Archive · 12 Nov 2025
World's First Microwave Neural Network Chip Outperforms Traditional CPUs
Imagine a new kind of computer brain chip that works like magic with tiny radio waves called microwaves, instead of regular electricity signals.
This new chip can think much, much faster than the chips in most computers, like going from a bicycle to a race car.
It's designed for jobs that need super-quick thinking, like understanding what radar sees.
The chip uses special patterns of these waves, like a special ruler, to measure things very fast and accurately.
It's like the chip has a simplified brain that can learn and figure things out without needing lots of extra parts or power.
It uses very little energy, about as much as your phone uses to send a signal, making it great for small devices.
Scientists want to make it even smaller and better for teaching computers new things.
Researchers have developed the first fully functional Microwave Neural Network (MNN) chip.
The MNN chip operates using microwaves and can outperform traditional CPUs in processing speed.
It processes data streams at tens of gigahertz, significantly faster than typical CPUs operating at 2.5-4 GHz.
The chip boasts remarkably low power consumption, using less than 200 milliwatts.
This new technology is suitable for high-bandwidth applications and personal devices due to its speed and efficiency.
- Who
- Researchers from Cornell University
- What
- Developed the world's first fully functional Microwave Neural Network (MNN) chip capable of outperforming traditional CPUs.
- When
- Published on August 14
- Where
- Cornell University (researchers), Nature Electronics (publication)
- Why
- To achieve exceptionally fast processing speeds for high-bandwidth applications like radar imaging, by leveraging analogue microwaves which handle data at higher rates than digital systems.
Microwave Neural Network (MNN) Chip
Traditional CPUs
Processing Speed
Microwave Neural Network (MNN) Chip
Operates on tens of gigahertz (tens of billions of operations per second) using analogue microwaves.
Traditional CPUs
Operates between 2.5 and 4 GHz (billions of operations per second) using digital circuitry.
Power Consumption
Microwave Neural Network (MNN) Chip
Consumes less than 200 milliwatts (0.2 watts).
Traditional CPUs
Requires at least 65 watts.
Design and Functionality
Microwave Neural Network (MNN) Chip
Leverages analogue waves in the microwave range within an AI neural network, using interconnected electromagnetic nodes and tunable waveguides. Processes spectral components and can be repurposed for multiple computing tasks by distorting frequencies.
Traditional CPUs
Uses traditional digital circuitry. Mimics the structure of digital neural networks.
Accuracy and Efficiency
Microwave Neural Network (MNN) Chip
Achieves 88% accuracy rate for wireless signal classification, recognizes patterns and adapts dynamically without additional circuitry, energy, or error correction for accuracy.
Traditional CPUs
Maintaining accuracy often demands additional circuitry, energy, and error correction.
- Chip Type
- Microwave Neural Network (MNN)
- Operating Frequency
- Microwave analogue range (tens of gigahertz)
- Operations Per Second
- Tens of billions
- Power Consumption
- Less than 200 milliwatts (0.2 watts)
- Accuracy Rate
- 88% (for wireless signal classification)
- Published In
- Nature Electronics
- Lead Author
- Bal Govind (Cornell University)