[tucnak]

It says so on the tin. "Escape the chokehold of hyperscalers" is all that matters, really. Everything else will follow nicely from it. Compute density is so good these days, you don't even need major datacenter investment. There are modular DC designs that fit in a shipping container. You tow one around, connect power, fiber, cooling lines (to intercoolers in another shipping container) and that's it. You would be surprised how much can be accomplished with so very little. There are many advantages to this approach, like being able to bring up SCIF-equivalent inspectable spaces on the cheap, but considering we're all probably going to war sooner than later, it might as well become necessary. This is akin to how SAAB, and perhaps to a larger extent Ukraine, have changed airplane logistics.

Unless you're a hyperscaler yourself, hyperscaling is overrated.

[nkoren]

Great, now how do you make that actually happen at a political / regulatory / cultural level?

It's already an uphill battle, because humans in large organisations seem to have an innately conservative bias which says that "nobody ever got fired for choosing ${giganticEvilStatusQuoCorporation}". That, combined with the fact that the US hyperscalars have, I dunno, hundreds of billions of dollars worth of ability to put their thumb on the political and regulatory scales, make this an uphill battle. There will need to be a specific plan for leveling the playing field.

What is that plan?