How Fiber Optics Carry the Internet at the Speed of Light
99% of intercontinental internet traffic travels through glass threads thinner than a human hair, buried under the ocean floor. The physics is total internal reflection — light bouncing through glass for thousands of kilometres without escaping.
Table of Contents
The Invisible Infrastructure
Right now, as you read this, roughly 99% of all data traveling between continents is moving through glass. Not through satellites. Not through the air. Through hair-thin strands of ultra-pure silica glass, laid on the ocean floor, carrying pulses of light at two-thirds the speed of light in vacuum.
There are about 550 submarine fiber optic cables currently in service, totaling over 1.4 million kilometres — enough to wrap around the Earth 35 times. Every email between New York and London, every video call between Tokyo and São Paulo, every financial transaction between Frankfurt and Singapore passes through these cables. The global internet, for all its apparent ethereality, is very, very physical — and it runs on optics.
Most people don’t know this. They imagine data floating through the cloud, bouncing off satellites, traveling wirelessly. The reality is more mundane and more impressive: your data is being encoded as laser pulses, injected into a glass thread 9 micrometres wide, and shot across the Atlantic Ocean at 200,000 kilometres per second, arriving in about 28 milliseconds. The physics that makes this possible is total internal reflection — one of the simplest phenomena in optics, known since at least the 1840s.
Total Internal Reflection: Light That Can’t Escape
When light passes from a denser medium (glass, refractive index ~1.47) into a less dense medium (air, refractive index 1.0), it bends away from the normal — Snell’s law. At shallow angles, some light refracts out and some reflects back. But at a critical angle — about 43° for glass-to-air — the refracted ray runs parallel to the surface. Beyond this angle, no light escapes. All of it reflects back into the glass. Total internal reflection.
This is not a gradual effect. It’s absolute. Below the critical angle, light escapes. Above it, 100% of the light reflects. Not 99%, not 99.9%. One hundred percent. This makes total internal reflection far more efficient than any mirror — the best metallic mirrors reflect about 95–99%, losing a few percent with each bounce.
A fiber optic cable exploits this by surrounding a glass core (slightly higher refractive index) with a glass cladding (slightly lower refractive index). Light entering the core at angles greater than the critical angle bounces off the core-cladding boundary and stays trapped inside the core. It zigzags down the length of the fiber, reflecting off the boundary millions of times per kilometre, without ever escaping. The refraction physics is identical to a mirage on a hot road — only controlled, confined, and engineered.
The Glass: Purer Than You Can Imagine
Here’s something that I find genuinely remarkable. Ordinary window glass is transparent enough to see through — maybe a few centimetres, maybe a metre. You wouldn’t call it opaque. But it’s far too absorptive for fiber optics. If you made a fiber from window glass, the signal would be undetectably faint after 100 metres.
Telecom-grade silica fiber is transparent to an almost absurd degree. The attenuation — the rate at which light is lost — is about 0.2 dB per kilometre at 1,550 nm wavelength. In practical terms, this means that after traveling 1 kilometre, about 95.5% of the light is still there. After 10 km, about 63%. After 100 km, about 1%.
One percent after 100 kilometres. Through a glass thread. That’s astonishing purity. If the ocean were as transparent as telecom fiber, you could see the seafloor from the surface — even in the deepest trenches. The glass in fiber optic cables is the purest manufactured material on Earth, with impurities measured in parts per billion.
Achieving this purity required decades of materials science. The key breakthrough came in the 1970s at Corning Glass Works (now Corning Inc.), where researchers developed the modified chemical vapour deposition (MCVD) process for producing ultra-low-loss silica fiber. The loss at the time of the first usable fibers (1970) was about 20 dB/km. It dropped to 0.2 dB/km by the 1980s, approaching the theoretical minimum set by Rayleigh scattering — the same scattering of light by nanoscale density fluctuations in the glass that makes the sky blue.
