16 going each way.

What is inside a submarine cable at shallow and deep depths?

The core is the same everywhere

Strip the armour off any modern repeatered telecom cable and you find the same basic core. The 2009 report Submarine cables and the oceans: connecting the world, written for the International Cable Protection Committee (ICPC) and the UN Environment Programme World Conservation Monitoring Centre (UNEP-WCMC), describes it from the inside out:

1. Optical fibres, the hair thin strands that carry light, held in a protective tube filled with thixotropic jelly so water cannot run along the cable if it is cut.
2. A strand of steel wires around that tube, which gives the cable its tensile strength so it can hang several kilometres down from a ship during laying and recovery.
3. A copper conductor, described in the report as the layer that powers the repeaters or amplifiers that process the light signal.
4. Polyethylene insulation, which isolates the conductor at high voltage from seawater. Polyethylene, discovered in 1933 according to the report, is what made long coaxial telephone cables practical, and it is still the dielectric today.

That core is the lightweight (LW) cable used in deep water. The report's life size drawing labels LW cable at about 17 mm across, and the text puts deep ocean cable at about 17 to 21 mm, "about the size of a domestic garden hose". TeleGeography's public FAQ makes the same comparison.

Armour is added as the water gets shallower

The main threats to a cable, according to the ICPC and UNEP-WCMC report, sit in water shallower than about 1,500 m, which it describes as the present limit of most bottom trawl fishing (with some boats extending to 2,000 m). Anchors, trawl doors and wave driven abrasion all live there. So cable designers add protection in steps as the route climbs toward shore.

G.977 lists the mechanical families: lightweight, lightweight protected, light armoured, single armoured, double armoured and rock armoured. The ICPC and UNEP-WCMC drawing shows them at roughly these sizes:

The diameters come from the report's figure, which lists representative cables rather than one manufacturer's catalogue, so treat them as typical rather than exact. The text of the same report gives the general rule: in water shallower than about 1,500 m, protective wire armouring can take a cable to 40 to 50 mm across.

Armour is galvanised steel wire wound around the core, one or two layers, held in place with layers of polypropylene yarn and compound. Each layer adds weight, cost and stiffness, which is why designers use the lightest type that the seabed and the local fishing allow.

Why the deep sea cable can be so thin

Most of the length is the thinnest cable because the deep ocean has almost no trawlers or anchors. A lighter cable lets a ship carry more per voyage and costs less per kilometre. The trade-off is reach: repairs in 5,000 m of water take longer and need more slack than repairs on the shelf.

A short history, as a dated timeline

The design above did not appear at once. Each generation solved a failure of the last.

The analogue coaxial era also explains why repeaters exist at all. The ICPC and UNEP-WCMC report notes that the highest capacity analogue systems needed repeaters every 6 to 9 km. Optical amplifiers stretched that to tens of kilometres.

How do optical repeaters work?

Amplifying light without converting it

Early fibre systems used electronic regenerators: they turned light into an electrical signal, cleaned it up, and turned it back into light. Modern systems do not. G.977 clause 8.3 says the optical fibre amplifiers (OFAs) in a submarine repeater use erbium doped fibre (EDF). A short length of glass fibre doped with the rare earth element erbium is spliced into the signal path, and pump lasers inside the repeater excite the erbium ions. When the weak signal light passes through, the excited ions release energy at the same wavelength, and the signal comes out stronger. Every wavelength in the band is amplified at once, which is why one repeater can serve dozens of wavelengths per fiber with no knowledge of what data they carry.

G.977 adds a few details that matter in practice:

• The EDF may be pumped in the same direction as the signal, against it, or both, by one or several redundant pump lasers. Redundancy matters because a pump laser is the part most likely to age.
• Optical isolators stop reflections from bouncing back through the amplifier.
• Automatic power control can regulate output power or pump power.
• Supervisory facilities must let operators monitor the status and performance of the amplifiers remotely. The repeater reports on itself.

Inside one repeater housing, G.977 says, the internal unit can hold several power feed modules and several OFA pairs, amplifying both directions of one or more fibre pairs. An OFA pair is one amplifier for each direction of one fiber pair.

The housing is a pressure vessel

The amplifier electronics are the easy part. The hard part is keeping them alive for 25 years at the bottom of the sea. G.977 clause 8.1 lists what a repeater housing must handle:

• Operation, laying, recovery and re-laying at large depths with no degradation.
• Large load transfer from the cable through a flexible coupling, because the repeater is pulled through the ship's cable machinery and hangs from the cable during recovery.
• No corrosion from seawater.
• Large water pressure. At 6,000 m, the pressure is roughly 600 times atmospheric.
• High voltage insulation between the housing and the internal unit.
• Heat dissipation by conduction through the housing, since there is no fan and no air.
• Sealing against water and gas, including water that might run along the cable core after a break close to the repeater.
• A controlled internal atmosphere for humidity and any gas the components may produce.

