Optical networking is a method of transmitting data using pulses of light through glass or plastic fibers instead of electrical signals through copper wire. It forms the backbone of the internet, long-distance telecommunications, and modern data centers, carrying the vast majority of the world’s digital traffic. A single optical fiber can carry 160 separate channels of data simultaneously, and commercial systems now operate at speeds of 800 gigabits per second, with 1.6 terabit products already in testing.
How Light Carries Data
At its core, optical networking works by converting electrical signals (the ones and zeros your devices produce) into light, sending that light through thin glass fibers, and converting it back to electrical signals at the other end. The component that does this is called an optical transceiver: a small module roughly the size of a pack of gum that plugs into networking equipment.
Inside each transceiver, a tiny semiconductor laser diode converts electrical current into light. That light travels through the fiber to a receiving port, where a photodetector diode does the reverse, turning the incoming light back into electrical signals that can be read by a computer, switch, or router. The entire conversion happens in nanoseconds.
Light travels through glass fiber at roughly 200,000 kilometers per second. That’s about two-thirds the speed of light in a vacuum, because the glass has a refractive index of about 1.5, which slows the light slightly. In practical terms, a signal crosses the Atlantic Ocean in under 30 milliseconds, making fiber the fastest medium available for long-distance communication.
Why Fiber Outperforms Copper
Traditional copper cables carry data as electrical pulses, which makes them vulnerable to electromagnetic interference from nearby power lines, motors, or other cables. Fiber optic cables are immune to this problem entirely because they use light, not electricity. In environments with heavy electrical noise, like factories, hospitals, or dense server rooms, this immunity is a significant practical advantage.
Distance is the other major difference. Copper cables lose signal strength quickly and typically max out at around 100 meters before needing a repeater. Fiber optic cables can transmit data over distances exceeding 100 kilometers without any signal boosting, depending on the fiber type. Copper cables top out at about 10 gigabits per second over short runs, and that speed drops further in noisy environments. Fiber routinely carries 100 Gbps and far beyond.
Single-Mode vs. Multi-Mode Fiber
Not all fiber is the same. The two main types serve very different purposes.
Single-mode fiber has an extremely thin core (about 9 micrometers) that allows only one path of light to travel through it. This eliminates signal distortion and allows data to travel up to 40 kilometers or more without amplification. It’s the standard choice for long-haul telecommunications, undersea cables, and connections between buildings or campuses. Single-mode fiber supports data rates of 100 Gbps and beyond.
Multi-mode fiber has a larger core (50 or 62.5 micrometers) that lets multiple light paths bounce through simultaneously. This limits its effective range to about 2 kilometers, but it’s cheaper to manufacture and easier to work with. Multi-mode fiber handles speeds from 10 Gbps up to 400 Gbps depending on the specific cable grade, and it’s the dominant choice inside data centers and within buildings where distances are short.
Multiplexing: Sending Multiple Signals at Once
One of the most powerful capabilities in optical networking is wavelength division multiplexing, or WDM. Instead of sending a single stream of light through a fiber, WDM sends multiple streams simultaneously, each on a slightly different wavelength (color) of light. Think of it like a highway with multiple lanes: each wavelength is its own lane, carrying its own independent data stream, and they all share the same physical fiber without interfering with each other.
There are two main flavors. Coarse wavelength division multiplexing (CWDM) spaces its channels 20 nanometers apart and supports up to 18 channels on a single fiber. It’s simpler and less expensive, making it a good fit for metro-area networks and enterprise connections where moderate capacity is enough.
Dense wavelength division multiplexing (DWDM) packs channels much more tightly, with spacing as narrow as 0.2 nanometers. This allows 80 to 160 channels on a single fiber. DWDM is what makes undersea cables and long-haul backbone networks possible at enormous scale. The tradeoff is cost: the lasers and filters needed to maintain such precise wavelength separation are significantly more expensive than CWDM equipment.
How Traffic Gets Routed
Early optical networks were simple point-to-point links. If you needed to reroute traffic, someone had to physically reconnect cables or swap components. Modern optical networks use devices called reconfigurable optical add-drop multiplexers (ROADMs) that can redirect individual wavelengths of light to different destinations without converting the signal back to electricity first.
ROADMs are controlled by software, so network operators can reroute traffic remotely and dynamically. If a particular link gets congested or a cable is damaged, the network can shift wavelengths to alternate paths in real time. This flexibility is what allows large-scale optical networks to function as true mesh networks, where data has multiple possible routes between any two points, rather than relying on a single fixed path.
Current Speeds and Where They’re Heading
The commercial standard at the high end today is 800 Gbps, widely deployed in data centers and backbone links. The next step, 1.6 terabits per second, is already being demonstrated by multiple companies, though the formal IEEE specification isn’t finalized yet. Some vendors already have functioning 1.6T products in interoperability testing.
The push for these speeds is driven largely by artificial intelligence workloads. Training large AI models requires moving massive amounts of data between thousands of processors, and traditional electrical connections create bottlenecks in both speed and energy consumption. Optical connections between chips, between servers, and between data centers are increasingly seen as the solution.
Silicon Photonics and Optical Computing
The most significant shift happening in optical networking right now is the move to build optical components directly onto silicon chips using the same manufacturing processes that produce computer processors. This approach, called silicon photonics, is growing at roughly 30% per year as an industry.
The goal is to replace the electrical wires that connect chips inside a server with tiny optical links, eliminating the energy loss and heat that come with pushing electrical signals through copper at high speeds. Broadcom is launching a 200-gigabit-per-lane optical engine in 2025, designed to scale toward 1.6T and 3.2T speeds. TSMC, the world’s largest chip manufacturer, has demonstrated technology that integrates optical and electronic components into a single advanced chip package, with commercial production expected by 2026.
Japan’s NTT is pursuing an even more ambitious vision with its IOWN project, which aims to use optical connections all the way from circuit boards to individual chips, creating what it calls a “full-optical frame.” The practical result would be networks that consume far less power while handling far more data, a combination that matters enormously as data center energy use becomes a growing global concern.
Where Optical Networks Are Used
Optical networking operates at every scale of modern communications. Undersea cables connecting continents are optical. The backbone links between cities that carry internet traffic are optical. The connections between data centers run on fiber. Increasingly, fiber reaches directly to homes and businesses through fiber-to-the-home (FTTH) deployments, which is what services marketed as “fiber internet” actually deliver.
Inside data centers, short-reach optical links connect rows of servers using multi-mode fiber. Between data centers in the same metro area, single-mode fiber with DWDM carries enormous volumes of traffic. And at the longest distances, submarine cables use optical amplifiers spaced every 60 to 80 kilometers along the ocean floor to keep signals strong across thousands of kilometers. Every layer of the internet, from the local to the global, depends on light moving through glass.