Optical computers: everything you need to know

Optical computing - the light fantastic?
24 May 2023

• Optical computers work through photonic transfer.
• They could be fast, with minimal heat loss during transfer.
• There’s controversy over the promises of photonic technology.

Optical computing is fast becoming a major player, especially in the realm of AI. You’d be forgiven for never having heard of it, but it involves lasers and light-speed, so why not find out more?

What is optical computing?

Optical computers, also known as photonic computers, perform digital computations using – you guessed it – photons. Light waves produced by lasers or incoherent sources are used as a primary means for carrying out numerical calculations, reasoning, artificial intelligence, data processing, data storage and data communications for computing.

Like any computer, an optical computer needs three things to function well:

  • An optical processor
  • Optical data transfer, such as a fiber optic cable
  • Optical storage.

The history of optical computing is interlinked with the development of radar systems. In the 1960s, the invention of the laser saw the first schemes for an all-optical computer proposed, and since the 1990s, the emphasis has shifted to optical interconnection of arrays of semiconductor smart pixels.

How is it different from traditional computing?

Traditional computers use electrons to carry out calculations, but photons have the capacity to enable a higher bandwidth; visible and infrared (IR) beams flow across one another without interacting, unlike electrons, so they can be constrained to what is effectively two-dimensional computing.

Three-dimensional wiring is necessary in traditional computers to direct electrical currents around one another. So, a photonic computer can be smaller than its more common counterpart. Like traditional computing, optical computers use logic gates and binary routines to perform calculations, but the way these calculations are performed differs.

Optical computing can achieve similarly efficient and reliable computation to the silicon channels and copper wires that enable electronic computers to function, by using plasmonic nanoparticles. Further, the absence of physical wires means that optical computers are less prone to damage from heat or vibrations.

Because photons can be easily manipulated and controlled, photonic computers are faster and more efficient. Photon movements can be guided and controlled in such a way that they can turn corners and carry on without a significant loss of power. Light can be easily contained and loses less information during travel, which is especially useful in situations where the interconnects might heat up, which slows electrons’ movement.

Photonics have a high throughput of >1TB/s per channel (of which there can be many in close proximity), compared to copper wire’s capability of 1GB/s per channel.

The hope is that the use of light or information shuttling will result in the development of exascale computers. Exascale computers could perform billions of calculations every second, 1000x faster than the current fastest systems.

So, we can weigh up the advantages and disadvantages of this alternative mode as follows:

Advantages of optical computing:

  • Fast density; small size; minimal junction heating; high speed; dynamic scaling and reconfigurability into smaller/larger networks/topologies; vast parallel computing capability, and AI applications.
  • Not prone to electrical short circuits and immune to electromagnetic interference.
  • Provides low-loss transmission and a lot of bandwidth, so multiple channels can communicate simultaneously.
  • Data processing on optical components is less expensive and simpler than data processing on electronic components.
  • Photons are able to pass across one another.
  • Optical materials are more accessible and have higher storage desnity than magnetic materials.

The disadvantages are:

  • It’s hard to develop photonic crystals.
  • Due to the interaction of several signals, computation is a complex process.
  • Current optical computer prototypes are bulky.


There are disagreements among researchers when it comes to the capabilities of optical computers. Whether or not they can compete with semiconductor-based electronic computers in terms of speed, power consumption, cost, and size is an open question.

Critics argue that real-world logic systems require “logic level restoration, cascadability, fan-out and input-output isolation,” all of which are currently provided by electronic transistors at low cost, low power, and high speed. For optical logic to be competitive beyond niche applications, major breakthroughs in non-linear optical device technology would be required, or even a change in the nature of computing itself.

Another option would be creating a hybrid system that integrates optical solutions into digital computing. However, there are impediments to the use of optics in digital computing “that perhaps demand a much more guarded view of the ability of optics to compete with digital electronics.”

Digital computing requires nonlinear elements to process digital data. The required functionalities of nonlinear elements are all delivered by transistor circuits in electronic computing. For large scalable logic circuits, no optical element or circuit, active or passive, can do all that and also compete with transistors in the metrics of energy consumption and small device footprint.

Fiber optics: already familiar?

In digital communications, fiber optic data transfer is already prevalent. Fiber optics use light for data manipulation. This is the area in which optical technology has advanced the most: it’s used enough that it’s already common in the lexicon of data transfer.

Fiber optic cables can contain a varying number of glass fibers, along which information is transmitted as light pulses. Fiber optic cables have advantages over copper cables, including higher bandwidth and transmit speeds. You might have noticed that these pros echo those of optical computing.

However, making the switch is much simpler when it comes to fiber optics cables, which are already used for internet, television and telephone connections.

Overcoming limitations

Areas of active research aiming to overcome some of the current limitations of photonic computing include:

  • High-speed integrated optoelectronic devices: Researchers are working on developing high-performance photodetectors, modulators, switches, amplifiers, and other components that operate at speeds approaching those of electronic devices.
  • Low-loss waveguides: Developing low-loss materials for guiding light through chips would allow for longer distances between components and reduce signal loss, increasing the efficiency of optical circuits.
  • Large-scale integration: One challenge facing optical computing is integrating large numbers of components onto a single chip without compromising performance.
  • Nonlinear effects: Nonlinear effects like self-phase modulation and four-wave mixing can limit the distance over which signals can be transmitted, and make it difficult to control the behavior of light within an optical circuit. Researchers are exploring methods to mitigate these effects and increase the range of optical communication systems.
  • Compatibility with existing technologies: Efforts are being made to develop hybrid optoelectronic systems that integrate optical elements with conventional electronic components, allowing for compatibility with existing infrastructure while leveraging the advantages of both types of technology.

A spinout of MIT,  Lightelligence is developing the next generation of computing hardware. Founded in 2017, the company claims to have “transformed the cutting-edge technology of photonics into groundbreaking computing solutions, which not only bring exponential improvements in computing power, but also dramatically reduce energy consumption.”

In basic terms, its research uses a silicon fabrication platform used for traditional semiconductor chips, but in a novel way. In the optical domain, arithmetic computations are done with physics instead of with logic gate transistors that require multiple clocks.

Yichen Shen, co-founder and CEO of Lightelligence, said that because the system it’s developing generates very little heat, it has a lower power consumption than electron-powered chips.

“We’re changing the fundamental way computing is done, and I think we’re doing it at the right time in history,” says Shen. “We believe optics is going to be the next computing platform, at least for linear operations like AI.”

Yes – like all of the tech world at the moment, optical computing has a vested interest in AI. However, instead of thinking about how artificial intelligence could help it, photonic computing might facilitate the further development of AI.

For example, self-driving vehicles rely on cameras and AI computations to make quick decisions. The conventional chip doesn’t “think” fast enough to make the split-second decisions necessary, so faster computational imaging is needed for quick decision making. That’s what Lightelligence says it’s achieving using photonics.

Photonic versus quantum computing

We couldn’t talk about radical changes to computational systems without touching on quantum computing. Due to the unique properties of quantum mechanics, quantum computing can solve problems beyond the capabilities of the most advanced computers, including photonic.

The area in which optical computing is ahead of quantum is the speed at which (simpler) calculations can be performed. In some cases, optical computing is faster than quantum. In many cases, optical computing is being researched for use in tandem with quantum computers. Both have the potential to revolutionize computation and data processing.

We’ve yet to see an optical computer, but we’re at the frontier of developments. Since 2012, Moore’s law (that the number of transistors in an integrated circuit doubles every two years) has been defunct: AI compute doubles every 3.4 months. We’ve come incredibly far, incredibly fast.

Photonic computers might be closer than we think.