Submarine cables carry 99 per cent of intercontinental internet traffic, even as satellite broadband receives most of the attention today. Over 570 cable systems are currently in service, with dozens more being added each year. Investment in the sector is growing just as fast. As per estimates, the submarine cable systems market stood at close to $15 billion in 2025, and is expected to reach around $26 billion by 2031, representing a compound annual growth rate (CAGR) above 10 per cent.
The main forces behind this growth are hyperscaler-funded routes across oceans, offshore wind cables exporting power to shore and government pressure to develop more resilient digital corridors of their own. From enabling early transoceanic communications to powering today’s digital economy, these systems remain among the world’s most critical infrastructure assets, with their role now expanding well beyond telecommunications.
Emerging trends
AI, cloud computing and hyperscalers
Hyperscale cloud companies now lead most of the new subsea cable construction and this shift has happened fast. Before 2012, the major hyperscalers together accounted for less than a tenth of international bandwidth usage. Today, they control 70-75 per cent of global subsea capacity and 90 per cent of capacity on the transatlantic route. This is not simply about buying more bandwidth. These companies are increasingly building, owning and routing cables themselves, on their own schedules, to connect their own data centres.
This growth can be attributed to rising artificial intelligence (AI) demand. Training large models means moving huge volumes of data between clusters of processors, first within a single data centre, then between data centres in the same region and finally increasingly between data centres on different continents. That is a different order of demand from the browsing and streaming traffic that shaped cable design for decades. Subsea bandwidth has grown at roughly 30 per cent a year for some time now, and forecasters, like Nokia, expect AI-driven traffic alone to add over a thousand exabytes a month of extra demand by 2033. In practice, this is pushing cable operators to design networks end to end, from one data centre straight through to another, rather than treating the cable as a separate piece from the terrestrial network on either side. The gains are largely about cost and speed of delivery rather than raw performance.
Regulators are adjusting to this shift too. In June 2026, the US telecom regulator approved its first major overhaul of subsea cable licensing in two decades, tightening security checks while fast-tracking approvals for AI-era builds. Going forward, this combination is likely to become the template for other governments as they try to keep pace with hyperscaler-led demand.
Green energy
Subsea cables are no longer just about data. Part of the appeal is emissions as well as capacity, since high voltage direct current (HVDC) loses far less power than older methods over the same distance. A growing share of new seabed projects carries electricity using HVDC at present. These cables link offshore wind farms to shore, connect national grids for cross-border power trading and help countries balance renewable supply that is often generated far from where it is consumed. In Southeast Asia, this is being organised formally through the ASEAN Power Grid framework, which is built around multi-country subsea power links.
The connection to data cables is not just geographic. Data centres drew around 485 terawatt hours of electricity in 2025, accounting for 1.5-2 per cent of the world’s total, and the International Energy Agency expects that to roughly double to around 950 terawatt hours by 2030, or close to 3 per cent of global demand. That is turning power availability into as important a site-selection factor as bandwidth or latency once was. It is also why grid operators now rely on fibre networks, often the same subsea systems, to monitor and balance increasingly complex and distributed power systems in real time.
Geopolitical shifts
The same ownership shift that is reshaping the AI trend is also a geopolitical fact in its own right. Four major US companies now control most of the world’s international bandwidth, which is one reason European policymakers have started describing their reliance on hyperscaler-owned cables as a strategic vulnerability rather than just a commercial trend. On the construction and repair side, the picture is even more concentrated. Around 98 per cent of the world’s cables are built and maintained by just four firms, American SubCom, French Alcatel Submarine Networks, Japanese NEC and Chinese HMN Technologies. Further, HMN’s share has been growing significantly faster than that of any of its rivals.
