Tech Transfer Central
Tech Transfer eNews

Tech Transfer E-News provides a weekly round-up of current news and information in the world of tech transfer, delivered every Wednesday (sign up here). It is published by Technology Transfer Tactics newsletter, which is available as a monthly subscription. For more information or to order a subscription click here, or for a sample issue, click here.

U College London start-up’s technology addresses rising electricity usage in the age of AI

A start-up from University of College London (UCL) aims to address the rising levels of electricity being consumed as power-hungry AI technologies continue to evolve.

Oriole Networks recently raised £10 million in seed funding to further develop its technology, which uses light beams running through optical fibers to speed up the rate at which packets of information can travel between GPUs by up to 100 times, according to the start-up.

GPUs, which are used in AI training, are more commonly connected through ethernet cables, which can lead to a bottleneck when the amount of information passing through them rises to a certain point. The slower a network’s computing power becomes, the more energy is consumed. Optical networks, on the other hand, use far less energy, and Oriole’s optic-based solution has been shown to reduce network energy consumption from 20% of an AI cluster’s energy needs down to 2%-3%.

The start-up licensed the technology from UCL Business, the university’s tech transfer arm. Once product development is complete, Oriole plans to begin selling its technology to customers, which may take a couple of years, according to CEO James Regan.

“Now we’ve actually got to go build it,” Regan says. “All the simulations and the prototyping and testing in the lab have shown that this works.”

The £10 million seed round was led by investors including XTX Ventures, Clean Growth Fund, and UCL’s tech transfer arm.

Source: Sifted

Posted under: Tech Transfer e-News