Researchers Build Stretchable Neuromorphic Arrays With 10,000 Transistors/cm² for On-Body Edge Computing
Updated
Updated · Nature.com · May 20
Researchers Build Stretchable Neuromorphic Arrays With 10,000 Transistors/cm² for On-Body Edge Computing
4 articles · Updated · Nature.com · May 20
Up to 10,000 transistors per square centimetre were integrated into intrinsically stretchable organic electrochemical arrays, addressing a key scaling bottleneck for wearable neuromorphic hardware.
The arrays showed synaptic behavior suited to edge AI, including linear and precise conductance programming, good retention and enough uniformity to run a stretchable artificial neural network in hardware.
Hardware tests processed health data for heart attack risk assessment and performed kernel convolution to locate cardiac propagation wavefronts, pointing to near-sensor analysis on the body rather than in the cloud.
The team also demonstrated reinforcement-learning algorithms on the soft circuits for robot navigation, extending the platform beyond health monitoring to soft robotics and other adaptive wearable systems.
When will this on-skin AI finally upgrade the smartwatches we wear every day?
How will we manage the ethics of wearables that constantly predict our fatal health risks?
Is processing sensitive health data on your body truly more secure than protecting it in the cloud?
In 2026, wearable technology entered a new era with the introduction of large-scale, intrinsically stretchable neuromorphic arrays built on organic electrochemical transistors (OECTs). These advanced arrays achieve densities of up to 10,000 transistors per cm², allowing complex computational power to be integrated directly into wearable devices. This breakthrough enables real-time processing of physiological data on the body, eliminating the need for cloud-based resources. As a result, wearable devices can now deliver immediate, energy-efficient, and private health insights, marking a significant leap forward in how personal data is processed and used.