The ICPC and UNEP-WCMC report includes a scale drawing of representative repeaters with a bend limiter and a flexible coupling at each end. The coupling is what lets a rigid steel cylinder pass over the ship's sheaves without kinking the cable.

How far apart are repeaters?

Repeater spacing is a design choice, not a constant. NEC's engineers explain the trade-off in their 2010 paper on repeatered transoceanic construction: shorter spacing improves the optical signal to noise ratio, but adds repeaters and cost, so designers "select the longest possible repeater interval" that still meets the performance target with manufacturing and repair margins.

The figures in primary sources line up well:

• The ICPC and UNEP-WCMC report gives a typical spacing of 70 km for erbium amplifier repeaters.
• Meta's September 2026 Petal announcement says a 7,000 km cable needs about a hundred repeaters, which is the same 70 km average.

Short cables can skip repeaters entirely. ITU-T G.973 (current edition 11/2016) covers repeaterless optical fibre submarine cable systems, where all amplification happens in the terminal stations, sometimes helped by remote pumping. These are common for island links and coastal festoons.

Equalizers and dispersion management

Erbium amplifiers do not amplify every wavelength by exactly the same amount. After a hundred repeaters, a tiny ripple becomes a large one. G.977 describes optical submarine equalizers that correct spectral ripple with shape equalizers and gain tilt with slope equalizers. NEC's paper explains that shape and tilt equalizers go in after an optimum number of repeaters, and that in one 9,000 km design the accumulated dispersion was corrected at adjustment sections placed after every group of ten repeater spans. The NEC engineers built those corrections from measurements of the actual manufactured cable and repeaters, adjusting the length of each section to hit the design.

How are repeaters powered from shore?

This is the part that surprises most people: there are no batteries on the seabed. G.977 clause 8.2.1 says repeaters are powered from the terminal station at a constant current through the electrical conductor in the cable. The repeaters sit in series on one copper conductor, like old Christmas lights, and the same current flows through every one of them.

The return path is the ocean. The far end of the circuit connects to a sea earth: G.977 says a branching unit has an attached sea earth, and the Los Angeles landing project described later includes ocean ground beds at the beach. The seawater and seabed between the two earth points carry the return current.

NEC's 2010 paper gives the clearest numbers in any vendor document:

• The repeaters run on a regulated current of 1.1 A from the PFE.
• The cable's conductor resistance is about 0.8 Ω per km.
• Voltage is also lost across each repeater and to differences in earth potential between the two ends, which NEC puts at about 0 to 0.3 V/km depending on region, and to spare repeaters and spare cable inserted for repairs.
• Adding it up, NEC estimates a feeding voltage of about 11 kV for a 9,000 km system with four fiber pairs.
• Power is fed from both ends. One PFE supplies positive voltage and the other negative, so each supplies half the total. If one PFE must be serviced, the other can feed the whole system alone.

Meta's Petal announcement adds the current ceiling: Petal is designed to stay within existing power feeding equipment limits, which Meta says are rated up to 18 kV, which avoids requalifying the subsea equipment for higher voltages. That sentence tells you something important about modern design. Electrical power, not fiber count or laser speed, is the binding limit on how much a cable can carry.

Worked example 1: the voltage budget of a 9,000 km cable

Take NEC's figures and check them with Ohm's law.

• Conductor resistance: 9,000 km × 0.8 Ω/km = 7,200 Ω.
• Voltage dropped in the copper alone at 1.1 A: 7,200 Ω × 1.1 A = 7,920 V.
• NEC's total for the system: about 11,000 V.
• What is left for everything else: 11,000 − 7,920 = 3,080 V.

That remaining 3 kV or so covers the drop across every repeater's power module, the earth potential difference (up to 0.3 V/km × 9,000 km = 2,700 V in the worst regions NEC mentions), and margin for repair sections added over 25 years. With dual end feeding, each station supplies about 5,500 V, one positive and one negative, and the point of zero volts sits somewhere mid ocean.

The total electrical power is modest by data centre standards: 11,000 V × 1.1 A is about 12 kW for the entire wet plant, every amplifier across 9,000 km. That is why each fiber pair added has to earn its keep in watts.

Worked example 2: counting amplifiers on a petabit cable

Meta describes Petal as 24 fiber pairs of 2-core fiber, which it calls equivalent to 48 fiber pairs, over about 7,000 km with about a hundred repeaters. Its repeater is a single body design with 96 amplifiers.

• 24 fiber pairs = 48 physical fibers.
• Each 2-core fiber carries 2 cores, so 48 × 2 = 96 cores.
• Each core carries one direction and needs one amplifier per repeater: 96 amplifiers per repeater, which matches Meta's figure.
• About 100 repeaters × 96 = roughly 9,600 amplifier stages in the cable, all fed by the same constant current within an 18 kV PFE limit.

