Takeaways

• Neuromorphic chips inspired by the brain could cut AI’s growing energy use.
• These adaptive chips process data more efficiently, reducing reliance on power-hungry data centres.
• Researchers stress the need to address ethical concerns as hardware itself begins to “learn.”

Artificial intelligence (AI) is expanding at breakneck speed, but its hunger for energy is becoming a serious challenge. A single AI prompt can use as much energy as charging a phone for ten minutes, says Wilfred, a researcher at the University of Twente. The real cost of AI is hidden in massive data centres, yet the impact is being felt everywhere in rising energy bills.

Instead of building bigger, faster servers, Wilfred’s team is developing neuromorphic chips, devices designed to mimic how the human brain processes information. These smart chips promise to do more with less energy, offering a radically different approach to AI’s energy problem.

Read More: Understanding AI Pollution: Environmental Impact and Sustainable Solutions

Smarter Chips That Mirror Neurons

At the NanoElectronics group in Twente, researchers are designing analogue, trainable processors that behave more like networks of neurons than traditional silicon chips. Unlike conventional hardware, their behaviour isn’t fixed. They can be tuned after production, much like a neural network learns.

By using reconfigurable nonlinear processing units, these neuromorphic chips perform complex operations more efficiently. Instead of repeatedly fetching data from memory, “the physics does the computing for you,” explains Wilfred.

One promising application is speech recognition. The team has filed a patent for a chip that processes raw audio directly. By skipping the heavy upfront signal processing, the system can run high-quality speech recognition with far smaller digital models, saving energy without sacrificing performance.

When Hardware Starts to Learn

The idea of hardware that can adapt and “learn” sounds revolutionary, but it comes with ethical questions. “We are developing physical systems that can adapt and learn, essentially bringing artificial intelligence into the hardware itself,” Wilfred notes. This raises issues of control, responsibility, and unintended consequences. To tackle these, the group collaborates closely with ethicists and philosophers.

From Data Centres to Everyday Devices

Neuromorphic chips are not meant to replace digital computing. Instead, they could complement existing systems, especially in data-intensive jobs where energy use matters. Wilfred points to potential applications in smartphones, cars, and even medical implants. Because these chips are lightweight, fast, and don’t require a constant internet connection, they are well-suited for real-world environments.

Also Read: How Chemistry Could Keep AI’s Energy Demands from Harming the Climate

Shaping AI’s Future Responsibly

As AI hardware evolves, its societal impacts cannot be ignored. Wilfred believes the discussion about ethics and responsibility must happen now, while the field is still emerging. “We do not want to build a powerful technology and only later ask whether we should have done it differently,” he says. "That conversation has to happen now, as the field is still taking shape."

Neuromorphic chips may hold the key to balancing AI innovation with sustainability, but ensuring they are developed responsibly will be just as important as the technology itself.

Follow more news and views via our ESG Tech and Featured Articles sections, and stay updated on the top ESG events to attend in 2025 for industry insights and networking.

Source: University of Twente