Why is the internet
slower than light?
Light travels 300,000 km/s. A ping from London to Tokyo covers ~18,000 km. The math says 60ms. Reality says 150ms. Here's why.
What is Latency?
Latency is the time it takes for a packet of data to travel from one point to another. In networking, we almost always measure Round-Trip Time (RTT) — the time for a packet to leave your computer, reach a destination, and come back. When you "ping" a server, you're measuring RTT.
Latency is measured in milliseconds (ms). For reference: human reaction time is about 200ms; a blink of an eye is 300–400ms. A latency of 1ms feels instantaneous. 100ms starts to feel sluggish in interactive applications. Above 300ms, video calls become uncomfortable.
RTT is roughly twice the one-way latency — but not exactly. The return path may be physically different, pass through different routers, and experience different queue delays. For geographically distant connections, RTT is the practical metric because it reflects real-world round-trip experience.
The Speed of Light (and Why Fiber is Slower)
Light in a vacuum travels at 299,792 km/s. But internet traffic doesn't travel through vacuum — it travels through glass fiber. Light in fiber optic cable moves at roughly 2/3 of the speed of light in vacuum, or about 200,000 km/s. This is called the refractive index effect.
This means the theoretical minimum latency for any fiber connection is about 5 microseconds per kilometer — or 5ms per 1,000 km. A London–New York connection spanning ~6,500 km of cable has a theoretical floor of around 32ms RTT. In practice, measured RTT is 70–80ms.
| Route | Cable Distance | Theoretical Min | Typical Real RTT |
|---|---|---|---|
| London → New York | ~6,600 km | ~33ms | 70–80ms |
| Los Angeles → Tokyo | ~9,600 km | ~48ms | 100–115ms |
| London → Singapore | ~21,000 km | ~105ms | 160–180ms |
| New York → Sydney | ~16,000 km | ~80ms | 170–200ms |
| Frankfurt → Mumbai | ~11,000 km | ~55ms | 110–130ms |
Why Real Latency is Always Higher
The gap between theoretical and real latency comes from several compounding factors:
Cable routing is never a straight line
Submarine cables follow ocean floor topography, avoid seismic zones, connect multiple landing points, and are often laid in arcs rather than great-circle paths. The actual cable length between two cities is typically 15–40% longer than the straight-line distance.
Terrestrial segments add distance
Data doesn't go directly from your computer to a submarine cable. It travels through local networks, regional aggregation points, and metro fiber — often hundreds of kilometers — before reaching a landing station. After the ocean crossing, the same happens in reverse.
Routing through intermediate hops
BGP routing (the protocol that directs internet traffic) doesn't always choose the shortest physical path. Traffic between Frankfurt and Singapore may route via London, New York, or even Los Angeles depending on peering agreements, congestion, and carrier contracts. This "cold potato" routing can add hundreds of milliseconds.
Queueing and processing delays
Every router along the path introduces a small delay — typically 0.1–1ms each. With 20–40 hops between continents, this adds up. Congested routers queue packets, adding variable delay. Amplifiers on submarine cables (placed every 50–80 km) introduce negligible but non-zero signal processing time.
Protocol overhead
TCP's handshake and acknowledgment mechanisms add round trips. TLS encryption for HTTPS adds 1–2 extra round trips to establish a secure connection. CDNs and anycast routing exist specifically to reduce these application-layer latencies.
How Submarine Cables Affect Latency
Submarine cables are the longest single segments of any internet path. A cable like SEA-ME-WE-5, spanning 20,000 km from Europe to Southeast Asia, contributes roughly 100ms of propagation delay on its own. The choice of which cable carries your traffic matters significantly.
This is why cable failures are so impactful: when a major cable is damaged (by ship anchors, earthquakes, or fishing trawls — the most common causes), traffic reroutes through alternative cables that may be significantly longer, or through satellites with 600–800ms latency. Users notice immediately.
Cable capacity also matters. While modern cables carry terabits per second, during peak hours congestion on overloaded routes can add tens of milliseconds of queueing delay — a phenomenon visible in GeoCables' real-time RTT measurements.
Measuring Latency with RIPE Atlas
GeoCables uses RIPE Atlas — a global network of 10,000+ hardware probes — to measure real RTT between points and verify which submarine cables carry the traffic. RIPE Atlas probes perform traceroutes and ping measurements continuously, providing ground truth about actual routing paths.
By comparing measured RTT against the theoretical minimum (based on cable distance), GeoCables can detect anomalies: a sudden increase in RTT on a specific cable corridor often indicates degraded performance, partial capacity loss, or rerouting through a longer path.
The Future: Will Latency Improve?
Hollow-core fiber, currently in commercial trials, guides light through air rather than glass — reducing propagation delay to 99.7% of the vacuum speed of light. This could cut trans-oceanic latency by ~30%. High-frequency trading firms and cloud providers are watching closely.
Low Earth Orbit (LEO) satellite constellations like Starlink offer ~20–40ms latency for shorter routes, but still can't match fiber for long intercontinental paths due to the need for multiple satellite hops and inter-satellite links.
For most intercontinental routes, submarine fiber cables will remain the lowest-latency option for the foreseeable future. The limiting factor isn't technology — it's physics.
Calculate your route's latency
Use GeoCables to find the actual cable distance and estimated RTT between any two cities, verified against RIPE Atlas measurements.
Open Calculator →