https://surrogatemodels.com/ News, papers, datasets, and funding and deals in Physics AI for engineers: Energy & fusion. en https://surrogatemodels.com/papers/tokamak-rl-controlhttps://surrogatemodels.com/papers/tokamak-rl-controlThu, 12 Nov 2026 09:00:00 GMTA reinforcement-learning algorithm (MPO) trains a small neural network in a free-boundary plasma simulator to turn magnetic measurements and shape targets into voltage commands. The trained network then runs on the TCV tokamak with no further tuning and commands all 19 control coils at 10 kHz.Energy & fusion https://surrogatemodels.com/papers/qlknnhttps://surrogatemodels.com/papers/qlknnThu, 12 Nov 2026 09:00:00 GMTQLKNN is a set of feed-forward networks that reproduce the heat and particle fluxes of the quasilinear gyrokinetic transport model QuaLiKiz from nine local plasma parameters. The training enforces known physics, such as a sharp critical gradient below which turbulent flux is zero. The networks replace QuaLiKiz inside the tokamak simulators JINTRAC and RAPTOR.Energy & fusion https://surrogatemodels.com/market/zenithon-seedhttps://surrogatemodels.com/news/zenithon-seedWed, 30 Sep 2026 09:00:00 GMTUK startup Zenithon raised $10 million from BACKED, Lunar, Seraphim, MMC and SOSV to build what it calls world models for extreme physics: AI models that learn how complex physical systems behave and predict the outcome of new designs, for fusion reactors, rockets and semiconductor fabs. The company says its models will explore a million design points in the time one simulation takes and will learn directly from experiments; this has not been independently benchmarked. Its advisers include Zongyi Li of NYU and Dan Brunner, co-founder of Commonwealth Fusion Systems.Energy & fusion https://surrogatemodels.com/market/quanscient-series-ahttps://surrogatemodels.com/news/quanscient-series-aTue, 26 May 2026 09:00:00 GMTFinland's Quanscient closed a €10 million Series A led by 55 North and B&C Group, with Maki.vc, Crowberry Capital, QAI Ventures and First Fellow Partners. Its code-driven cloud multiphysics platform targets hardware such as motors, antennas and fusion magnets, and the company says it is built to produce the volume of simulation data that machine-learning models need.ElectromagneticsEnergy & fusion https://surrogatemodels.com/market/thinklabs-series-ahttps://surrogatemodels.com/market/thinklabs-series-aTue, 31 Mar 2026 09:00:00 GMTThinkLabs AI, based in New York, closed a $28 million Series A led by Energy Impact Partners, with new investors NVentures (NVIDIA's venture arm) and Edison International, the parent of Southern California Edison. Existing investors GE Vernova, Powerhouse Ventures, Active Impact Investments, Blackhorn Ventures and Amplify Capital took part, along with an unnamed North American investor-owned utility. The company describes its product as physics-informed AI for grid planning and operations. It says utilities have used it to bring a month-long study to under three minutes and to run 10 million scenarios in 10 minutes, with more than 99.7% accuracy on grid power flow. The funds go to customer delivery and a broader product line.Energy & fusion