Wavelength-Division Multiplexing: Colours of Data
A single fiber can carry one laser signal. But why settle for one? Different wavelengths of light — different colours, effectively — can travel through the same fiber simultaneously without interfering with each other. Each wavelength carries an independent data stream. At the receiving end, a demultiplexer (essentially a prism or diffraction grating) separates the wavelengths and routes each to its own detector.
This is wavelength-division multiplexing (WDM), and it’s the reason modern fibers carry mind-boggling amounts of data. A typical submarine cable fiber uses 100–200 wavelength channels, each carrying 100–400 gigabits per second. A single fiber pair can thus carry 10–80 terabits per second. A cable with 16 fiber pairs carries 160–1,280 terabits per second.
For perspective: a single modern submarine cable can carry more data than the entire global internet traffic of 2005. Every year, as laser and receiver technology improves, the capacity per fiber increases — and the old fibers don’t need to be replaced, just the electronics at each end.
The wavelength channels are typically spaced 50 or 100 GHz apart in the C-band (around 1,530–1,565 nm) and the L-band (1,565–1,625 nm) of the infrared spectrum. These bands correspond to the minimum absorption window of silica glass — the wavelength range where the glass is most transparent. The entire modern internet backbone operates in a sliver of the electromagnetic spectrum about 100 nm wide.
Submarine Cables: Engineering at the Ocean Floor
Laying a submarine fiber optic cable is one of the most ambitious engineering operations in regular practice. A cable ship — typically 100–150 metres long with a cable tank holding thousands of kilometres of cable — pays out cable from its stern at 6–8 km/h while navigating a planned route that avoids undersea mountains, earthquake zones, coral reefs, and shipping lanes.
The cable itself is engineered to survive decades on the ocean floor. In deep water, the cable is thin (~17 mm) and unarmoured — the seafloor is relatively benign at depth. In shallow water near coastlines, where fishing and anchoring pose risks, the cable is armoured with layers of galvanised steel wire, increasing the diameter to ~40 mm. At shore landings, it’s buried in trenches or encased in conduit.
Every 60–100 km along the cable, an optical amplifier (repeater) boosts the signal. These amplifiers use erbium-doped fiber amplifiers (EDFAs) — sections of fiber doped with erbium ions that, when pumped with laser light at 980 or 1,480 nm, amplify the signal optically without converting it to electrical form. The amplifiers are powered by a constant-current electrical supply carried by a copper conductor in the cable — the entire cable carries about 1 ampere at 5,000–15,000 volts, powered from shore stations at each end.
These amplifiers are designed to operate for 25 years on the ocean floor without maintenance. If one fails, the cable’s capacity is reduced until a repair ship can reach it. The reliability requirements are extraordinary — the electronics sit in pressurised housings at depths of up to 8,000 metres, in total darkness, at temperatures near 2 °C, for decades.
Latency: The Speed of Money
For most internet users, a few extra milliseconds of latency don’t matter. For high-frequency financial trading, they matter enormously. Trades are executed by algorithms that respond in microseconds, and the speed of light through fiber sets a hard limit on how fast information can travel between exchanges.
The New York–Chicago route is a famous case study. The fiber optic path between the NYSE and the Chicago Mercantile Exchange is about 1,400 km, giving a one-way latency of about 7 milliseconds. Trading firms have invested in straighter routes — including microwave relay towers that send data through the air at the full speed of light (faster than through fiber, because the refractive index of air is ~1.0003 vs. ~1.47 for glass) — to shave this to about 4.2 milliseconds.
The difference — 2.8 milliseconds — is worth millions of dollars per year to trading firms. This is probably the most expensive demonstration of the difference between the speed of light in glass and in air that has ever existed.
The Next Frontier: Hollow-Core Fiber
Current fiber optics have a fundamental speed limitation: light travels at about 2/3 of c in glass. Hollow-core fiber — where the light travels through air (or vacuum) inside a microstructured glass tube — could bring the speed to nearly c, reducing latency by about 30%.
Hollow-core fibers guide light not by total internal reflection but by photonic bandgap effects or anti-resonant reflection — engineered microstructures in the glass cladding that prevent light from escaping the hollow core. The physics is more complex, but the result is light traveling through air at air-speed, guided by glass that it never actually enters.
