Data Take to Cross the ocean through undersea cable?

Data usually crosses the Atlantic Ocean through an undersea fiber-optic cable in roughly 30 to 40 milliseconds one way. A direct trans-Pacific crossing is commonly closer to 50 to 80 milliseconds one way, depending on the cable route and landing points. Those are propagation estimates for the ocean-spanning path—not the total time required to load a website, complete a video call, or download a file.

The reason is physical distance. Light travels through ordinary communications fiber at about 200,000 kilometers per second, or approximately 4.9 to 5 microseconds per kilometer. A 6,400-kilometer cable therefore creates about 31 milliseconds of one-way propagation delay before adding the time spent on land, in network equipment, in queues, and at the destination server.

A ping result is normally a round-trip measurement. The packet travels to the destination and a reply returns, so an Atlantic connection that needs about 31 milliseconds in each direction already has a physical round-trip floor near 63 milliseconds. Real-world round-trip times are generally higher because the complete route is longer than the undersea portion and the network is not frictionless.

Large data transfers take longer for a different reason: all the bits must be transmitted through the available connection. The first bit of a file might cross the ocean in about 30 milliseconds, while the last bit may arrive seconds or minutes later depending on file size, actual throughput, congestion, packet loss, and the user’s internet speed.

Quick Answer

Allow approximately:

30 to 40 milliseconds one way for a relatively direct transatlantic undersea crossing
60 to 80 milliseconds for the Atlantic portion of a round trip
50 to 80 milliseconds one way for many trans-Pacific routes
100 to 160 milliseconds or more for the Pacific portion of a round trip
Longer end-to-end times after including terrestrial networks, routers, security handshakes, server processing, congestion, and indirect routing

There is no single ocean-crossing time because the Atlantic, Pacific, Indian, Arctic, and regional seas use cables with very different lengths and routes. A signal does not always travel between the two geographically closest landing points. Network operators select paths according to connectivity agreements, cost, available capacity, failures, maintenance, and traffic engineering.

The best general conversion is:

Fiber distance in kilometers × 4.9 microseconds = approximate one-way propagation delay

For quick mental arithmetic, use 5 microseconds per kilometer. That equals about 5 milliseconds per 1,000 kilometers one way.

Typical Ocean-Crossing Times

These ranges describe the undersea or long-distance propagation component. They are not promises for a particular internet connection.

Route type	Approximate cable distance	Fiber-only one-way estimate	Approximate fiber-only round trip
Short regional sea crossing	500 to 2,000 km	2.5 to 10 ms	5 to 20 ms
Direct Atlantic-scale crossing	5,500 to 7,000 km	27 to 35 ms	54 to 70 ms
Longer Atlantic or multi-landing route	7,000 to 10,000 km	35 to 49 ms	70 to 98 ms
Direct North Pacific-scale crossing	9,000 to 12,000 km	44 to 59 ms	88 to 118 ms
Long Pacific system or branched route	12,000 to 18,000 km	59 to 88 ms	118 to 176 ms

An internet measurement from a home in one country to a server in another includes much more than the cable. The packet may travel hundreds or thousands of kilometers over land before reaching the landing station, take an indirect ocean route, cross several carrier networks, and travel inland again after it lands.

The Calculation Behind the Answer

The speed of light in a vacuum is about 299,792 kilometers per second. Light moves more slowly through glass because the material has a refractive index. A practical networking estimate is approximately 4.9 to 5 microseconds per kilometer of fiber.

One millisecond contains 1,000 microseconds. Therefore:

1,000 km × 4.9 microseconds per km = 4,900 microseconds = 4.9 milliseconds

Example: A 6,400-Kilometer Atlantic Cable

Google describes the Dunant transatlantic cable as approximately 6,400 kilometers long.

One-way propagation estimate:

6,400 km × 4.9 microseconds = 31,360 microseconds

31,360 microseconds ÷ 1,000 = 31.36 milliseconds

Fiber-only round-trip estimate:

31.36 ms × 2 = 62.72 milliseconds

That is a theoretical propagation estimate for the stated cable length. It does not include the trip from the user to the cable landing, the trip from the opposite landing to the server, or network and server delays.

Example: A 6,605-Kilometer Atlantic Cable

TeleGeography lists MAREA, which connects Virginia Beach and Bilbao, at 6,605 kilometers.

