The cloud may look virtual, but much of its global reach still depends on physical cables lying on the ocean floor.
Key Takeaway
Submarine cables carry more than 99% of international data traffic. More than 170 cable repairs were reported worldwide in 2025, equivalent to almost four repairs per week. The strategic risk is not that one broken cable will bring down the internet. It is that concentrated routes and landing points can turn routine physical failures into serious regional bottlenecks for cloud services and other digital infrastructure.
The Internet's Hidden Backbone
The modern cloud is often described as if data simply moves between invisible digital platforms. In reality, cloud computing depends on an enormous physical infrastructure: data centres, terrestrial fibre, exchange points, power systems and, between continents, submarine cables.
That last layer is becoming increasingly important as cloud computing, artificial intelligence and distributed digital services generate more international traffic.
The International Telecommunication Union (ITU) describes submarine cables as the hidden backbone of global connectivity. Its current data shows that submarine cables carry more than 99% of international data flows, while more than 500 active and planned submarine cable systems span the globe.
The question is therefore no longer simply whether enough bandwidth exists.
It is whether that bandwidth is diverse, geographically distributed and resilient when physical infrastructure fails.
The Bottleneck Is Often Geography, Not Bandwidth
Submarine cables do not form a perfectly distributed global mesh.
Cables converge at landing stations, coastal gateways and established international corridors. Those locations become strategically important because multiple systems can depend on the same geographic area.
The World Bank has highlighted this problem particularly in lower-income countries. Nearly half of low-income countries with submarine cables rely on only one or two landing points. The same analysis found that low-income countries account for just 0.2% of global data-centre capacity, while lower-middle-income countries account for about 6%.
That creates an infrastructure imbalance.
A country can technically be connected to the global internet while still having relatively little redundancy between its domestic networks and the international cloud infrastructure it depends on.
For cloud providers, the implication is straightforward: resilience cannot be measured only by the number of servers inside a data centre. The paths connecting those servers to users and other cloud regions matter too.
Why Cable Failures Are Not Rare
Submarine cables are engineered for long operating lives, but they operate in a difficult physical environment.
The ITU identifies accidental human activity, including fishing and anchoring, as major sources of damage, alongside natural hazards such as earthquakes and underwater landslides. ITU also reports that more than 170 cable repairs were recorded worldwide in 2025, equivalent to almost four cable failures per week on average.
That creates an important distinction between failure frequency and service impact.
A cable can fail without most users noticing if sufficient alternative capacity exists.
The same failure can become disruptive if a region has few alternative routes, if those routes are already congested, or if another cable fails before the first one is repaired.
This is why redundancy matters as much as raw capacity.
Cloud Providers Are Designing Around Failure
Cloud companies are not unaware of this vulnerability. Their infrastructure strategies increasingly show that geographic diversity itself is a form of resilience.
AWS, for example, is developing Sta'O'Nuk, a planned transpacific submarine cable connecting the United States and Japan. AWS says the system is designed to add 420 terabits per second (Tbps) of capacity to its global network and is expected to become operational in 2029.
The significance is not simply the size of the number.
A new cable can provide additional bandwidth, but strategically different routes can also provide an alternative path when existing infrastructure is disrupted.
Google Cloud is pursuing a similar approach in the Americas.
In August 2026, Google announced Alisios, Canoa and OlaLuz, three new subsea cable systems designed to create additional and geographically diverse routes across the Caribbean, Latin America, Europe and North America. Google specifically describes the systems as improving the reach, reliability and resilience of its connectivity infrastructure.
The Real Risk Is Concentration
It is tempting to describe submarine-cable risk as a simple question of whether cables break.
That misses the more important issue: where the cables break, what other routes exist, and how much spare capacity those routes have.
Imagine a region served by five independent cable systems. One fails and traffic shifts to the other four. The event may be almost invisible.
Now imagine that several of those systems depend on the same landing area or broadly similar corridor. A localized disruption could affect several systems simultaneously.
This is why the ITU's 2026 resilience recommendations emphasize route diversity, infrastructure redundancy, improved risk monitoring, stronger preparedness and faster response.
The policy direction is increasingly clear: resilience is becoming an infrastructure-design objective rather than simply a repair problem.
What This Means for the Cloud and AI Economy
The consequences extend beyond conventional web browsing.
Cloud platforms increasingly support financial transactions, enterprise software, video distribution, remote work, scientific computing and AI workloads.