Recent advances have brought hollow-core fiber loss down to about 0.1 dB/km — actually lower than conventional solid-core fiber. If manufacturing costs can be reduced and long-length production scaled up, hollow-core fiber could become the next major upgrade to the global internet backbone. Not for bandwidth (solid-core fiber handles that fine) but for latency — and in a world where financial trades, cloud gaming, and autonomous vehicle networks need the absolute fastest possible connections, that 30% speed improvement matters.
Glass Threads Holding the World Together
There’s something almost fragile about the physical internet. The data carrying your video call, your bank transfer, your medical records is moving through a glass thread thinner than a hair, lying on the mud at the bottom of the Atlantic, protected by nothing more than some steel wire and plastic sheathing. Sharks occasionally bite the cables (attracted by the electromagnetic fields, possibly). Anchors drag across them. Earthquakes snap them.
And yet the system works, almost all the time, because there’s enough redundancy — enough parallel cables, enough rerouting capability — that individual failures don’t take down the network. The physics is simple and reliable: total internal reflection doesn’t fail. Silica glass doesn’t degrade. Laser diodes last for years. The weakest links are always mechanical — the cable lying exposed on a continental shelf where a trawler might snag it.
The entire foundation of the global internet — everything from cat videos to cryptocurrency, from academic papers to emergency communications — rests on a physical law that you can demonstrate with a glass of water and a laser pointer. Light goes into the glass and doesn’t come out. Everything else is engineering.
Frequently Asked Questions
How fast does data travel through fiber optic cables?
Light in glass fiber travels at about 200,000 km/s — roughly two-thirds the speed of light in vacuum (the refractive index of silica glass is about 1.47, so light slows down by a factor of 1.47). A signal from London to New York (about 5,600 km by undersea cable route) takes roughly 28 milliseconds one way — faster than you can blink. The latency you experience in practice is longer (typically 60–80 ms round trip) because of signal processing at repeater stations, routing equipment, and the fact that cables don't follow perfectly straight paths.
How thin is a fiber optic strand?
The core of a single-mode fiber — the part that actually carries the light — is about 9 micrometres in diameter. That's roughly 1/10th the diameter of a human hair. The cladding (outer glass layer) brings the total to about 125 micrometres. A protective coating and jacket add more, giving a total diameter of about 250 micrometres — still thinner than a human hair including the root. Modern submarine cables bundle many fibers together (typically 8–24 fiber pairs) inside layers of steel wire, copper conductor, and polyethylene insulation, giving the final cable a diameter of about 17–22 mm — roughly the thickness of a garden hose.
What happens when a submarine cable breaks?
It happens more often than you'd think — about 100–200 cable faults per year worldwide, mostly caused by fishing trawlers dragging anchors or nets across the cable, and occasionally by earthquakes or underwater landslides. When a cable breaks, traffic is automatically rerouted through redundant cables (most major routes have 10–20 parallel cables). The break is located using optical time-domain reflectometry (OTDR) — sending a light pulse down the fiber and timing the reflection from the break point, like sonar but with light. A specialised cable repair ship then sails to the location, grapples the cable from the seafloor (sometimes from depths of 4,000+ metres), splices the broken fibers on board, and relays the repaired cable. Repair typically takes 1–3 weeks depending on weather and depth.
Could wireless replace fiber optics?
Not for backbone internet traffic. The fundamental advantage of fiber is bandwidth density — a single fiber pair can carry 400+ terabits per second using modern wavelength-division multiplexing. A 5G wireless tower, by comparison, serves a few gigabits per second shared among all users. Wireless is essential for last-mile connectivity (connecting your phone to the network), but the backbone — the long-distance trunks between cities and continents — will remain fiber for the foreseeable future. The physics favours guided light: there's no signal spreading, no atmospheric absorption, no interference, and essentially unlimited bandwidth per fiber. Wireless has none of these advantages over distance.