The Physical Reality Behind Every Web Request
When you load a webpage hosted on a server in another continent, the round trip feels instantaneous. That speed is earned through one of the most remarkable engineering achievements in history: thousands of kilometers of hair-thin glass fiber stretched across the floor of the ocean, carrying your IP packets as pulses of light moving at roughly two-thirds the speed of light in a vacuum.
The common mental model of the internet as something wireless or cloud-based is accurate for the last mile — the Wi-Fi in your home, the LTE signal on your phone. But the moment your data leaves a regional network and crosses an ocean, it almost certainly enters a submarine cable system. As of 2024, there are over 550 active submarine cable systems, representing more than 1.3 million kilometers of deployed fiber. Satellites carry less than 1% of intercontinental data traffic. The ocean floor is the actual backbone of the global internet.
How IP Packets Become Light Pulses
Your data does not travel through an undersea cable as electrical signals. It travels as light. The conversion happens at the cable's landing station — a coastal facility where the undersea cable comes ashore and connects to terrestrial fiber networks.
At the landing station, electrical signals from land-based routers are converted by optical transceivers into modulated laser light. Each wavelength of light corresponds to a separate data channel. Modern systems use Dense Wavelength Division Multiplexing (DWDM), which allows dozens to hundreds of different wavelengths to travel simultaneously through a single fiber strand, each carrying independent data streams. A modern cable pair can carry multiple terabits per second of aggregate capacity.
Your IP packet's header — containing the source IP, destination IP, TTL, and protocol fields — travels encoded in this light signal. Every router along the path reads only the IP header to make forwarding decisions. The payload is never examined by intermediate routers during normal operation.
Cable Architecture: From Shore to Shore
A submarine cable system has several physical components that work together to carry IP traffic across an ocean basin:
The fiber core is made of ultra-pure silica glass. The core diameter is typically around 8-10 micrometers for single-mode fiber. Even at this scale, light can travel hundreds of kilometers before the signal attenuates significantly — but not thousands of kilometers without help.
Optical amplifiers (Erbium-Doped Fiber Amplifiers, EDFAs) are placed every 60-100 kilometers along the cable route on the ocean floor. These devices boost the optical signal without converting it back to electrical form. They are powered by a constant-current DC power feed running through a conductor wrapped around the cable — typically at tens of thousands of volts to minimize resistive losses over such long distances. A cable crossing the Pacific Ocean may have over 150 repeaters/amplifiers on the seafloor.
Branching units are subsea nodes that split the cable to reach multiple landing stations. They allow a single cable system to serve multiple countries without requiring a separate cable for each destination.
The cable itself is armored in shallow coastal waters where it is most vulnerable to damage from ships' anchors, fishing trawls, and underwater landslides. The armor consists of steel wire strands wrapped around the fiber core. In deep water, where less physical intervention risk exists, the cable is much thinner and lighter — sometimes only slightly thicker than a garden hose.
IP Routing and the Role of BGP
The undersea cable provides the physical transport layer. IP routing over that transport is handled at the network layer by Border Gateway Protocol (BGP), the routing protocol that connects autonomous systems (AS) — the large networks operated by ISPs, cloud providers, and content delivery networks.
When your packet enters a submarine cable, it is being carried on a circuit that was negotiated between BGP peers at the cable landing stations. The cable capacity is typically sold to ISPs and content providers as IRUs (Indefeasible Rights of Use) — essentially long-term leases on a specific amount of fiber capacity. These ISPs then announce their IP prefixes to their BGP peers at the landing station, and packets destined for those prefixes are routed over the cable.
BGP's path selection is based on AS-PATH length, local preference, MED values, and other attributes — not on physical cable distance or latency directly. This is why IP traffic sometimes takes seemingly indirect routes: the BGP path attributes have been configured to prefer certain providers, and the physical reality of which cables carry that traffic follows from those decisions.
Real-World Use Cases
Content delivery networks: Companies like Cloudflare, Akamai, and Amazon CloudFront operate servers in data centers adjacent to cable landing stations specifically to minimize the number of submarine hops required to serve content. When you stream a video, the CDN edge node nearest to your ISP's peering point — often reachable via a single submarine cable segment — serves the content instead of a server on the other side of the world.
Financial trading: High-frequency trading firms pay premium prices for the lowest-latency submarine cable routes between financial centers. The difference between a cable taking the most direct geographic route versus a longer route can represent multiple milliseconds — which at trading speeds is significant. Dedicated cable systems have been built specifically to shave latency between New York and London or between the US and Japan.
Cloud provider connectivity: Hyperscale cloud providers (AWS, Google Cloud, Microsoft Azure) have invested in or fully own submarine cable systems to interconnect their global data center regions. Google's Dunant cable and Meta's 2Africa cable are examples of private cable systems built for cloud provider traffic, independent of shared carrier infrastructure.
