Latest
Subscribe

How photonic chips move data using light instead of electrons

Photonic chips route information as pulses of light through silicon rather than as electrical current, and that simple swap could solve some of computing's biggest bottlenecks.

quantum computer laboratory
Photo · Anita Fors (Chalmers) / Wikimedia Commons (CC BY-SA 4.0)

The problem electrons cause

Every conventional computer chip works by pushing electrons through tiny wires etched into silicon. This has powered decades of progress, but electrons have drawbacks. They generate heat as they move, because the wires resist their flow. They interfere with each other when packed close together. And beyond a certain speed, sending electrical signals down a wire starts to behave less like a clean pulse and more like a smear, which limits how fast you can reliably send data.

These effects become a serious headache inside data centres, where huge numbers of chips need to talk to each other constantly, and inside AI accelerators, where moving data between memory and processors is often the actual bottleneck, not the calculations themselves. Photonic chips are one answer to that problem: instead of asking electrons to carry information, they ask photons, particles of light, to do it.

What a photonic chip actually is

A photonic chip looks similar to an ordinary silicon chip under a microscope, but instead of wires it has waveguides: microscopic channels etched into the material that guide light much like an optical fibre does, just on a chip instead of running between buildings. Light is generated by a laser, usually a separate small component attached to or integrated near the chip, and then split and steered through these waveguides.

Information is encoded onto the light by modulating it, rapidly switching its intensity, phase or wavelength in a pattern that represents data, similar in principle to how a lighthouse could send Morse code by flashing on and off. At the receiving end, a detector converts the light pulses back into an electrical signal that the rest of the chip’s circuitry can use. This matters because most photonic systems today are hybrid: they use light for the parts of the journey where light is best, and electronics for logic and computation, where transistors remain more practical.

Why light is genuinely better for this job

Light has three properties that make it attractive for moving data. First, it travels extremely fast and suffers far less signal loss over a given distance than electrical current in a thin wire, so a light-based connection can carry a strong, clean signal further without needing to be boosted.

Second, light beams do not interfere with each other the way electrical signals in adjacent wires can, through a phenomenon called crosstalk. You can send multiple separate streams of light down the same tiny channel simultaneously by using different wavelengths, a technique called wavelength-division multiplexing, dramatically increasing how much data a single connection can carry at once.

Third, moving data with light generates much less waste heat than pushing it through resistive metal wires. Since cooling is one of the largest running costs and engineering challenges in a modern data centre, any technology that reduces heat output while increasing data throughput is commercially significant, not just academically interesting.

Where photonic chips are actually being used

The most mature application is in the connections between chips and racks inside data centres, and in the undersea and long-distance fibre optic cables that carry the internet’s traffic between countries, which have used light for the long-haul journey for years. What is newer is bringing photonics much closer to the processor itself, sometimes called co-packaged optics, where light-based connections replace the electrical links between a chip and its nearest neighbours, rather than light only being used for the long journey across a data centre or between cities.

Research is also progressing on doing actual computation with light, not just moving data. Certain mathematical operations that are central to training AI models, particularly matrix multiplication, can in principle be performed optically, extremely quickly and with low energy use. This is more experimental and further from everyday deployment than photonic data links, but it is a serious area of research because AI workloads are pushing electrical chip designs towards their practical limits on power and heat.

Why this counts as UK-relevant deep tech

The UK has research strength in photonics through university groups and specialist companies working on integrated photonics and optical communications, building on the country’s long history in fibre optics and telecoms research. Photonic chips also sit squarely inside the wider strategic conversation about semiconductor supply chains, since manufacturing them well requires precision fabrication skills that overlap with, but are not identical to, conventional chipmaking.

What to watch for, and what to check

As with any emerging hardware technology, be cautious about bold claims of chips that are simply “faster” or “greener” without detail on what specific function is being replaced with light and what remains electronic underneath. Most real systems today are hybrid, and full all-optical computing remains a longer-term goal rather than a shipping product. For current, verified information on UK research funding and strategy in this area, check UK Research and Innovation, and for context on the broader semiconductor and photonics supply chain, the Department for Science, Innovation and Technology publishes updated policy material worth reading alongside any specific technical claim.

Sources