What happens electrically when the cable breaks

A cut cable does not simply go dark. If the copper touches seawater, the current finds a new path to earth at the break, which G.977 calls a shunt fault: "a current leakage path between the power conductor and the sea". The PFE's job then becomes keeping as much of the cable powered as possible. G.977 requires every repeater and branching unit to be protected against surges caused by a sudden loss of the high voltage supply, whether from a cable break, a shunt fault or a short circuit in the PFE.

Branching units help here. G.977 clause 9.3.2 says the BU can connect any two incoming power conductors together while the third cable is switched to the BU's own sea electrode, and that the switching should let operators restore traffic on the healthy segments during a fault or a repair. In plain terms, the BU can reroute power so that one broken branch does not take down the whole trunk.

What is space division multiplexing, and how many fiber pairs do new cables carry?

The problem SDM solves

For two decades, cable capacity grew mostly by sending more wavelengths down each fiber (wavelength division multiplexing) and by using better modulation and coherent receivers in the SLTE. Physics eventually caps what one fiber can carry over thousands of kilometres at a given optical power, and the shore end caps the total electrical power. Meta's Petal post puts it simply: to get past this limit, the industry moved to spatial division multiplexing (SDM) and put more fibers in each cable.

SDM means using more spatial paths, which in practice means more fiber pairs (and, next, more cores per fiber), each run at a slightly lower and more power efficient operating point. Google's Dunant announcement describes the second half of the idea: earlier cables gave each fiber pair its own dedicated pump lasers, while in Dunant's SDM design the pump lasers and related optics are "shared among multiple fiber pairs". Sharing pumps saves power and parts, so more pairs fit inside the same power budget.

Fiber pair counts, from primary sources

The 2Africa announcement in 2020 said it doubled the eight fiber pair maximum of older technology, and Google described Dunant's twelve pairs as a step up from the six or eight of past generations.

Worked example 3: capacity per fiber pair goes down, total goes up

Divide each cable's headline capacity by its pair count:

• Marea: 224 Tbps ÷ 8 pairs = 28 Tbps per pair.
• Dunant: 250 Tbps ÷ 12 pairs ≈ 20.8 Tbps per pair.
• Anjana: 500 Tbps ÷ 24 pairs ≈ 20.8 Tbps per pair.
• Petal: 1,000 Tbps ÷ 48 equivalent pairs ≈ 20.8 Tbps per pair.

The pattern is the point of SDM. Per pair throughput fell from Marea to Dunant, then held flat, while total capacity quadrupled. Designers chose to run each pair less aggressively and add more pairs, because a lower power operating point per pair lets many more pairs fit within the PFE's voltage limit. (Headline capacities are design or potential figures. TeleGeography's FAQ explains that potential capacity assumes all transmission equipment is installed, while lit capacity is what is actually running.)

Where SDM goes next

Meta's Petal post names three ways to double again: 48 fiber pairs, adding the L-band to the C-band, or multi-core fiber. Petal takes the third route with 2-core fiber, splitting each fiber's cores to ordinary single-core amplifiers inside every repeater. Petal is planned, not built, so treat its numbers as design targets.

How is a cable system built, from raw material to service?

NEC's paper says the period from design to completion is about one year for a small system and about two years for a large one, and stresses that the process runs continuously from manufacture to laying. This table follows the order NEC and the ICPC and UNEP-WCMC report describe.

G.977 sets the reliability target that all of this is aiming at: fewer than three component failures needing a cable ship over the designed lifetime, which is usually 25 years from provisional acceptance. That target applies to component failures inside the system. External damage from anchors and trawls is a separate problem, which is why routing and burial get so much attention.

How is a route surveyed and a cable laid by a cable ship?

The desktop study and route survey

According to the ICPC and UNEP-WCMC report, work starts with a desktop study (DTS): everything already known about the seabed, the fishing grounds, the shipping lanes, other cables and pipelines, protected areas and permit requirements. The DTS includes visits to each landing to see where the cable will cross the beach and how it will reach the station, and to meet local officials about hazards and permits.

The marine survey then confirms the route. The report describes these tools and limits:

• Multibeam echosounders map a swath of seabed three to five times as wide as the water is deep, so in deep water one pass covers a very wide strip. The report's figure shows a survey ship mowing the lawn, running parallel lines across a swath that can be up to 20 km wide.
• The survey corridor is commonly about 1 km wide in depths down to about 1,500 m, because that is where cables usually need burial. In deeper water, a single track of multibeam bathymetry generally suffices.
• Seismic sub-bottom profilers measure the type and thickness of sediment under the seabed wherever burial is planned, and look for natural hazards.
• Sediment coring and geotechnical tests check whether the seabed is stable and suitable for a plough.

The aim is to avoid fault zones, steep slopes, rock outcrops, fishing and anchoring areas, and other infrastructure. TeleGeography notes that operators take considerable care to keep cables on the safest path. For how the Earth's structure and seabed processes shape that choice, see our explainer on Earth's layers.

