Results from couple of efforts I have been working with Microsoft Research (Weiwei Yang, Andrea Britto Mattos Lima, Thiago Vallin Spina, Spencer Fowers) have been annouced:

  1. We released an open dataset of approximate U.S. transmission-grid topologies built entirely from public sources like OpenStreetMap, spanning 48 states and interconnection-scale networks up to the 21,697-bus Eastern Interconnection, with every model solvable under AC optimal power flow. Because the models are geographically grounded and electrically coherent, they enable physics-based studies that price data alone cannot answer, such as where new transmission can physically fit, where superconducting links would relieve urban congestion, and where large new datacenter loads can be sited without triggering overloads. Full methodology is in the companion paper, with code and models available on GitHub and Hugging Face.

  2. We introduce GridSFM, a small foundation model that approximates AC optimal power flow in milliseconds across grids from 500 to 80,000 buses, unlocking decisions that can impact up to $20B/year in congestion losses and 3.4 TWh of renewable curtailment. Unlike conventional surrogates that require retraining per topology, a single GridSFM model generalizes across 150+ grids and produces full AC operating points plus a feasibility verdict, running roughly 1,000× faster than a full AC solver and ~100× faster than the DC-approximation baseline. The open-research tier is available now on GitHub and Hugging Face, with more details on the project page.