6,605 km × 4.9 microseconds = 32,364.5 microseconds

32,364.5 microseconds = about 32.36 milliseconds one way

32.36 ms × 2 = about 64.73 milliseconds round trip

Again, an actual ping between devices near those regions will be higher because it measures an end-to-end path rather than only the wet cable.

Example: An 11,629-Kilometer Pacific Cable System

TeleGeography lists the FASTER system at 11,629 kilometers, with landings in Japan, Taiwan, and Oregon. Treating the full published system length as one continuous illustration gives:

11,629 km × 4.9 microseconds = 56,982.1 microseconds

56,982.1 microseconds = about 56.98 milliseconds one way

56.98 ms × 2 = about 113.96 milliseconds round trip

This does not mean every endpoint pair on a branched system traverses the full published length. The example demonstrates why Pacific distances usually create a higher physical latency floor than a direct North Atlantic crossing.

Does Packet Size Change the Ocean Propagation Time?

Packet size does not materially change how fast the optical signal propagates through a fixed cable route. A larger packet does, however, take longer to place onto a link. This separate delay is called serialization or transmission delay.

A 1,500-byte packet contains 12,000 bits. On a 1-gigabit-per-second link, the ideal serialization calculation is:

12,000 bits ÷ 1,000,000,000 bits per second = 0.000012 second, or 12 microseconds

On a 100-megabit-per-second link, the same calculation produces 120 microseconds. Once each bit is in the fiber, it still needs about 31 milliseconds to traverse a 6,400-kilometer cable. This is why propagation dominates the first-packet delay over an ocean, while access speed and throughput dominate the completion time for large transfers.

One-Way Time, Round-Trip Time, and Download Time Are Different

Many confusing claims about internet speed come from comparing three different measurements.

One-Way Propagation Time

This is how long a signal takes to travel from one point to another. It is the closest match to the question of how long data takes to cross an ocean.

Accurate one-way measurement is difficult because clocks at both endpoints must be synchronized precisely. For ordinary troubleshooting, people often estimate one-way delay as half of the round-trip time. That is only an approximation because outbound and return traffic can follow different routes and experience different queues.

Round-Trip Time

Round-trip time, or RTT, measures the trip to a destination and back. The ping command usually reports RTT, not one-way delay. It can include:

access-network delay
terrestrial fiber
the undersea crossing
router and switch handling
queueing during congestion
destination response behavior
a different route on the return trip

A 70-millisecond ping does not mean the data took 70 milliseconds to cross the Atlantic one way. A substantial portion of that result represents the return journey.

Full Transfer Time

Transfer time measures how long it takes all requested data to arrive. It depends heavily on throughput and file size. Latency affects how quickly communication begins and how transport protocols expand their sending rate, but bandwidth determines how much data can be in motion at once.

For example, a 100-megabyte file contains approximately 800 megabits when using decimal units.

At an ideal 100 megabits per second:

800 megabits ÷ 100 megabits per second = 8 seconds

The first part of that file can cross an Atlantic cable in roughly 30 milliseconds, yet sending the entire file still requires at least about eight seconds at that connection rate, plus protocol overhead and any slowdowns.

What Actually Travels Through an Undersea Cable?

Modern submarine telecommunications cables contain pairs of optical fibers. Networking equipment encodes digital information onto optical signals, and those light signals travel through the fiber. Data is divided into packets by networking protocols, but the wet plant in the ocean is fundamentally an optical transport system rather than a row of underwater internet routers reading web addresses.

Google’s description of its Firmina cable explains that data travels as pulses of light and that the signal is amplified about every 100 kilometers. The cable also contains a conductor that allows landing stations to feed electrical power to the submerged equipment.

Multiple wavelengths can operate on a fiber pair, allowing enormous amounts of traffic to share the physical system. A modern cable may contain many fiber pairs and carry traffic for multiple owners or customers. Capacity can be increased by upgrading terminal equipment and lighting additional wavelengths or fiber pairs, within the cable system’s design limits.

The cable does not carry one complete email, movie, or cloud backup as a single object. Packets from many users and services are interleaved across high-capacity links. The receiving network reconstructs each flow according to the relevant protocols.

