The Cloud Mirage Ends for Europe's Internet Backbone
Here's the thing about cloud migration that nobody likes to admit until it's too late: you trade a monthly bill for a dependency, and dependencies have a way of becoming liabilities when the geopolitical wind shifts.
RIPE NCC — Europe's regional internet registry, the organization that actually keeps the DNS root hints working for half the planet — just announced it's walking away from its cloud-first strategy. Managing Director Hans Petter Holen laid out the reasoning in a blog post and a presentation at last month's General Meeting, and the short version is this: relying on US hyperscalers for critical internet infrastructure just doesn't sit right anymore, especially after the Trump administration made "reliability" feel like a conditional promise.
I've been tracking this space for years, and I'll say it plainly — RIPE NCC is doing the right thing, even if it hurts their balance sheet. The alternative was quietly accepting that Europe's most critical network coordination services lived at the mercy of a foreign government's trade policy. That's not resilience. That's hope.
And hope isn't an architecture.
What Happened to the Cloud-First Assumption
Like a lot of organizations, RIPE NCC adopted cloud-first as a sensible default. Move core services and databases to cloud providers, reduce hardware overhead, flex with demand. The math worked on paper. Opex replaced capex. Teams stopped buying servers and started buying subscriptions.
But here's where the trade-off got ugly. Holen was explicit about it: during the cloud-focused years, hardware that should have been refreshed got deferred. Some of it has reached end-of-life. Some of it has passed it. The capex/opex balancing act that looked elegant in a five-year model fell apart when the underlying assumptions about cloud reliability stopped holding.
"To start with, we will need to replace hardware that has reached, or in some cases passed, the end of its lifecycle," Holen said. "This is the result of trade-offs between capex and opex over the period in which we were focused on cloud deployments, as well as various assumptions and decisions about how this balance would evolve over the long term."
Translation: you can't outsource your infrastructure judgment and then act surprised when the vendor landscape changes. RIPE NCC assumed cloud would stay stable, geopolitically and operationally. The assumption was wrong.
Rebuilding On-Premises: The Hard Parts
So RIPE NCC is going greenfield. Full datacenter rebuild, on-premises, with the kind of capital expenditure that hasn't been seen since before 2020. Holen put a number on it: an additional €5 million, with a target completion date of 2028.
That's not a small project. The registry needs to think carefully about its datacenter footprint — how many facilities, where they sit geographically, and crucially, how to minimize interdependencies between them. If you've got three datacenters that all route through the same undersea cable landing station, you don't have redundancy. You have three copies of the same single point of failure.
Geographically redundant storage and backups are non-negotiable. Holen also flagged the need to pick virtualization platforms that limit vendor lock-in — because apparently learning that lesson once wasn't enough. The irony here is thick: RIPE NCC spent years moving away from on-prem to avoid vendor lock-in, only to discover that hyperscaler dependency is just vendor lock-in with better marketing.
The timeline is aggressive but not unrealistic for an organization that's been planning the pivot internally. Greenfield by 2028 means construction starts now, hardware procurement begins next year, and the old infrastructure gets retired in phases. It's going to be expensive. It's also going to be sovereign.
The Membership Fee Fight Nobody Saw Coming
Here's where the story gets genuinely interesting — and a little embarrassing for RIPE NCC's communications team.
To fund this rebuild, the registry needed to balance internal cost savings against membership fees. Holen acknowledged that some members are already grumbling about what they've paid in recent years and don't want to see more rises. Fair enough.
So RIPE NCC put a charging scheme vote to the General Meeting. The proposed alternative was a sliding scale: members with more internet number resources would pay more, and 74% of members would actually pay less under the new model. The math was defensible. The equity argument was strong.
And it lost. Barely.
51.1% voted to keep the flat fee. 48.9% wanted the sliding scale. A swing of just 35 votes flips the entire outcome.
Of 19,415 eligible members, only 3,421 registered to vote and 3,049 actually cast ballots. That's a 15.7% turnout — described as one of the highest on record, which says something about how engaged most members are with governance decisions. (Something, I suspect, is wrong with that metric.)
RIPE NCC has since published a blog post dissecting why the vote went the way it did, and the answers are telling. Misconceptions persisted — some members believed paying more would mean greater voting rights, an idea that was explicitly opposed by the registry. Consultations on different fee ideas were published at various stages, some abandoned after member input, but not everyone tracked which proposals were still live. Mixed messages, apparently, even when you're trying really hard to be clear.
The registry's conclusion is worth quoting: "Perhaps our initial assumption that many of you would prefer a tiered model was inflated. It is true there was a long-standing demand for it. But we also hear from people who believe in equality over equity when it comes to financial contribution. To them, the varying amounts of resources members hold shouldn't be a reason for them to pay accordingly."
Translation: the membership wanted fairness defined as sameness, not proportionality. It's a governance lesson that applies far beyond internet registries.
What This Means for Cloud Strategy Across Europe
RIPE NCC isn't alone in second-guessing cloud dependency. The broader pattern is clear: European organizations — especially those running critical infrastructure — are recalibrating after years of treating cloud as a default rather than a decision.
The EU's sovereignty push has been building for years. The DMA targeting AWS and Azure for stricter competition rules, national cloud strategies across member states, the whole alphabet soup of digital autonomy initiatives — it all points in one direction. RIPE NCC is just the latest high-profile organization to make it concrete.
For security and compliance teams watching this unfold, the implications are real. Cloud migration isn't wrong. But it's not neutral either. Every hour your critical services spend on a hyperscaler's platform is an hour you're betting that platform stays reliable, stays accessible, and stays outside the jurisdiction of governments that might not share your risk tolerance.
RIPE NCC's decision to rebuild on-premises with geographic redundancy, vendor-diverse virtualization, and a hard 2028 deadline is a statement: some infrastructure is too important to outsource your judgment on. The €5 million price tag and the capex returning to pre-2020 levels are the cost of that statement.
I think it's worth paying. Not because cloud is bad — it isn't — but because the alternative to a deliberate infrastructure decision is drift. And drift, in critical internet coordination, looks a lot like vulnerability.
The Bottom Line
RIPE NCC's reversal on cloud migration is a case study in what happens when geopolitical reality catches up with technology strategy. The registry isn't abandoning cloud entirely — it's abandoning the assumption that cloud should be the default for everything, regardless of risk profile.
The rebuild will take until 2028. It'll cost €5 million more than originally planned. Membership fees stay flat, at least for now, despite the sliding-scale proposal losing by 35 votes. Hardware needs replacing. Datacenter interdependencies need untangling. Virtualization platforms need selecting without repeating the same lock-in mistakes.
It's a lot of work. It's also exactly the kind of infrastructure decision that critical internet coordination demands. Europe's internet registry just proved that sovereignty isn't a slogan — it's a capex line item, and sometimes you have to write the check yourself.