DARPA Launches ScAN Program to Develop Analog Neural Networks for AI Efficiency

inductionheads · x · 2026-08-07

To address the significant power consumption issues of current digital neural networks, DARPA has officially launched the Scalable Analog Neural-networks (ScAN) program.

Today's digital AI systems face energy bottlenecks that limit the deployment of advanced AI in size-, weight-, and power-constrained environments. The ScAN program aims to achieve orders-of-magnitude improvements in energy efficiency by designing analog neural networks that connect directly to analog sensor outputs, eliminating the need for power-hungry analog-to-digital converters. Launched in 2025, the 54-month, two-phase effort will scale up to large networks.

Original post →

More from Infra

Infra channel →