They're

doing in the first instance more so, I would say, grid compliance support.

So by balancing out not just that sudden drop of the complete data center, that's one subset of the use cases.

Another one is that very choppy and fast intermittent load.

You have some volatility in the load of these AI training models, but also in between their compute and their checkpoint save phase, you have these massive power gradients from 90% to 50%, let's say.

These are huge swings.

And coupled with the fact that now these data centers themselves are huge, it becomes a bigger issue for the grid.

So what happens if you scale that up and you have more fault current than you need?