Published June 5, 2026 in Nature Communications. Researchers from UMass Amherst, MIT, and Cambridge prove that asynchronous neural networks — updating one randomly selected neuron per step without global clocks — can achieve Turing universality under specific design constraints.
Asynchrony eliminates global updates and reduces energy use. The paper proves universality for both asynchronous fixed architectures with varying-precision neurons and variable architectures with fixed-precision neurons.
These results advance theoretical understanding of asynchronous networks, suggesting they preserve full computational power, remain amenable for efficient training, and may achieve substantial reductions in energy use for large-scale neural network deployment.
DOI: 10.1038/s41467-026-73830-6. Accepted May 21, 2026.