The End-to-End Journey of an International Request

The ocean segment is only one part of a request to an overseas server.

1. Device and Local Network

A phone, computer, television, or other device sends data through Wi-Fi, Ethernet, or cellular service. Local radio interference, a busy router, or poor signal can add more delay than expected before the packet reaches the provider.

2. Internet Provider Access Network

The service provider carries the packet from the home, office, or mobile network into its regional infrastructure. Cable, fiber, fixed wireless, cellular, and satellite access networks have different latency characteristics.

3. Regional and Long-Haul Terrestrial Fiber

The packet moves toward a major exchange point, data center, or submarine cable landing. A person living far inland may add thousands of kilometers of terrestrial travel before the packet reaches the coast.

4. Cable Landing Station

The dry portion of the submarine cable system connects to terminal equipment. Optical line systems generate, receive, monitor, and correct the high-capacity signals used on the wet plant.

5. Undersea Cable

The signal propagates through the submerged fiber. Optical repeaters compensate for signal loss along long routes. Branching units can connect additional landings or direct fiber paths within the cable system.

6. Opposite Landing and Terrestrial Network

After reaching land, the data enters terrestrial networks again. The destination may be near the landing station—or on the other side of the continent.

7. Destination Network and Server

Firewalls, load balancers, application servers, databases, and storage systems may all participate before a response is created. A slow application can make the user wait even when the undersea network path is functioning normally.

8. Return Journey

Interactive services need a response. The return traffic may use the same cable, another fiber pair, a different cable system, or a different carrier route. Internet routing is not required to be symmetric.

Why the Cable Is Longer Than a Straight Line on a Map

Cable planners do not simply draw the shortest great-circle line and drop a cable onto the seabed. A route must account for:

suitable landing locations
water depth and seabed conditions
steep slopes, trenches, and geologic hazards
fishing and anchoring areas
shipping routes
environmental and permitting requirements
existing cables and pipelines
repair access
political and jurisdictional considerations
valuable connections to additional countries or islands

The installed cable also includes engineered slack. The seabed is not flat, and the cable must conform safely to the route. Near shore, it may be armored and buried for protection. In deep water, much of it can rest directly on the seabed.

Every extra kilometer adds approximately 4.9 to 5 microseconds one way. A route that is 1,000 kilometers longer adds about 5 milliseconds one way, or roughly 10 milliseconds to the physical round-trip floor.

Do Underwater Repeaters Add Much Delay?

Repeaters are necessary because optical power weakens as light travels through fiber. In a modern repeatered submarine system, submerged optical amplifiers strengthen the signal so it can continue across thousands of kilometers.

They are not ordinary packet routers. They generally do not open internet packets, choose the next country, or wait for a web server. Their delay is small compared with the propagation time created by the cable’s length. Digital signal processing, forward error correction, transponders, switching, and terminal equipment can add latency at the system level, but distance remains the dominant unavoidable component of a long ocean crossing.

This is why adding cable capacity does not make light travel faster. A new system can provide a shorter route, better terminal design, less congestion, or more direct interconnection, but it cannot eliminate the propagation limit imposed by distance and fiber.

Bandwidth Does Not Eliminate Distance

Bandwidth is the amount of data a link can carry over time. Latency is the delay before data arrives. A cable capable of hundreds of terabits per second can carry an extraordinary number of simultaneous streams, but each signal still needs tens of milliseconds to cross an ocean.

Imagine a very wide conveyor belt. Making the belt wider allows more packages to travel at once. It does not necessarily make the front package reach the other side sooner.

High bandwidth is especially valuable for video, backups, cloud replication, and large transfers. Low latency is especially important for actions that require repeated back-and-forth exchanges, such as:

online games
voice and video conversations
remote desktops
interactive cloud applications
financial trading
database queries
authentication sequences
webpages that request many dependent resources

A service can have ample bandwidth and still feel sluggish if it repeatedly waits for distant responses.

Why High Latency Can Reduce a Large Transfer’s Speed

The simple file-size calculation assumes the connection immediately uses all available bandwidth. Real transport protocols must manage congestion, acknowledgments, packet loss, and receiver capacity.

TCP performance on a high-bandwidth, long-delay path depends on having enough data in flight. The IETF’s TCP high-performance extensions address paths with a large bandwidth-delay product by supporting larger receive windows and round-trip measurements.

