The Scale of the System
There are roughly 550 submarine cable systems currently in service or under construction, totaling more than 1.4 million kilometers of fiber on the ocean floor — enough to wrap around the equator about 35 times. These cables collectively carry more than 95% of all intercontinental internet traffic. Satellites, despite their visibility and public attention, handle a tiny fraction of global data volume. Geostationary satellites add 500–700ms of latency by physics alone (the signal travels 35,786km up and 35,786km back). Fiber adds roughly 5ms of latency per 1,000km in the cable — a New York to London path of approximately 6,600km has a theoretical minimum latency of about 33ms. In practice it is 70–80ms due to routing, amplifier delays, and the cable route not following a straight line.
How a Cable Is Built
A submarine cable looks nothing like a garden hose. From the center outward, a typical deep-water cable has: a central steel strength member, optical fiber pairs embedded in a gel-filled core, multiple layers of steel wire armor, copper conductor (which carries electrical power to the repeaters), polyethylene insulation, and an outer sheath. In deep water where abrasion is not a concern, the cable can be as thin as 17–22mm in diameter — lighter than a garden hose. In shallow coastal water where anchors, fishing gear, and currents pose real threats, the cable is armored with additional steel layers and can be 50–70mm thick. Cables in the highest-risk zones near shore are often buried under the seabed.
| Component | Function | Why It Matters |
|---|---|---|
| Optical fiber pairs | Carry data as modulated light pulses | Each fiber pair can carry dozens of wavelengths (DWDM), each wavelength at 100–400 Gbps |
| Optical amplifiers / repeaters | Boost signal strength every 50–100km | Without them, light would attenuate to noise before crossing an ocean |
| Copper power conductor | Delivers DC electrical power to repeaters from shore | Repeaters have no battery or local power — the shore end pumps up to 15,000V DC through the cable |
| Steel wire armor | Protects against abrasion, anchor strikes, fishing | Concentrated in shallow-water sections; absent in deep-water segments to save weight |
| Landing station | Connects undersea system to terrestrial networks | Houses the cable termination, DWDM transponders, power feed equipment, and colocation for carrier handoffs |
Notable Cable Systems
The modern cable map has shifted significantly in recent years. Historically, submarine cables were consortia built and owned by telecommunications carriers. Today, technology companies are the primary builders of new capacity:
- MAREA (Microsoft and Meta, opened 2017): crosses the North Atlantic from Virginia Beach, Virginia to Bilbao, Spain. Initial capacity of 160 Tbps across 8 fiber pairs. One of the highest-capacity transatlantic cables.
- Grace Hopper (Google, 2022): New York to Bude (UK) and Bilbao (Spain), with branches to the UK and Spain separately, allowing independent routing. Named after computing pioneer Grace Hopper.
- DUNANT (Google, 2021): Virginia Beach to Saint-Hilaire-de-Riez, France. Capacity of 250 Tbps, one of the highest-capacity cables in service at its launch.
- FASTER (Google consortium, 2016): Pacific cable connecting the US West Coast to Japan and other Asian landing points. 60 Tbps initial capacity across 6 fiber pairs.
- 2Africa (Meta-led consortium, under construction): planned to be one of the longest cables ever deployed at ~45,000km, encircling Africa and connecting to Europe, Middle East, and India.
Google, Meta, and Microsoft now own or have significant shares in a substantial fraction of global submarine cable capacity. This is partly cost-driven (building your own cable is cheaper than buying capacity from carriers at scale) and partly strategic — a company that controls its cable infrastructure has guaranteed bandwidth for its own traffic between data centers and users.
Cable Cuts: What Actually Happens
Submarine cables break more often than most people realize. The International Cable Protection Committee estimates approximately 100–150 faults per year globally. The causes:
- Anchors and fishing: the most common cause in shallow water, particularly in busy shipping lanes near ports. Trawl nets from fishing vessels can catch cable and pull it until it breaks.
- Submarine landslides: underwater slope failures triggered by earthquakes or sediment instability. The 2006 Hengchun earthquake near Taiwan caused multiple cable cuts in a single event, disrupting internet traffic across Southeast Asia for weeks.