Laying: speed, slack and tension

A cable ship lays cable by paying it out over the stern through cable engines that control tension and speed. The ICPC and UNEP-WCMC report reduces the problem to three key parameters: the ship's speed over the ground, the speed at which cable is paid out, and the water depth. Pay out slightly more cable than the distance travelled and the cable drapes into the seabed's contours instead of hanging between peaks. Pay out too much and you waste cable and leave loops that can snag.

The report says a ship can lay at up to about 11 to 15 km per hour (6 to 8 knots), slowing down from time to time to pass repeaters through the cable handling machinery. In deep water, it notes, as much as 6 km of cable can be suspended between ship and seabed during laying, which is why the strength member matters as much as the glass.

NEC's paper adds the quality step that is easy to forget: during laying, the crew periodically measures insulation resistance, voltage drop, the optical fibre trace and optical signal to noise ratio, to confirm the cable is not being damaged as it goes over the side.

Burial with ploughs, jets and ROVs

In water shallower than about 1,500 m, the report says, modern cables are usually buried as protection against human activity. It calls the sea plough the most effective burial method. The plough sits on skids on the seabed, towed by the cable ship. The cable is fed through the plough, which places it in a narrow furrow cut by the ploughshare. Plough designs vary: a traditional ploughshare suits mud, while sandy sediments may need a water jet to cut the trench.

Burial has limits:

• Where the route crosses another cable or a pipeline, the plough must be recovered or flown over the crossing and set down again on the other side.
• Those skipped sections are buried later by divers or by a remotely operated vehicle (ROV) with jetting or trenching tools.
• Some ground cannot be buried at all, such as bare rock or highly mobile sand, so armour and route choice have to do the work instead.

How deep? It depends on the seabed and the threat. Knowing which trawl gear is used in an area lets engineers pick a burial depth, the report says. One concrete current example: the City of Los Angeles notice for a trans-Pacific landing at Dockweiler Beach says the cables would be buried about 3.3 feet below the seabed where feasible, in water shallower than 3,600 feet, and laid on the seafloor deeper than that. That is about 1 m of cover out to about 1,100 m of water.

Which cable type goes where: a decision table

What happens at a landing station and in the beach manhole?

The shore end

The cable has to get from the seabed, through the surf zone, across a beach that people use, and into a building, without being exposed to waves or excavators. Modern landings usually do this with horizontal directional drilling (HDD): a steel bore pipe is drilled from a pit behind the beach out under the seabed, and the marine cable is pulled through it to land.

The City of Los Angeles project notice for the Dockweiler Beach trans-Pacific landing lists exactly what that involves:

• Four offshore bore pipes drilled by HDD, extending about 4,000 feet (about 1.2 km) offshore.
• Two beach manholes (BMHs) about 400 feet landward of the mean high water line, where the HDD bores end.
• Up to four ocean ground beds (OGBs) at the beach.
• One HDD bore from a beach manhole to a nearby road intersection, then about four miles of open trenched terrestrial conduit to the data centre.
• A phased plan: the landing infrastructure is built first and the cables are installed later, depending on planning, approvals and ship availability.

The beach manhole is an underground concrete vault where the marine cable ends and the land cable begins: the beach joint, or point B in G.977's terms. The two cables are joined there in a splice enclosure. The ocean ground bed is the sea earth electrode for the power circuit, buried near the shore so that the return current reaches the ocean through a low resistance path.

G.977 Annex A notes a safety point that drives a lot of shore end design: the land cable carries the repeater and BU feeding current, so a high potential difference can exist between the conductor and ground, and personnel need protection. Several kilovolts are present on the conductor that runs under a public beach, so access and work procedures are tightly controlled.

Inside the landing station

The landing station (or cable landing station, CLS) is where the dry plant lives. It typically holds:

• Power feeding equipment, redundant, at both ends of the system as NEC describes, with the high voltage output going to the land cable.
• SLTE, the transponders and line equipment that light each fiber pair. The FCC's 2025 order treats SLTE as a significant component of the cable system itself, and in many modern systems more than one party owns SLTE on different fiber pairs.
• Line monitoring equipment, including the coherent optical time domain reflectometry (C-OTDR) gear that measures each fibre span.
• The element management system for the submarine plant. NEC describes EMS server and clients connected to a router in the station, linked to a remote network operations centre.
• Backhaul, the terrestrial fiber that carries traffic onward, often to a data centre or a carrier hotel inland. In the Los Angeles example, the conduit runs about four miles to a data centre.

For how landings and new cables have changed connectivity across one continent, see our post on Africa's subsea cables. If you are interested in the terrestrial side of the same chain, our guide to building your own internet service covers transit, peering and backhaul.

How are undersea cables repaired?

Fault frequency and causes are covered in detail in how often undersea internet cables break. Briefly, TeleGeography cites ICPC data for "roughly 200 faults each year", most caused by fishing and anchors. This section is about the repair itself.

Step 1: find the fault from shore

Before a ship sails, the operator narrows down where the cable is broken.