The bandwidth-delay product can be thought of as the amount of data needed to fill the path. For example, a 1-gigabit-per-second connection with a 100-millisecond RTT has an idealized product of:

1,000,000,000 bits per second × 0.1 second = 100,000,000 bits

100,000,000 bits ÷ 8 = 12,500,000 bytes, or about 12.5 megabytes

Roughly 12.5 megabytes would need to be in flight to keep that idealized path full. Modern operating systems and transport protocols can manage large windows, but packet loss, congestion controls, server limits, and application design can still prevent full use of the available link.

Why Loading a Website Can Take Much Longer Than One Crossing

A website may require several round trips before useful content appears. The browser may need to:

resolve the domain name
establish a transport connection
negotiate encryption
send the request
wait for server processing
receive the initial document
discover additional scripts, styles, images, and data
make more requests

Modern protocols reduce some repeated setup, combine requests, and reuse connections. Even so, a page with dependent operations can multiply the effect of distance. A 70-millisecond RTT repeated several times becomes noticeable before download size is considered.

This is one reason content delivery networks place copies of popular content near users. If a video segment, image, software file, or webpage is served from a nearby cache, the request may never cross an ocean even though the company operating the website is based overseas.

Does Visiting a Foreign Website Always Send Data Across an Ocean?

No. A country-code domain, foreign company, or overseas news organization does not prove the server is in that country. Modern services use distributed cloud regions, content delivery networks, edge caches, and anycast routing.

A user in Florida might retrieve a European company’s images from a nearby U.S. cache. At the same time, a U.S. company’s specialized database request might travel to Europe. Some pages combine resources from several networks, so one part loads locally while another crosses an ocean.

The reverse is also possible: traffic between two nearby places may take an indirect international route because of carrier relationships or available paths. Geography influences routing, but commercial and technical policy also matter.

How Does the Internet Choose an Undersea Cable?

The internet is a network of independently operated networks, often called autonomous systems. The Border Gateway Protocol, or BGP, distributes reachability information between them. Operators apply policies that influence which routes they accept and prefer.

The selected path may depend on:

which providers exchange traffic
business agreements
route preference and policy
available capacity
cable ownership or leased capacity
maintenance
outages
congestion-management decisions
security or regulatory requirements

BGP does not calculate a consumer’s shortest physical route in the way a road-navigation app finds the shortest drive. A path with fewer network-level hops can still cover more kilometers. A carrier may send traffic to a distant interconnection point before it crosses the ocean.

What Happens When an Undersea Cable Is Cut?

A cable fault does not necessarily stop international internet service. Major routes have multiple cable systems, and large networks purchase or operate diverse paths. Traffic can be shifted to another fiber pair, cable, landing station, or carrier route.

Rerouting can increase latency because the alternative path is longer. It can also create congestion when surviving cables absorb traffic that normally used the damaged system. Users may see slower downloads, higher ping, packet loss, or intermittent service even though connectivity continues.

Places served by only one or two systems are more vulnerable to severe disruption. The International Telecommunication Union reports that subsea cables carry more than 99 percent of international data traffic and that hundreds of faults can occur globally in a year. Fishing activity and anchors are major accidental causes.

The data does not remain paused inside the broken cable waiting for repair. Optical continuity is lost at the fault. Network operators restore or reroute services where possible while a cable ship and repair team locate, recover, splice, test, and return the cable to service.

Can Data Take More Than One Undersea Route at the Same Time?

Yes. Networks can distribute different flows across multiple links, and traffic engineering can shift demand among cable systems. A single ordered transport connection is usually handled carefully so that excessive path variation does not scramble performance, but separate connections or engineered packet paths can use different infrastructure.

Redundant design improves capacity and resilience. It also explains why two tests between the same cities may not produce identical results. Different providers, cloud networks, applications, or times of day can select different routes.

The outbound and return directions can also differ. Dividing a ping result by two is therefore a useful estimate, not a precise measurement of either direction.

Can Encryption Make the Ocean Crossing Slower?

Encryption does not materially change the speed at which light propagates through the undersea fiber. It can add processing and connection setup at endpoints.

