World's first light-powered neural processing units (NPUs) could massively reduce energy consumption in AI data centers

Somebody holding the Q.ANT photonic processor
The Q.ANT wafer based on Thin Film Lithium Niobate enables photonic integrated circuits with high-precision, high-speed optical modulation, low noise and reduced thermal dissipation. (Image credit: © Q.ANT)

A light-powered computer chip designed to drive artificial intelligence (AI) data centers and make high-performance computing (HPC) more sustainable has entered production.

In a statement published Feb. 24, representatives from quantum computing company Q.ANT said its photonic AI chip could deliver a 30-fold increase in energy efficiency and a 50-fold boost in computing speed compared with conventional, silicon-based computer chips.

Pilot production of the new chip is now underway at IMS Chips in Stuttgart, Germany, where Q.ANT has invested 14 million euros ($15.1 million) to repurpose an existing semiconductor factory to fabricate its new, light-powered chip.

Because the chip is being produced on a repurposed facility instead of a specialist production line, the company believes it can bring the technology to market much more quickly. The chip can also integrate with the existing HPC servers, potentially accelerating adoption, Q.ANT representatives said.

"By 2030, we aim to make our photonic processors a scalable, energy-efficient cornerstone of AI infrastructure," Michael Förtsch, chief executive of Q.ANT, said in the statement.

Photonic computing

Photonic chips could solve a massive challenge faced by existing processor technology, particularly as AI and other data- and resource-intensive computing applications grow.

Traditional silicon chips control electrical signals using tiny switches called transistors. Photonic chips, by contrast, process data using light particles (photons), which are massless and can travel much faster than electrons do in conventional computer chips.

Photons don't emit heat in the same way electrons carrying an electrical charge do. As such, using photonic chips in applications involving complex, energy-intensive computations like AI could overcome the limitations of classic silicon chip architecture and thus vastly accelerate the computers' processing speed and reduce their energy consumption.

Related: 'Crazy idea' memory device could slash AI energy consumption by up to 2,500 times

"This comes at a critical time for the computing industry, as the exponential growth of AI and data-intensive applications will soon overwhelm the current data center infrastructure," Jens Anders, a professor at the University of Stuttgart and director and chief executive of IMS Chips, said in the statement. Anders added that the two companies aimed to establish "a scalable model for energy-efficient computing."

Q.ANT's chip is built using thin-film lithium niobate (TFLN), a crystalline compound applied to a wafer that forms the basis of the company's photonic chip. TFLN is increasingly catching the attention of photonics researchers and quantum scientists for its potential in next-generation computing. When an electric field is applied to the material, it can be used to control the speed and phase of light waves, thereby enabling it to modulate optical signals with extreme precision.

The pilot production line has been set up specifically to manufacture chips that incorporate TFLN, with Q.ANT aiming to fabricate 1,000 wafers per year.

"As AI and data-intensive applications push conventional semiconductor technology to its limits, we need to rethink the way we approach computing at the core," Förtsch said. "With this pilot line, we are accelerating time to market and laying the foundation for photonic processors to become standard coprocessors in high-performance computing."

Owen Hughes

Owen Hughes is a freelance writer and editor specializing in data and digital technologies. Previously a senior editor at ZDNET, Owen has been writing about tech for more than a decade, during which time he has covered everything from AI, cybersecurity and supercomputers to programming languages and public sector IT. Owen is particularly interested in the intersection of technology, life and work ­– in his previous roles at ZDNET and TechRepublic, he wrote extensively about business leadership, digital transformation and the evolving dynamics of remote work.

You must confirm your public display name before commenting

Please logout and then login again, you will then be prompted to enter your display name.