AI adds another layer because modern systems increasingly depend on distributed infrastructure. Training, inference, storage, content delivery and applications do not necessarily occur in one physical location.
AWS explicitly connects resilient subsea connectivity with latency-sensitive applications and the broader infrastructure supporting cloud computing.
Google's Americas Connect strategy similarly links geographically separated subsea routes with access to multiple cloud regions.
This makes subsea infrastructure part of the AI infrastructure story.
The limiting factor for some digital workloads may not be processor availability alone. It can also be the reliability, capacity and latency of the networks connecting compute resources, data stores and users.
That does not mean a cable failure automatically stops an AI service. Major cloud networks are specifically designed with redundancy.
It means that as digital workloads become more distributed, the economic value of network diversity rises alongside compute capacity.
The Next Infrastructure Race Is About Diversity
The next phase of submarine-cable investment is therefore likely to be judged by more than raw terabits per second.
The more consequential questions are:
What Comes Next
Submarine cables are unlikely to become less important as cloud computing expands.
The opposite is more likely.
The global digital economy is becoming increasingly dependent on cloud regions, AI infrastructure and continuous international data exchange. At the same time, submarine cable systems remain physical assets exposed to geography, maritime activity, natural hazards, ageing infrastructure and repair constraints.
The answer is not to eliminate that vulnerability—something impossible for a global physical network.
The answer is to design around it.
More routes. More landing diversity. More redundancy. Better monitoring. Faster repair. Stronger international coordination.
For cloud companies, governments and businesses, resilience is becoming a question of infrastructure architecture, not simply uptime promises.
The cloud may be virtual from the user's perspective.
Its weakest links are still very physical.
Does the route create genuine geographic diversity?
Does it reduce dependence on existing concentration points?
Are there multiple landing locations?
Can traffic be rerouted when a system fails?
Is sufficient spare capacity available during an outage?
Can damaged infrastructure be repaired quickly?
Are there enough specialized resources to support maintenance and recovery?
Google's Americas Connect expansion illustrates this approach. Its new systems are designed to create additional routes and redundant paths linking Latin America, the Caribbean, North America and Europe.
That points to a broader change in how connectivity infrastructure should be evaluated.
The value of a cable increasingly depends not only on what it can carry when everything works, but on what it allows the network to do when something fails.
Selected Infrastructure Data
Indicator | Current Figure | Why It Matters |
International data traffic carried by submarine cables | 99%+ | Shows the critical role of subsea infrastructure in global connectivity |
Active and planned submarine cable systems | 500+ | Demonstrates the scale of the global network, while not eliminating geographic concentration |
Global cable repairs reported in 2025 | 170+ | Shows that physical cable faults are a recurring infrastructure event |
Typical repair challenge | Days to weeks, depending on location and conditions | Demonstrates why route redundancy and preparedness matter |
AWS Sta'O'Nuk planned capacity | 420 Tbps | Illustrates the scale of new cloud-driven subsea investment |
AWS Sta'O'Nuk expected operation | 2029 | Shows the long planning and construction cycle for major connectivity infrastructure |
Sources: ITU and AWS.
A Question for Readers
If a region had to choose between one extremely high-capacity international cable and several smaller, geographically diverse cables, which infrastructure model would provide greater long-term resilience—and what trade-offs should governments and cloud companies consider?
Primary Sources
International Telecommunication Union (ITU) — Submarine cable resilience and global connectivity data.
World Bank — International connectivity, landing-point concentration and data-centre distribution.
AWS — Sta'O'Nuk transpacific submarine cable project.
Google Cloud — Americas Connect, Alisios, Canoa and OlaLuz.
Editorial note: The previously suggested “~200,000 km of new cable installed in 2024” figure has been removed because we could not establish a sufficiently direct primary-source basis for that exact number. This is the safer editorial choice for The Wider Lens.
Source URLs (editable)
ITU: https://www.itu.int/en/mediacentre/backgrounders/Pages/submarine-cable-resilience.aspx
World Bank: https://blogs.worldbank.org/en/voices/digital-technology-is-driving-the-future-of-jobs
AWS: https://aws.amazon.com/blogs/infrastructure-sustainability/staonuk-how-a-new-transpacific-subsea-cable-will-reshape-cloud-connectivity-between-north-america-and-japan/
Google Cloud: https://cloud.google.com/blog/products/infrastructure/introducing-americas-connect
Comments
Post a Comment