Summary

  • RIPE NCC’s 2027 draft raises planned capital expenditure to €10.48m from €0.93m in the 2026 budget; €7.68m is allocated to infrastructure renewal, with €1m for migration and overlap.
  • A budget line can show capacity purchased, but it cannot establish service resilience. The useful public evidence is a service-level map of dependencies, cutovers, recovery tests, decommissioning and recurring cost.

The renewal programme is already under way and is scheduled to run through 2028. Its 2027 plan is more concrete than the initial €5m estimate discussed earlier: roughly €7.7m in hardware capital expenditure for two data centres, €1m in migration costs, and an expected recurring operating cost of about €1.4m a year once both centres are established. The draft says 2027 operating costs are €1.2m because not all costs begin in January; that figure excludes depreciation. The exact capital/operating split may change as the project develops. Draft Activity Plan and Budget 2027

Those figures describe inputs and timing. They do not, by themselves, show which services will stop depending on a shared cloud architecture, what each service will depend on after migration, or whether the new design removes a meaningful failure mode. “Two data centres” is an asset count. Resilience depends on how power, network paths, identity, control planes, backups, staff access and third-party services are separated—and on whether a service can actually operate when one of those domains is unavailable.

That makes a service-level resilience ledger a practical accountability tool. For every critical service, it would identify the current and target locations, material dependencies, migration window, rollback condition, recovery objective, test date and evidence, and the point at which old capacity can be retired. The ledger should distinguish a planned test from a passed test and a passed test from sustained operation. Where details are security-sensitive, a public summary can report coverage and exceptions without publishing exploitable configuration.

The financial bridge matters just as much. The draft attributes about €1.5m annual depreciation to €7.7m of renewal hardware over five years, while the project section estimates €1.4m annual recurring operating cost, excluding depreciation. These are different cost categories and should not be added or compared as if they were the same measure. Migration overlap is temporary only if the old environment is actually decommissioned; a dated retirement decision and actual-versus-budget cost series would show whether transition spending declines as intended.

There is also a small but material reconciliation question in the publication. The detailed financial table lists 2027 expense per LIR at €2,209, against a €2,056 2026 budget and a €2,016 2026 forecast. The 30 September announcement instead calls €2,056 the 2027 amount while describing it as a 7% rise over the 2026 budget. The draft therefore needs a corrected, consistent member-facing figure before readers use it to judge affordability. This discrepancy does not establish why it occurred.

The decision is not whether renewal is desirable in the abstract. It is whether the proposed architecture and expenditure produce verifiable independence for the services members rely on. Publishing the service map, test coverage, exceptions and cost bridge would let members assess that claim without disclosing sensitive implementation detail.

Sources