HTTPS, VPNs, and other secure protocols may require cryptographic handshakes or route traffic through an additional server. A VPN can add noticeable latency if the VPN endpoint is far from the natural route. For example, sending traffic from the United States to a VPN server in Asia before reaching a European site can force an unnecessary Pacific and Eurasian detour.

After a secure connection is established, modern hardware can encrypt large amounts of traffic efficiently. The larger performance penalty often comes from extra distance, additional round trips, or an overloaded endpoint rather than the mathematical encryption operation itself.

Undersea Cable Versus Satellite Timing

Submarine fiber is the primary carrier of international data, not satellites. The ITU states that more than 99 percent of international data traffic uses subsea cables.

Traditional geostationary satellite connections have much greater propagation distance because the signal travels from Earth to a satellite tens of thousands of kilometers above the equator and back down. A two-way internet exchange repeats portions of that space path, producing far higher latency than a direct ocean cable.

Low-Earth-orbit satellite systems operate much closer to Earth and can provide substantially lower latency than geostationary systems. Their traffic may still enter terrestrial fiber and undersea cables after reaching a ground station. Satellite service is valuable for ships, aircraft, remote regions, disaster recovery, and places where cable access is unavailable, but it does not replace the capacity of the global submarine network.

How to Measure an International Connection

Use Ping Carefully

ping provides round-trip time when the destination responds to its test messages. Run several probes and look at minimum, average, maximum, and variation. One result can be distorted by a temporary queue.

Some networks block or deprioritize the messages used by ping. A poor ping response does not always mean ordinary application traffic performs identically.

Use Traceroute as a Clue

Traceroute-like tools reveal responding network hops and their approximate round-trip times. They can help show where latency increases. However, routers may hide, filter, rate-limit, or label responses inconsistently. Hostnames can suggest a city or cable but are not guaranteed proof of the physical path.

Compare Nearby and Distant Destinations

Test a nearby server first. If local latency is already high, the ocean cable is not the only problem. Then compare a destination across the Atlantic or Pacific. The difference helps reveal the cost of long-distance routing.

Check More Than One Provider

Different cloud services and websites may use different cables and network agreements. Testing multiple destinations provides a better picture than treating one server as “the internet.”

Measure Throughput Separately

A speed test measures more than propagation. It can be limited by the home connection, Wi-Fi, test server, provider peering, device, and congestion. Use latency, packet loss, jitter, and throughput as separate indicators.

Why Latency Changes During the Day

The physical propagation time of a fixed cable does not change significantly between morning and evening. Variable delay usually comes from queueing, routing, access-network load, server load, or a path change.

When a link approaches capacity, packets may wait in buffers. A few milliseconds of fiber propagation can be overshadowed by long queues. Packet loss can force retransmission and trigger transport controls that reduce the sending rate.

Jitter—the variation in delay—is especially noticeable in voice calls, gaming, and remote control. A stable 80-millisecond RTT may feel better than a connection jumping unpredictably between 50 and 200 milliseconds.

Can New Technology Beat the Speed-of-Light Limit?

No networking upgrade can make information propagate through ordinary fiber faster than the physical medium permits. Engineers improve performance by reducing avoidable delay rather than eliminating propagation.

Useful approaches include:

building shorter and more direct cable routes
landing near major data-center regions
improving terrestrial backhaul
placing content in edge caches
reusing secure connections
reducing the number of application round trips
increasing bandwidth and equipment capacity
improving congestion control and packet-loss recovery
choosing a nearby cloud region

These changes can make a service feel dramatically faster even though the light in the undersea fiber still takes about 5 microseconds per kilometer.

You may also find these related technology guides helpful:

Frequently Asked Questions

How long does data take to cross the Atlantic Ocean?

A relatively direct Atlantic undersea cable commonly creates about 30 to 40 milliseconds of one-way propagation delay. The complete end-to-end route is longer because it includes terrestrial networks, equipment, queueing, and server processing.

How long does data take to cross the Pacific Ocean?

Many trans-Pacific routes require roughly 50 to 80 milliseconds one way for the long-distance fiber path. The range is wider than the Atlantic because cable systems, branches, landing points, and routes vary greatly.

Why is my ping about twice the ocean-crossing time?

