Brain-Inspired Chips Could Slash AI’s Soaring Energy Appetite

As artificial intelligence’s power consumption threatens to double by 2026, an international team of researchers has unveiled an ambitious roadmap for brain-inspired computing technology that could dramatically reduce AI’s environmental footprint while boosting its capabilities.

The comprehensive review, published January 22 in Nature by 23 leading experts from academia and industry, outlines how neuromorphic computing—chips that mimic the brain’s architecture—could revolutionize everything from smartphones to smart cities while using just a fraction of the energy of conventional systems.

“Neuromorphic computing is particularly relevant today, when we are witnessing the untenable scaling of power- and resource-hungry AI systems,” explains Gert Cauwenberghs, Distinguished Professor in UC San Diego’s Department of Bioengineering and one of the paper’s coauthors.

The field appears to be reaching a critical moment. “We are now at a point where there is a tremendous opportunity to build new architectures and open frameworks that can be deployed in commercial applications,” says Dhireesha Kudithipudi, the Robert F. McDermott Endowed Chair at the University of Texas San Antonio and the paper’s corresponding author.

The potential applications span an impressive range, from scientific computing and artificial intelligence to augmented reality, wearable devices, smart farming, and urban infrastructure. Recent breakthroughs have already demonstrated the technology’s promise—in 2022, a team led by Cauwenberghs developed a chip that could run diverse AI applications using dramatically less energy than conventional systems while maintaining equivalent accuracy.

But to achieve widespread adoption, the researchers argue that neuromorphic systems need to better emulate one of the brain’s key features: its selective pruning of neural connections. The human brain initially forms numerous connections before strategically eliminating most of them, a process that optimizes both spatial efficiency and information retention.

“The expandable scalability and superior efficiency derive from massive parallelism and hierarchical structure in neural representation,” Cauwenberghs notes, describing how the systems combine dense local connections within core units (like the brain’s gray matter) with sparse long-distance communications (similar to white matter).

The impact could be substantial. At the San Diego Supercomputer Center, where new computing architectures are regularly evaluated, researchers see immense potential. “This publication shows tremendous potential toward the use of neuromorphic computing at scale for real-life applications,” says Amitava Majumdar, director of Data-Enabled Scientific Computing at the center and a coauthor of the paper.

The roadmap emphasizes that success will require extensive collaboration between academia and industry, along with development of more user-friendly programming tools to make the technology accessible to a broader range of developers and researchers.

Rather than proposing a single solution, the authors envision a spectrum of neuromorphic hardware designs tailored to different applications. This flexible approach could help accelerate adoption across various sectors, from healthcare and robotics to environmental monitoring and autonomous systems.

Progress is already underway. Last year, Cauwenberghs and Kudithipudi secured $4 million from the National Science Foundation to launch THOR: The Neuromorphic Commons, a pioneering research network that will provide open access to neuromorphic computing hardware and tools, fostering collaborative innovation in this rapidly evolving field.