Light-based computers are getting close to a commercial launch

Computers that use light rather than electricity to represent and manipulate data could slash the power demands of data centres and simultaneously speed up calculations. Two studies published today describe breakthroughs in running real-world problems on light-based computers, making the technology on the verge of commercial application, say researchers.

Electronic computers, like those we all use today, historically followed Moore’s Law: the power of machines doubled every two years. But in recent years, progress has slowed as the miniaturisation of transistors hit fundamental physical limits.

Researchers are working on numerous potential solutions, including quantum computing and photonic computing. But while quantum computing still struggles to achieve true usefulness, photonic computing has now reached a point where chip designs, such as those set out in the two new studies, are performing genuine calculations. On top of that, the same factories that manufacture silicon chips for electronic computers can be used to fabricate these photonic chips.

Photonic computers offer huge potential advantages over electronic computers. For one, because photons move faster than electrons do in a circuit, they could speed up calculations and also reduce the pauses between each step of a calculation. Secondly, because photons move without resistance and are rarely absorbed by the material the chips are made of, they could do the same job using less energy than electrical computers, which require energy-intensive cooling.

Lightelligence, a Singapore-based company, shows in its study that its device, called a photonic arithmetic computing engine (PACE) that combines a photonic chip and a microelectronic chip, can successfully run Ising problems, which have direct applications to the logistics industry and many other areas.

Meanwhile, US start-up Lightmatter claims its own chip, Envise, can run the AI model BERT to create text in the style of Shakespeare, with an accuracy rate similar to that of conventional electronic processors. New Scientist was unable to reach Lightmatter for comment.

Bo Peng at Lightelligence says the field is increasingly busy with start-ups and that the technology is rapidly maturing. “We are more or less in pre-production,” says Peng. “It’s more like a real product, not just a lab demonstration.”

Just as the world of quantum computers is trying to demonstrate quantum advantage – the point where a quantum machine can offer something useful that a classical computer cannot – Peng says he is focused on demonstrating photonic advantage. Although he won’t be drawn on when this will occur, he says the technology is close to being ready for commercial applications – most likely as a photonic chip that works alongside electrical chips, rather than entirely replacing them, to handle specific tasks where it can provide a boost.

Tellingly, the hardware on which Lightelligence based its research is made in a PCI Express format. This is the standard motherboard add-on format for desktop computers that allows graphics cards or other devices to be added. The company’s device can already be added to any commercial desktop – although it would need the right software to communicate with it.

Robert Hadfield at the University of Glasgow, UK, says the two studies show “this is an area which is kind of coming to the boil”. “It’s close to a point where the industry may consider photonic processors as a viable alternative,” he says. “It’s really interesting to see how much this architecture has matured. And these are photonic chips which are fabricated in one of the world’s leading foundries, so it can be scaled up for mass production.”

Stephen Sweeney, also at the University of Glasgow, says that we have already seen optical data transmission rolled out around the world with fibre optics, and that photonic computing is now also close. “Photonics allows you to do things with higher speed and lower loss than you can do with electronics,” says Sweeney. “And if you need to be able to do vast amounts of computation, you need to start looking at that.”