• Optical measurements. G.977 says C-OTDR is the method used especially for fault location in long distance amplified systems. Each repeater includes a loopback path so a reflectometry signal sent from shore can return, letting operators see which span has lost light. G.977 notes the COTDR path should not disturb in service traffic and can monitor bidirectionally.
• Electrical measurements. Because a shunt fault changes the resistance and voltage of the power circuit, PFE readings and electrode tests at both ends give an independent estimate of how far out the conductor touches seawater.
• Repeater supervisory data. Repeaters report their own status, so a failing pump laser can be seen before it becomes an outage.

Step 2: mobilise a ship

According to the ICPC and UNEP-WCMC report, cable ships are kept on standby around the world, ready to sail at short notice. Many owners pay into maintenance arrangements that keep a ship and a stock of spare cable and repeaters ready for a region. The report stresses that the correct spare cable and fibre type must be used, so the repair authority has to carry a comprehensive stock. ICPC Recommendation 4 covers coordination procedures for repair operations near in service cable systems, so that a repair does not damage a neighbour.

Step 3: grapple and recover

The report sets out the recovery sequence: locate the cable and the faulted section, retrieve it with purpose designed grapnels deployed from the repair vessel, and lift it to the surface. On a buried cable, the grapnel may have to drag it out from 1 to 3 m below the seabed, and the strain is substantial. Recovery speed and angle are limited by the manufacturer's recovery tension tables for each cable type and depth. Biological growth on an old cable can add drag.

A common approach is to cut the cable, seal and buoy one end, and recover the other end first. Each end is tested from the ship back toward shore to prove that the remaining cable is healthy before anything is joined.

Step 4: splice on deck

The report gives the core of the job in one line: a repair usually means cutting out the damaged section and jointing in a replacement. In the telegraph era a joint took 3 to 6 hours. A fibre-optic joint now takes 10 to 24 hours of specialised work: every fibre must be aligned and fusion spliced, tested, then enclosed in a mechanical joint that restores the cable's strength, the conductor's continuity and the insulation.

A repair needs at least two joints: one end of the cable to the spare section, then the spare section to the other end. If the fault was a failed repeater, a spare repeater goes in with the spare cable.

Step 5: lay back and restore

Because the final joint is made on deck, the repaired cable is longer than the original route. The extra length is laid back on the seabed as a loop beside the original line. On a buried section, the repaired cable is reburied with an ROV. The new length and joints are added to the system's records, and the power and optical budget designed into the system at day one (NEC calls it the repair margin) absorbs the extra loss.

Repair checklist at a glance

Who regulates landing licences?

United States: the Cable Landing License Act and the FCC

In the US, the legal basis is short and old. Section 34 of Title 47 of the US Code, enacted on 27 May 1921, says no one may land or operate a submarine cable that connects the United States with a foreign country, or one part of the United States with another, without a written licence issued by the President. Cables lying wholly within the continental United States are exempt.

The FCC's guide explains how that works in practice. Executive Order 10530 of 10 May 1954 delegates the President's authority to the FCC, subject to approval from the Secretary of State and advice from other agencies. A licence is needed to land a cable connecting the continental US with a foreign country; Alaska, Hawaii or US territories with a foreign country, the mainland or each other; and points within the US where the cable runs through international waters. The FCC's guide still describes rules adopted in 2001 and contained in section 1.767, which have now been replaced.

The 2025 rules

Early planning notes for this post pointed to 2024 rules. The primary record shows a 2024 notice of proposed rulemaking followed by a 2025 order. The FCC's Office of International Affairs (DA 25-934, released 28 October 2025) confirms the dates:

• The FCC adopted the Submarine Cable Landing License Report and Order and Further Notice of Proposed Rulemaking on 7 August 2025 (FCC 25-49, released 13 August 2025, OI Docket No. 24-523).
• Both were published in the Federal Register on 27 October 2025 (90 Fed. Reg. 48648 for the order).
• Most of the new rules, codified in a new subpart beginning at 47 CFR 1.70000, took effect on 26 November 2025. Some information collection provisions waited on separate approval.

According to the Federal Register summary, the order:

• Updates the licensing process and adopts rules against risks from entities owned or controlled by a foreign adversary, or subject to its jurisdiction or direction, including a presumption of denial for certain such applicants.
• Requires certain licensees to file an annual report on the licensee, ownership and operations.
• Requires applicants and licensees to certify that they have "created, updated, and implemented a cybersecurity and physical security risk management plan".
• Requires applicants to certify that the system will not use equipment or services on the FCC's Covered List.
• Adds a one time information collection to identify how many entities own or operate SLTE on existing licensed cables, and bars licensees from new or extended capacity deals that would let a foreign adversary linked entity install, own or manage SLTE on a US landed cable.
• Keeps the 25-year licence term, after the FCC concluded that a shorter term could harm the deployment and resilience of submarine cables.
• Streamlines circuit capacity reporting.

