Juni

Now, after training in simulation, they transferred the learned controller to a real machine with safety limits wrapped around it.

unreal plasma shots or runs, the system would create and maintain several different shapes that matter for fusion research.

And it could switch between them during the shot while staying within the safe operating envelope.

On the industry side, CFS is trying to move from experiments to power plants.

SPARC is their next device, built around high-temperature superconducting magnets and designed to show net energy gain.

Their published roadmap describes SPARC as a bridge to ARC, which is meant to be a grid-connected, commercially relevant plant.

In October, the AI alliance with Google DeepMind means that Spark is designed and commissioned.

And the AI will be part of the toolbox for simulation, control, and optimization from the beginning, not just bolted on after the hardware is finished.

And that's a meaningful shift in how these systems are conceived.

Let's zoom out from fusion, and let's look at the wider pattern, or what I like to call the red string theory.

Grid researchers talk about autonomous energy systems, where AI agents respond to changes in load and renewables to keep frequency and voltage within systems.

Industrial vendors are demoing autonomous plant control for refineries and process plants using digital twins and learned control policies.

National labs are building self-driving material labs where AI chooses which experiment to run next and then robots to carry it out.

And then weather agencies are adopting AI-based models that can forecast storms and temperature patterns as well as, or better than, some traditional physics-based models, but at a far lower compute cost.

So in each case, you have a digital model, a learning system on top of it, and a gradual shift from suggestion to direct control.

So here's the reflective part.

First, we have the energy loop.

AI and data centers consume a growing share of electricity, and the same companies build the largest models are now investing in and helping to control and advance energy sources like fusion.

There's a feedback loop where AI both pushes up

demand and offers tools to match that.