The Defense Advanced Research Projects Agency's (DARPA) September 24 update describes its Mapping Machine Learning to Physics (ML2P) program. The program aims to relate machine-learning performance to physical electrical characteristics so designers can optimize not only for accuracy, but also for the useful performance returned by each joule.
At the tactical edge, that is not an efficiency exercise. It is mission engineering.
Originally published in 2024; substantially revised in 2026 to deepen the analysis and incorporate additional sources.
“Cloud to edge” is often presented as an infrastructure continuum: train and aggregate centrally, then move inference closer to users and sensors. That description is technically sound but strategically incomplete.
The important design question is not where the compute happens. It is where data becomes knowledge, where a recommendation becomes a decision, and which parts of that chain must continue to function when connectivity, latency, power, or trust changes.
Cloud-to-edge AI is a decision-placement problem expressed through architecture.