Some law firm summaries described the order as asserting FCC authority to license SLTE operators. The Federal Register text is more cautious: the FCC declined, for now, to require SLTE owners and operators to become licensees, and used the one time information collection instead. The Federal Register text wins.

Team Telecom

Security review runs alongside the FCC process. The Department of Justice describes the Committee for the Assessment of Foreign Participation in the US Telecommunications Services Sector, "known as Team Telecom", formalised by Executive Order 13913 in 2020. Its members are the Departments of Justice, Defense and Homeland Security, with other agencies as advisers. Team Telecom can tell the FCC an application raises no risk, should be granted with mitigation conditions, or should be denied; it can also re-review existing licences. The 2025 order confirms that EO 13913 continues to apply when the FCC refers an application.

Other countries, summarised

Outside the US, the pattern is similar even though the instruments differ:

• International law sets the frame. The ICPC and UNEP-WCMC report explains that the 1884 Convention for the Protection of Submarine Telegraph Cables remains the only treaty with detailed procedures (cable ship lights and shapes, minimum distances, claims for sacrificed anchors and gear), and that UNCLOS Article 311(2) recognises its continued use. Coastal states control activity in their territorial seas and regulate to varying degrees in their exclusive economic zones.
• Protection zones. Australia's Telecommunications and Other Legislation Amendment (Protection of Submarine Cables and Other Measures) Act 2005, cited in the same report, created zones where activities like trawling and anchoring are restricted near critical cables.

ICPC itself is not a regulator. It describes itself as a body founded in 1958 that helps members "improve the security of undersea cables" through shared technical, legal and environmental information, with more than 255 members from 77 countries or territories as of September 2026.

Security and monitoring of cable systems

Physical protection comes first: careful routing, armour, burial, charting cables so fishers and ship masters know where they are, and protection zones where governments create them. Monitoring comes next. Operators watch PFE voltage and current, repeater telemetry and C-OTDR traces continuously, and some also watch ship traffic near their routes so they can warn a vessel before an anchor drags. NEC's paper shows the shape of the monitoring network: EMS servers and clients in each station on a router, connected to a remote monitoring centre.

That monitoring layer is software. The SLTE, the EMS and network management system, the PFE controllers and the operator's own tooling are built from operating systems, libraries and containers like any other enterprise software, and the FCC's 2025 order now asks every applicant to certify a cybersecurity and physical security risk management plan. This is where Safeguard's fit is real but indirect. Safeguard does not monitor cables, inspect repeaters, scan firmware or analyse hardware. What it does is generate and analyse SBOMs and scan the source repositories, container images and SBOM files of the software that teams build and run, match components against vulnerability databases, show which projects carry a given component and version, rank findings with reachability analysis, and open fix pull requests when a safe version exists (autonomous remediation ships switched off by default). If you build or operate management software for network infrastructure, that is the part of the risk plan it can help with.

Walkthrough: see the physics from your own terminal

You cannot touch a submarine cable, but you can see its effects and do the engineering arithmetic yourself. Nothing here sends traffic anywhere unusual or changes any system.

Step 1: measure a trans-oceanic round trip

Pick a host you know is on another continent. Run a short trace that reports per hop latency:

# macOS: brew install mtr   Debian/Ubuntu: sudo apt install mtr-tiny
mtr --report --report-wide --report-cycles 20 www.example.org

# Or, without mtr:
traceroute -q 3 www.example.org     # Linux and macOS
tracert www.example.org             # Windows

Look for the hop where latency jumps by tens of milliseconds in one step. That jump is usually a long haul link, often a submarine cable.

Step 2: compare with the speed of light in fiber

Light travels through silica fiber at roughly two thirds of its speed in vacuum, a common rule of thumb is about 5 microseconds per kilometre. A short script turns cable length into a minimum round trip time:

# rtt_floor.py: minimum round trip time over fiber, ignoring routing and queuing
C_VACUUM_KM_S = 299_792.458
GROUP_INDEX = 1.47            # typical value for standard single-mode fiber (approximation)

def rtt_ms(route_km: float) -> float:
    one_way_s = route_km / (C_VACUUM_KM_S / GROUP_INDEX)
    return 2 * one_way_s * 1000

for name, km in [("Ireland to UK, CeltixConnect, 131 km", 131),
                 ("France to US, Petal design, about 7,000 km", 7000),
                 ("Asia America Gateway, 20,000 km system length", 20000)]:
    print(f"{name}: at least {rtt_ms(km):.1f} ms round trip")

Run it with python3 rtt_floor.py. The 7,000 km route gives a floor of about 69 ms. If your measured trans-Atlantic RTT is 75 to 90 ms, the difference is terrestrial backhaul, routing detours and equipment. (Cable lengths are from TeleGeography's FAQ and Meta's Petal announcement. The 20,000 km figure is the total system length of a cable with many branches, not a single path, so it overstates a real trip.)

