The Cloud Does Not Need a Divorce. It Needs Boundaries.

Option A: Hand the hyperscaler your data, your controls, your recovery plan, and most of your future negotiating leverage. Trust the branding. Sign the enterprise agreement. Enjoy the keynote.

Option B: Declare technological sovereignty, convene a steering committee, and spend three years explaining why phase one is delayed but "strategically critical."

Neither is especially satisfying.

The real problem isn't that AWS, Azure, and Google Cloud exist. They're good at what they do. Combined, they held 63% of global enterprise cloud spend in Q3 2025. AWS alone did $128.7 billion in revenue last year. This is not a cottage industry.

The problem is what happens when one of them becomes your entire trust model.

At that point, you're not really buying cloud infrastructure. You're renting dependency with exceptional marketing.

Customers are finally asking better questions.

Not "which region is the data in?" or "can someone put a flag on a slide?"

But: Who can reconstruct this data? Who controls the operational dependency? Who becomes impossible to leave?

These are infrastructure questions wearing a suit. And once you start asking them, a lot of very polished sales language starts to sound like scented tape over a structural crack.

The hyperscalers have a problem too. They'd just prefer not to discuss it.

Gartner projects sovereign cloud IaaS spending will hit $80 billion in 2026 — up 35.6% — with roughly 20% of workloads shifting from global to local providers. That's not symbolic European protest. That's a spending trend with teeth and a spreadsheet behind it.

The risk isn't dramatic collapse. It's something finance departments find far more unsettling: a slow leak. A bit less European growth. A bit more discounting. A bit more compliance overhead. A bit more customization for sovereign deals.

Still revenue. Just not the kind that arrives with the same margins — or the same smugness.

Microsoft's cloud gross margin already fell to 69% in 2025 from AI infrastructure costs. Sovereignty requirements are now queuing up to add another layer of operational complexity, at the precise moment hyperscalers would prefer to sell standardized cloud plus AI plus a gently expanding sphere of dependency.

Bad timing, really.

The answer isn't rejection. It's redistribution of control.

Most organizations still want the infrastructure, performance, and ecosystem that hyperscalers provide. They just don't want it bundled with total dependency and the quiet realization that leaving later will require either a miracle or a very uncomfortable budget hearing.

This is where architectures like SkyeConnex change the conversation.

By splitting, encrypting, and distributing data across providers — so no single cloud holds a complete, usable copy on its own — it does something useful for both sides.

For the customer: real sovereignty, real portability, real bargaining power. Not a flag on a slide.

For the provider: a way to stay in deals they'd otherwise lose, without being asked to carry the entire trust boundary alone.

The conversation shifts from "We can't use you because you control too much" to "We can use you as part of an architecture where no one controls too much."

Better sentence for the customer. Better sales outcome for the provider.

Turns out reducing dangerous concentration is good for everyone — including the concentrated party.

Originally published by Ross Norrie, founder of SkyeConnex, on LinkedIn.

Published March 26, 2026 · More from the SkyeConnex blog