Physicists at Loughborough University have developed a new computer chip that processes data directly in hardware. The chip can make some AI tasks up to 2,000 times more energy efficient than conventional software. It solves the growing problem of AI systems consuming too much electricity.
Loughborough University researchers created a device that handles time-dependent data — like weather patterns or sensor readings — without relying on software running on standard computers. Their study was published in Advanced Intelligent Systems.
READ ALSO: 5-Hour Flight: Solar Drone Uses 110W Power and 1 Smart Backup Mid-Air
Many real-world AI problems involve predicting what will happen next in dynamic systems, such as weather forecasting or biological processes. A technique called reservoir computing is normally used for this, but it runs on software and uses a lot of energy.
The new chip is a type of memristor made of nanoporous oxide. It contains random microscopic holes that create multiple electrical pathways. These pathways act like the hidden processing layer of a neural network, letting the material itself do part of the computation.
The team tested the chip on a famous chaos model linked to the “butterfly effect,” as well as recognizing simple number images and basic logic operations. Dr. Pavel Borisov, senior lecturer in physics, who led the research, said the chip successfully predicted short-term chaotic behavior and reconstructed missing data. The same device handled all these different tasks.
Tests were done on relatively simple problems. The system is still in an early stage, and more work is needed to scale it up for noisy, real-world data. Dr. Borisov said the next steps include increasing network complexity and testing with more signal noise.
WATCH ALSO: China sets world record, maglev train reaches 700km/h speed in 2 seconds
Why Does This Matter?
AI’s energy demands are rising fast, raising long-term sustainability concerns. By shifting computation from software to hardware, similar results may be possible with far less electricity. Dr. Borisov said this is a scalable approach to creating small, industry-ready AI devices with much better energy efficiency.