Cables have also become a tool of everyday coercion rather than just a subject of trade rivalry. Taiwan’s Matsu Islands have had cables cut repeatedly by suspected Chinese vessels, cutting off residents for weeks at a time. The Baltic Sea has seen a string of similar incidents since 2023, involving Chinese- and Russian-linked ships dragging anchors across cables and pipelines. NATO responded by launching a dedicated Baltic Sentry mission in January 2025, adding frigates, patrol aircraft and naval drones to monitor the area, and by mid-2026 the alliance reported the mission had cut its response time to suspicious incidents from 17 hours to just one. In February, the European Commission announced a Euro 347 million funding in cable security toolbox, its largest ever investment in subsea protection.
The Red Sea tells a similar story. Houthi-linked attacks on shipping have damaged multiple cables since February 2024, most caused by ships dragging their anchors after being struck, and further cuts followed in September 2025. By 2026, wider conflict around the Strait of Hormuz added a second chokepoint on top of the Red Sea one and work on the Gulf extension of the 2Africa cable was paused.
Governments are responding with tighter rules on who gets to build and land cables. However, ownership is concentrating in another sense too. Meta’s Project Waterworth, a single company’s plan spanning around 50,000 km, shows how mega projects are increasingly the work of one owner rather than a shared consortium.
Satellites versus cables
Despite all the attention on satellite broadband, it has not changed the basic economics of global data. Even optimistic projections put total global satellite capacity at around 50 terabits per second by 2026, against roughly 8,750 terabits per second projected for subsea cables, a gap of well over a hundred times. Starlink remains the largest player by far, with thousands of satellites in orbit. Newer entrants such as Amazon’s satellite broadband arm and Blue Origin’s TeraWave are adding competition rather than closing that gap.
Satellites do have a clear role. They reach islands, ships, aircraft and remote regions where laying fibre makes little economic sense and they offer useful backup when cables fail. What they cannot do is carry the sustained, high-volume traffic that hyperscale cloud and AI workloads need between data centres. This means that satellites will continue to help in extending connectivity to places cables cannot reach, while fibre cables will continue to carry the weight of global traffic.
The industry’s most immediate bottleneck
The cable repair fleet is ageing faster than it is being replaced, making it the industry’s most immediate bottleneck. As per industry estimates, total cable kilometres will grow by 48 per cent by 2040, while two-thirds of dedicated maintenance ships will be nearing the end of their working lives. Bridging that gap will need at least 15 replacement ships, plus five more in Asia, at a cost of around $3 billion, just to hold current service levels rather than improve them. Most of today’s fleet comprises converted second-hand vessels from other industries and only around 60 cable ships exist worldwide to cover the entire planet.
Fault rates are not evenly spread either. Southeast Asia, particularly the South China Sea, accounts for close to half of all cable faults each year, driven by heavy fishing activity and overlapping maritime claims that slow down repair permits. The Red Sea and the waters around Taiwan add further risk, this time from a mix of geology and geopolitics.
The same ownership shift running through the other trends is evident here too. Hyperscalers are now among the biggest funders of new maintenance capacity, alongside carriers and governments, which is starting to change how repair priorities are set.
Other challenges
Building a subsea cable typically takes one to three years and costs hundreds of millions of dollars. The delays rarely come from engineering issues. Permitting is usually the biggest obstacle. National approval for a cable landing station can be straightforward, while municipal-level permits for the same site often are not. With only around 60 cable-laying ships available worldwide, any delay risks losing a booked vessel to another project entirely.
Financing has adapted according to this risk. Many projects now pre-sell capacity well before the cable is ready for service, which helps stabilise cash flow, while some operators use supplier-led models that standardise equipment and cut costs. Even so, geopolitics can any time or at any stage override good planning altogether. For instance, several attempts to build cables across the Arctic collapsed the same way, as partners withdrew for political reasons.
Outlook
Every trend here points the same way. Cables are no longer just background infrastructure. They are assets that hyperscalers and governments want to own, protect or fight over. Bandwidth is not really the problem as it keeps growing on trend anyway. The real problem is ships, permits and trust, that is, who gets to build and fix cables and where.
Moving ahead, more government money is expected to flow into cable protection and repair, with other regions following the Baltic model. Similarly, single-company ownership of mega projects is expected to face greater scrutiny and the network itself is expected to continue fragmenting along political lines rather than growing as one interconnected system.