What Happens When a Cable Breaks
Cable breaks are not rare. Approximately 100-150 submarine cable faults are reported globally per year. The vast majority are caused by human activity: ship anchors dragging on the seafloor, fishing trawl gear, and submarine landslides near continental shelves. Shark bites, despite popular mythology, account for a very small fraction of cable faults.
When a cable is cut, IP routing responds through BGP reconvergence. Routers at the affected landing stations detect the loss of the optical signal (or the BGP session drops), withdraw the affected prefixes, and BGP propagates updated routing information across the internet. Traffic is rerouted to alternative paths — other submarine cables, satellite links, or terrestrial routes. This reconvergence typically takes seconds to minutes depending on BGP timer configurations.
The end user experience during a cable cut depends on redundancy: major routes between continents have multiple parallel cable systems, so a single break typically causes congestion and latency increases rather than complete outages. Routes with limited cable diversity — some island nations or smaller markets — can experience more significant disruption.
Comparison: Submarine Cable vs. Satellite vs. Terrestrial
| Parameter | Submarine Cable | Geostationary Satellite | LEO Satellite (e.g., Starlink) | Terrestrial Fiber |
|---|---|---|---|---|
| Typical latency (intercontinental) | 60-150 ms | 550-650 ms | 20-60 ms | N/A (regional only) |
| Capacity per link | Multiple Tbps | Gbps range | Gbps range | Multiple Tbps |
| Geographic coverage | Coastal/connected regions | Global | Global (expanding) | Land-connected regions |
| Resilience | High (multiple cables) | High (orbit) | High (constellation) | High (route diversity) |
| Primary use | Bulk intercontinental traffic | Remote/broadcast | Last-mile broadband | Regional/metropolitan |
Common Misconceptions
Satellites carry most of the internet's international traffic
This is a persistent myth. Geostationary satellites carry less than 1% of intercontinental internet traffic. Their 550+ ms round-trip latency (due to the 35,786 km orbital altitude) and limited capacity make them unsuitable for bulk data transport. Submarine cables carry the overwhelming majority of international IP traffic. Low Earth Orbit (LEO) satellite constellations like Starlink are changing some edge cases, but they do not come close to submarine cable capacity for international backbone traffic.
A single cable cut can bring down the internet
No single submarine cable carries so much traffic that its loss would cause widespread internet disruption. Major routes between large markets have five to fifteen or more cable systems providing redundancy. While a cut can cause congestion and latency spikes on alternative paths, the internet's routing protocols are specifically designed to route around failures.
IP packets travel faster in vacuum than through fiber
Light travels at approximately 299,792 km/s in a vacuum but only at roughly 200,000 km/s through optical fiber (about two-thirds the speed of light). This means the actual propagation delay through a fiber cable is higher than the theoretical minimum set by the speed of light. The latency from London to New York via fiber is approximately 28 ms one-way based purely on propagation; in practice it is 70-80 ms due to routing and equipment delays.
Undersea cables are always on the deepest ocean floor
Cable routes are carefully planned around ocean bathymetry and geological hazards. In shallow coastal zones, cables are deliberately buried beneath the seafloor using specialized plows to protect them from anchors and trawls. In deep water, cables rest on the bottom but are chosen to avoid areas with known turbidity currents or seismic activity that could cause landslides. Route planning is as much a geological exercise as an engineering one.
Pro Tips
- Use traceroute to observe actual submarine hops. Run
traceroute(ortracerton Windows) to a server on another continent. The large latency jumps between specific hops correspond to transits across submarine cable segments. You can cross-reference the hop IP addresses with WHOIS data to identify which carriers are involved. - Check the TeleGeography Submarine Cable Map (submarinecablemap.com) to see the actual routes, landing stations, and capacity of every major cable system. This is the definitive public reference for submarine cable infrastructure.
- Monitor cable disruption news if your business depends on connectivity to specific regions. Services like ISOC's pulse and network operator mailing lists report cable faults as they occur, giving advance warning of potential latency or capacity impacts.
- For latency-sensitive applications, choose cloud provider regions and CDN configurations that minimize submarine hops. A Singapore-based application serving users in Europe will always have higher latency than one hosted in Frankfurt — the submarine cable transit adds irreducible propagation delay regardless of how optimized the software is.
- Understand that BGP path and physical cable path can diverge. Traffic between two cities may travel through a third continent because of BGP policy decisions by ISPs. Tools like RIPE's RIS or BGPlay can show you the AS-path that traffic takes, which often reveals surprising routing detours.
- Diversify your connectivity providers if operating infrastructure in regions with limited submarine cable diversity. Islands and developing regions may have only one or two cable systems. Combining terrestrial ISPs, satellite links, and multiple cable-connected carriers reduces single-point-of-failure risk.
Every IP packet you send across an ocean is a small marvel of physics and engineering — converted to light, amplified dozens of times on the seafloor, reconverted to electrical signals, and handed off to routers that make forwarding decisions in microseconds. Understanding this physical layer is essential for anyone designing global network infrastructure. Trace your own IP's global path here.