The chip relies on two pairs of ring-shaped microresonators to achieve unmatched performance. Université Laval
Scientists at Canada’s Université Laval have developed a tiny yet game-changing optical chip capable of transmitting 1,000 gigabits per second while consuming only a fraction of the energy used by conventional systems.
The research team at the Centre for Optics, Photonics and Lasers (COPL) set out to address the high energy demands of artificial intelligence systems like ChatGPT by developing a unique, energy-efficient optical chip.
The revolutionary photonic device, which is as thin as a strand of hair, harnesses the power of light instead of electricity as its primary medium for data transmission.
It can reportedly transmit massive amounts of data at ultra-high speed, reaching 1,000 gigabits per second (Gbps), marking a significant leap from current systems that max out at around 56 Gbps.
As per the researchers, that’s enough to transfer the equivalent of 100 million books in under seven minutes, the exact amount time it takes to brew a cup of coffee.
Tiny but full of potential
The chip uses the phase of light, in other words, its shift, to operate and transmit information more effectively. By adding this new dimension to the signal, the researchers achieved unprecedented levels of performance.
“We’re jumping from 56 gigabits per second to 1,000 gigabits per second,” Alireza Geravand, a PhD student at the Université Laval and first author of the study explained.
But speed is only half the story. What makes this chip truly groundbreaking is its ability to achieve that speed while consuming just four joules of energy, the same amount needed to heat a single milliliter of water by 1.8 degree Fahrenheit (one degree Celsius).
Alireza Geravand, a PhD student at the Université Laval working on the tech in the COPL laboratories.Credit: Dany Vachon / Université Laval
The breakthrough relies on microring modulators, which are tiny ring-shaped devices made from silicon that manipulate light to encode information. The system uses two pairs of these modulators, one of which to control the light’s intensity, while the other to adjust its phase.
This dual-channel approach allows for far greater bandwidth in a much smaller physical footprint. Geravand explained the chip offers a promising solution for AI data centers.
“At 1,000 gigabits per second, you could transfer an entire training dataset – the equivalent of over 100 million books – in under seven minutes,” he continued. “That’s about the time it takes to brew a cup of coffee.”
Current data centers rely on tens or even hundreds of thousands of processors that must constantly communicate, and often over long distances. Even though each processor measures just a few millimeters, collectively, the infrastructure grows rapidly, and so does the energy required to power it.
“You end up with a system that’s kilometres long,” the PhD student highlighted, adding that with the new technology, the units can communicate quickly and efficiently, as if they were only a few meters apart. This is a great advantage as demand for AI continues to rise.
While the technology is still in the lab, commercial applications may not be far off. Companies like NVIDIA have already begun using microring modulators, though current systems still rely solely on light intensity. However, the team’s innovation adds a brand new layer of functionality, thus bringing society one step closer to light-speed AI.
“Ten years ago, our lab laid the groundwork for this technology,” Geravand concluded in a statement. “Today, we’re taking it to the next level. Maybe in a few years, the industry will catch up, and this innovation will make its way into the real world.”
The study has been published in the journal Nature Photonics.
MasterCard