Ping normally measures a round trip. The test packet travels to the destination and a reply travels back. It also includes the networks on both sides of the cable and any handling or queueing delay.

Does data travel at the speed of light underwater?

It travels as optical signals inside glass fiber, not through seawater. Light moves more slowly in glass than in a vacuum. A practical estimate is about 4.9 to 5 microseconds per kilometer of fiber.

Do undersea repeaters slow the data significantly?

Repeaters and terminal equipment add some system delay, but the cable’s length is the dominant unavoidable factor. Submerged optical repeaters amplify the signal; they are not ordinary routers examining and forwarding individual internet packets.

Does a large file take longer to cross the ocean than a small packet?

The first bits have similar propagation time, but a large file contains many more bits that must be transmitted. Total transfer time therefore depends on file size, throughput, protocol behavior, congestion, and packet loss.

Why can a download take minutes if the crossing takes milliseconds?

Milliseconds describe when the signal begins arriving. A download is limited by how fast all the data can be sent through the slowest part of the end-to-end connection. The home internet link is often a much smaller bottleneck than the submarine cable’s total capacity.

Does the internet always choose the shortest undersea cable?

No. Routing depends on network policies, business relationships, available capacity, failures, and engineering decisions. The selected path can be longer than the geographically shortest route.

Why are undersea cables not laid in perfectly straight lines?

Routes must avoid hazards, follow safe seabed terrain, meet landing and permitting requirements, connect useful locations, and include installation slack. The ocean floor is not flat, and the shortest map line is not always practical or safe.

Will one broken cable disconnect the internet?

Usually not on well-connected routes because traffic can move to other systems. A fault may still increase latency and congestion. Regions with few cable connections can experience much more serious disruption.

Is a satellite faster than an undersea cable?

Traditional geostationary satellite service has much higher propagation delay. Low-Earth-orbit systems reduce that delay and serve important locations, but direct submarine fiber generally provides exceptional capacity and low intercontinental latency.

Does using a VPN increase ocean-crossing time?

It can. A VPN adds a server to the route. If that server is near the natural path, the increase may be modest. A distant VPN exit can force a long detour and even add another ocean crossing.

Can traceroute tell me which submarine cable my data used?

Sometimes hostnames and latency changes provide clues, but traceroute is not definitive proof. Routers can hide or filter responses, paths can be asymmetric, and public labels may not reveal the actual physical fiber.

Does visiting an overseas company’s website always cross the ocean?

No. The company may use a nearby content delivery network, cloud region, or edge cache. Some resources can load locally while other requests on the same page travel internationally.

Quick Summary

Data commonly crosses a relatively direct Atlantic undersea cable in about 30 to 40 milliseconds one way. Many Pacific crossings require roughly 50 to 80 milliseconds one way. A round-trip measurement is approximately twice the one-way propagation time before adding the rest of the network.

The calculation is based on about 4.9 to 5 microseconds of delay per kilometer of fiber. A 6,400-kilometer Atlantic cable therefore has an estimated one-way propagation delay near 31.36 milliseconds and a fiber-only round-trip floor near 62.72 milliseconds.

Real internet traffic also travels through local access networks, terrestrial fiber, landing-station equipment, carrier networks, security handshakes, and destination servers. Routing can be indirect, and the return path can differ from the outbound path. That is why an actual ping is higher than a cable-length calculation.

Propagation time is not download time. The first bit of a large file can cross the ocean in milliseconds, while the final bit may arrive seconds or minutes later. File size, bandwidth, throughput, transport behavior, congestion, and packet loss determine the total transfer time.

Submarine repeaters amplify optical signals along the route and add far less delay than the distance itself. More cable capacity allows more data to travel at once; it does not make light move faster. Engineers reduce perceived delay by building direct routes, using nearby cloud regions, caching content near users, and designing applications to require fewer round trips.

Sources & References

Editorial Review

Reviewed by Claire Bennett, Managing Editor

Last reviewed: August 2026

Quick Answer Guide publishes practical, research-based answers to common questions about money, technology, health, travel, home improvement, and everyday life. Content is reviewed using official government resources, educational institutions, industry publications, and other authoritative sources when appropriate. Articles are updated periodically to improve accuracy and usefulness.

How Long Does It Take for Data to Cross the Ocean Through an Undersea Cable?