Step 3: check a power budget

Use NEC's published numbers to see why PFE voltage limits the design:

# pfe_budget.py: series powered repeaters, figures from NEC Technical Journal Vol.5 No.1 (2010)
CURRENT_A = 1.1          # regulated line current
R_OHM_PER_KM = 0.8       # cable conductor resistance
EARTH_V_PER_KM = 0.3     # worst case earth potential gradient NEC mentions

def conductor_drop_v(km: float) -> float:
    return km * R_OHM_PER_KM * CURRENT_A

for km in (2000, 6000, 9000):
    copper = conductor_drop_v(km)
    earth = km * EARTH_V_PER_KM
    print(f"{km:>5} km: copper drop {copper/1000:.1f} kV, earth potential up to {earth/1000:.1f} kV, "
          f"before repeater drops and repair margin")

At 9,000 km the copper alone needs 7.9 kV. Add repeaters, earth potential and repair margin and you reach NEC's 11 kV. The calculation shows why designers care about every watt per amplifier, and why Meta designs Petal to stay inside an 18 kV PFE rating.

Common mistakes

• Thinking the internet runs mostly over satellites. The ICPC and UNEP-WCMC report put submarine fiber at well over 95 per cent of international traffic in 2009, and TeleGeography cites FCC data showing satellites at 0.37 per cent of US international capacity.
• Assuming the thick armoured cable you see in photographs runs across the ocean. Most of the length is 17 to 21 mm lightweight cable. The thick sections are near shore.
• Treating a repeater as a regenerator. Modern repeaters amplify light with erbium doped fiber. They do not decode or regenerate the data, which is why the wet plant can stay in place while SLTE upgrades raise capacity.
• Forgetting that the cable is a high voltage circuit. Thousands of volts sit on the conductor, including the land section under the beach. G.977 calls out personnel protection for exactly this reason.
• Citing outdated rules. The FCC's own guide page still describes the 2001 rules. The 2025 order replaced them, effective 26 November 2025 for most provisions.
• Ignoring the software layer. The EMS, network management and SLTE software are now part of a licensee's certified risk plan. Teams that build that software should treat its dependencies like any other production system.

FAQ

How thick is an undersea internet cable?

In deep water, about 17 to 21 mm across, roughly the width of a garden hose, according to the ICPC and UNEP-WCMC report and TeleGeography. Near shore, steel armour increases it to about 40 to 50 mm, and some rock armoured types are thicker still.

How are undersea cables powered?

From shore. Power feeding equipment in the landing stations sends a regulated constant current, about 1.1 A in NEC's example, through a copper conductor inside the cable, with the sea as the return path. Voltages are in the kilovolts: about 11 kV for a 9,000 km four pair system in NEC's paper, and up to an 18 kV equipment rating for Meta's planned Petal cable. Feeding from both ends splits the voltage between a positive and a negative supply.

How far apart are repeaters on a submarine cable?

Typically about 70 km. The ICPC and UNEP-WCMC report gives 70 km as typical for erbium amplifier repeaters, and Meta says a 7,000 km cable needs about a hundred repeaters. Exact spacing is a design choice traded against signal quality, cost and power.

What is inside a submarine cable repeater?

Erbium doped fiber amplifiers for each fiber in each direction, redundant pump lasers, optical isolators, power modules that draw from the line current, surge protection and supervisory electronics, all inside a sealed, pressure resistant, corrosion resistant housing with high voltage insulation, according to ITU-T G.977.

How many fiber pairs does a modern subsea cable have?

New systems commonly have 12 to 24. Dunant has 12, 2Africa up to 16, Amitié 16 and Anjana 24. Meta's planned Petal uses 24 pairs of 2-core fiber, which it calls equivalent to 48 pairs. Older designs topped out at about eight.

How deep are undersea cables buried?

In shallow water, often around 1 m below the seabed where the ground allows. The Los Angeles Dockweiler Beach project, for example, plans burial of about 3.3 feet (1 m) out to 3,600 feet of water depth. Below about 1,500 m, cables are generally laid on the seabed surface.

How long does it take to repair an undersea cable?

The splicing alone takes 10 to 24 hours per fibre joint, and a repair needs at least two joints. Total time depends on how fast a ship can reach the site, the weather, the depth and any permits needed in territorial waters. Our post on how often cables break covers typical repair durations and causes.

Do you need a licence to land a submarine cable in the US?

Yes. Under the Cable Landing License Act of 1921 and Executive Order 10530, the FCC issues cable landing licences, with advice from the State Department and Team Telecom. New rules adopted on 7 August 2025 took effect on 26 November 2025.

Why software supply chain hygiene still matters here

The cable on the seabed is fixed by ships, but the software that runs a cable system is maintained like any other codebase: management systems, SLTE software and operator tools ship open source libraries and container images that need patching on their own schedule. The fit here is indirect, and it applies only to the software you build and run, not to the wet plant.

Fix it without waiting for upstream

Whatever the vulnerability, most of the risk sits in two gaps. The first is the time between a fix being published and that fix actually running in your production builds. The second, and worse, is the time before any fix exists at all, while a maintainer works out a patch and a release schedule. Waiting through both gaps is a habit, not a requirement. Safeguard gives teams three ways to close them.