- Ship anchoring: accidental anchor drops over cable corridors in congested ports, despite cables being charted.
- Equipment failure: repeater failures, which require specialized repair (the failed repeater cannot be bypassed).
When a cable cuts, traffic does not simply stop — the internet's routing protocols detect the failure and reroute traffic within minutes across available alternate paths. But if several cables share a route (as often happens in the same undersea geography), a single geological event can degrade or cut multiple systems simultaneously. The result is usually not a blackout but measurable increases in latency (because traffic takes longer alternate paths) and congestion on backup routes.
Repair Operations
Repairing a submarine cable is a serious maritime operation. A fleet of specialized cable ships — operated by companies like Alcatel Submarine Networks, SubCom, and HMN Technologies — are on maintenance contracts for major cable systems. When a fault is located (using optical time-domain reflectometry, which measures the exact distance from shore to the break), a repair ship navigates to the site, grapples the cable from the sea floor, brings it to the surface, splices in a new section, and relays it. In deep water (3,000–6,000m), this process typically takes two to four weeks from fault detection to restoration. Shallow-water repairs near shore are faster — sometimes a few days. Weather is the main variable; repair ships cannot work in severe sea states.
Latency Physics and Why Geography Still Matters
Light travels through optical fiber at approximately 200,000 km/s — about two-thirds the speed of light in vacuum, because of the fiber's refractive index. This creates a hard physics floor for intercontinental latency that no technology can shrink. New York to London: ~70–80ms minimum round trip. US West Coast to Tokyo: ~100–130ms. These floors matter for real-time applications. A game server located in Europe will always have higher latency for a user in the US than a server hosted in Virginia, regardless of ISP, router, or Wi-Fi quality. When you run a traceroute and see a hop that adds 70ms suddenly, you have just watched your packet cross an ocean.
Frequently Asked Questions
Does most international internet traffic use satellites?
No — and not even close. Submarine fiber cables carry over 95% of intercontinental internet traffic by volume. Satellites handle a tiny fraction, primarily for maritime, aviation, military, and truly remote region connectivity where laying fiber is impractical. Geostationary satellites add 500–700ms of latency, which makes them unsuitable for real-time applications. Low-earth orbit satellites like Starlink reduce latency to 25–60ms but have dramatically lower aggregate capacity than submarine fiber — a single major submarine cable system can carry more traffic than the entire Starlink constellation combined.
Do submarine cable cuts break the internet?
In the most connected regions (North America, Western Europe, East Asia), a single cable cut causes rerouting within minutes and the user impact is often unnoticeable or a slight increase in latency. In less-connected regions — parts of Africa, island nations, Southeast Asia — a single cable can carry a much larger share of total international capacity, and a cut can cause significant degradation visible to users. The 2008 cuts of the FLAG Europe-Asia and SEA-ME-WE 4 cables within days of each other disrupted connectivity across the Middle East and South Asia measurably for weeks while repairs were underway.
Why are technology companies building their own cables?
Scale and economics. Google, Meta, and Microsoft together generate a substantial fraction of all internet traffic — primarily from users accessing their platforms and from data replication between their global data centers. At that scale, buying capacity from carrier cable consortia is more expensive per bit than building dedicated infrastructure. Private cables also give these companies control over capacity planning, routing priorities, and maintenance contracts without depending on shared-ownership consortium decisions. The trend is expected to continue: industry analysts estimate that hyperscaler-owned or funded cables will represent the majority of new submarine cable capacity added through the late 2020s.
Can submarine cables affect my ping to a game server?
Yes, directly. When you connect to a game server in another region, your traffic travels through submarine cables between the continents. The specific cable and routing path used depends on your ISP's peering arrangements and the server location. A game server in Frankfurt will always have higher latency for a player in Los Angeles than a server in Seattle, because the trans-Atlantic path adds unavoidable fiber propagation delay. The best mitigation is to choose the nearest regional server, which good game matchmaking does automatically. No home networking improvement — better router, faster Wi-Fi, fiber internet — can overcome the physics of intercontinental distance.