The AI-created chips were more efficient and took an new approach that no human designer would have ever considered.

Boffins from Princeton Engineering and the Indian Institute of Technology published their groundbreaking findings in Nature Communications, revealing how AI can revolutionise chip design.

They focused on millimeter-wave (mm-Wave) wireless chips, the intricate components used in 5G modems. These chips are notoriously difficult to design due to their complexity and need for miniaturisation. Traditionally, manufacturers rely on human expertise, trial and error, and established templates—an inefficient and slow process.

The AI, however, used an inverse design method and developed the inputs and parameters independently. This allowed it to create radical new chip structures, disregarding old design templates that may have hidden inefficiencies.

Lead author Kaushik Sengupta said: "The resulting structures look randomly shaped. Humans cannot understand them."

But the real shock came when Sengupta’s team manufactured the AI-designed chips. The results? Performance levels beyond anything seen before. AI has outpaced human engineers in speed and efficiency.

Sengupta stressed that AI is not yet ready to replace human designers entirely. Many AI-generated designs did not function properly, mirroring the ‘hallucinations’ often seen in generative AI tools.

“The point is not to replace human designers with tools. The point is to enhance productivity with new tools.”

The breakthrough could transform chip design, allowing for custom chips tailored for energy efficiency, raw performance, or broader frequency ranges. With wireless chips becoming ever more critical, the implications of this research are vast.

“This is just the tip of the iceberg regarding what the future holds for the field,” Sengupta added. If AI can expand into other areas of circuit design, it could fundamentally change how electronics are created forever.