1. Autonomous remediation: pull requests, not tickets

When a fixed version exists, Safeguard's Griffin models open a tested, policy gated pull request that moves the affected dependency to the safe version. You can run it on one repository or in bulk across every project that carries the vulnerable version. Remediation covers eight categories, including dependencies, licenses, security posture and findings. Reachability analysis ranks the vulnerable code paths your application actually calls first, so the first pull requests you merge are the ones that matter most. Autonomous remediation ships switched off, with every category set to Manual, so your team decides how much runs on its own and what still needs a human approval.

2. Gold Open Source: the safe version in seconds, free

gold.safeguard.sh is free and needs no login. Search any CVE or package to see affected and fixed versions, CVSS, EPSS, CISA KEV status and a recommended safe version, or use the "Am I affected?" checker. The same data is available through a free read API, README status badges and RSS threat feeds. The safeguard-sh/gold-scan-action GitHub Action can fail a build on a known exploited vulnerability, on any known vulnerability, or never, so a version you have just removed cannot quietly come back in next week's merge.

3. Gold Registry: start from components that are already fixed

registry.safeguard.sh provides hardened, drop in replacements for upstream packages and container images. Every Gold artifact is held continuously to zero exploitable critical or high CVEs, zero malware and a completed license check, with SLSA L3 or higher provenance, a Sigstore signature and a public evidence page you can hand to an auditor. An artifact loses its Gold badge if it regresses. Teams can point their own package directory at the registry so hardened equivalents resolve without changing how developers install anything.

When upstream has no fix yet, or will not ship one, Safeguard first works to get the fix accepted upstream. If the issue is severe or upstream will not move, Safeguard builds a fork, fixes the vulnerability there and publishes a hardened Gold build, so you are not stuck on a maintainer's release schedule. You can also request a hardened, compatibility tested build of a specific package. This fork and patch flow is in early access today, so ask which packages it covers before you plan around it.

One honest limit: when the vulnerable code lives in a vendor appliance, a hosted service or an operating system runtime you do not build yourself, the fix still has to come from that vendor or distribution. The steps above apply to the open source dependencies and container images you build and ship, which is where most teams carry the largest and least tracked share of their exposure.

See how Safeguard would help

None of these demos touch a cable. They show how a team that writes network management or operator software can find and fix vulnerable dependencies in that code.

The first is a real CLI session. You install the CLI, open the interactive terminal with safeguard, and ask it to fix the lodash CVE in a repository; it reads the finding and the advisory, asks before each command, edits package.json to lodash 4.17.21, runs npm install and npm test with your approval, prints a summary and pushes nothing. The same flow applies to a monitoring dashboard or a provisioning tool built on npm.

Click through it at your own pace, or open it full screen.

The second shows the local runner. You set up the runner on your own machine, approve it with its device code, then scan from the terminal. For operators who keep management software on isolated networks, running scans on your own hardware is often the only acceptable option.

Click through it at your own pace, or open it full screen.

The third connects Claude to the Safeguard MCP server with OAuth and lets you choose which tools Claude can use. It suits teams who already ask an assistant questions about their codebase and want vulnerability data in the same conversation, with tool access limited to what they approve.

Click through it at your own pace, or open it full screen.

To try it on your own code, download Safeguard or read the CLI documentation.

Quick check

Sources

• ITU-T Recommendation G.977 (01/2015): Characteristics of optically amplified optical fibre submarine cable systems
• ITU-T Recommendation G.973: Characteristics of repeaterless optical fibre submarine cable systems
• Carter et al., Submarine cables and the oceans: connecting the world (ICPC and UNEP-WCMC, 2009), UN Digital Library
• International Cable Protection Committee: About the ICPC
• International Cable Protection Committee: ICPC Recommendations
• NEC Technical Journal Vol.5 No.1 (2010): Construction Technology for Use in Repeatered Transoceanic Optical Submarine Cable Systems
• Meta Engineering: Inside Petal, building the world's first petabit-class transoceanic subsea cable (21 September 2026)
• Meta Engineering: Building a transformative subsea cable to better connect Africa (13 May 2020)
• Meta Engineering: 2Africa Pearls extension
• Google Cloud blog: The Dunant subsea cable is ready for service (3 February 2021)
• TeleGeography: Submarine Cable Frequently Asked Questions
• FCC: Submarine Cable Landing Licenses guide
• FCC: Submarine Cables
• FCC DA 25-934: Effective date of certain submarine cable license rules (28 October 2025)
• Federal Register, 90 FR 48648 (27 October 2025): Review of Submarine Cable Landing License Rules and Procedures
• 47 U.S. Code § 34: Cable landing license required (Cornell LII)
• US Department of Justice: Team Telecom
• City of Los Angeles Bureau of Engineering: LAX Transpacific Cable Landing project

Last reviewed: 24 October 2026.