Summary

  • The ‘minimum necessary’ test doing the work in this debate belongs to a member, not to the registry: the retrieved draft does not use the phrase, and RIPE NCC’s own case is argued in depreciation arithmetic — EUR 7.7M of hardware capex inside a EUR 10.5M capital programme, about EUR 1.5M a year, roughly EUR 77 per LIR.
  • The like-for-like comparison that member asked for is not in the document members read: the roughly EUR 5M less-resilient alternative appears in Executive Board minutes, while the instruments reviewed here make the annual contribution an obligation covering the whole association and offer no reviewed mechanism by which a member declines the data-centre share.

A question the budget does not answer

RIPE NCC, the not-for-profit membership association that is the regional internet registry for its service region (directory entry), published its draft Activity Plan and Budget 2027 in September 2026 1. On the members-discuss-unmoderated list, a member, Kaj Niemi, asked for the comparison the draft does not contain: the total costs of the proposed approach set against ‘a minimum necessary renewal of the existing infrastructure’, together with the additional resilience each provides 2.

That sentence deserves more weight than the thread gave it, because the phrase has since travelled. ‘Minimum necessary’ is not how RIPE NCC describes its own plan. It is a member’s test, and a demanding one: not whether the chosen design is good, but what the cheapest credible alternative would cost and what resilience it would surrender. The draft text retrieved for this review did not contain the phrase 3; that absence is scoped to the extraction obtained here, not to the document as a whole.

The capital line, item by item

The draft sets 2027 capital expenditure at EUR 10.5M, against 930 kEUR in the 2026 budget 3. EUR 7.7M of that is hardware capex for two data centres, described as infrastructure renewal; a further EUR 2.2M is budgeted for office repairs and renovation, depreciated over nine years 3.

The 7.7 figure then performs a conversion its authors clearly want read as reassurance: over a five-year depreciation period it represents approximately EUR 1.5M of opex per year, which the draft puts at roughly EUR 77 per LIR 3.

That conversion is arithmetic, and sound arithmetic. It allocates a capital sum across an asset life and across the membership. What it cannot do is answer the member’s question, because it has no second column. Nowhere in the comparison does a reader see what a smaller renewal would have bought, or what the extra EUR 2.7M purchased in resilience.

The operating line starts in 2028, not 2027

Recurring operating cost for the two data centres is put at around EUR 1.4M per year once established, excluding depreciation. The 2027 budget carries EUR 1.2M, because not all of that cost starts in January 2027, and migration is budgeted at EUR 1.0M in 2027 including overlap with the current infrastructure 3. Staff replies quoted in the member thread confirm the sequence: EUR 1.2M additional opex in 2027, then EUR 1.4M annually excluding depreciation from 2028-2031 4.

So the two numbers carrying this debate are not one capex-and-opex pair. EUR 7.7M sits inside a EUR 10.5M programme that also funds office renovation. EUR 1.4M is a steady state beginning in 2028, while the year actually being priced carries EUR 1.2M. A reader who treats 7.7 and 1.4 as the two sides of a single decision is comparing scopes taken from two sections of one file.

The year being priced is a deficit year

2027 income is budgeted at EUR 42.5M against costs of EUR 44.2M: an operating deficit of EUR 1.7M and an overall deficit of EUR 1.1M 3. The draft states that membership fees are not the only lever available and points to the Clearing House reserve and to outstanding Charging Scheme Task Force recommendations 3. Quoted staff replies in the member thread put the reserve at around EUR 34M, project deficits in 2027-2028, break-even in 2029 and surpluses in 2030-2031, and say income needs to increase by about 3-5% a year to break even from 2029 4.

The fee, as the instruments write it

Here the shape of the accountability problem becomes visible. Article 7.1 of the Articles of Association obliges members to pay an annual contribution pursuant to the Charging Scheme 5. Article 3 places registration activities, coordination activities supporting stable internet operation, administration and new activities inside the association’s object 5. Non-payment suspends the member, including attendance and voting rights 5.

The Charging Scheme 2027 prices accounts and registrations rather than services: EUR 1,894 per LIR account, up from EUR 1,800, on a one-account-one-fee model, with EUR 75 added for Legacy IPv4 registrations held under direct contract, the first ASN included, EUR 50 for each additional ASN and a EUR 1,000 sign-up fee 6.

The registry’s own Charging Scheme Task Force stated the principle plainly: members generally do not pay for individual services but for the rights that come with membership, so a significant majority of income should derive from membership fees, with separate charging reserved for resource-intensive or bespoke services used by a smaller group 7. Its report recommends that the scheme cover all services with no optional extras 8 — a short excerpt whose surrounding recommendations were not inspected for this review.

The registry’s service page draws a different line, between member services (registration including transfers, the LIR Portal, RPKI) and community services (the RIPE Database, K-root, DNS services, RIPE Atlas, RIPEstat, RIS, IPmap, research, outreach, meetings and the Academy) 9. That is a functional distinction. It states no funding distinction, and none of the instruments reviewed here converts the functional line into a fee a member could decline.

The alternative exists, as a minute

The comparison the member asked for does exist — in Executive Board minutes, which record a roughly EUR 5M IT infrastructure option that would provide less resilience and was called far from ideal, against the goal of at least two fully independent data centres. The same minutes record 500 kEUR a year spent on capex over the preceding five years, which left the two data centres dependent on each other 10. Minutes record discussion; the meeting date and the approval wording were not visible in the excerpt obtained.

The judgement

Judgement: the draft makes one decision legible and its alternative invisible. The depreciation arithmetic is printed in the document members read; the cheaper, less resilient option is recorded in a boardroom document most of them do not.

Judgement: because the annual contribution is collective and non-optional in the instruments reviewed, the meaningful lever is not a service menu a member could decline. It is the timing and content of the vote, and the comparison placed in front of members before it.

Judgement: roughly EUR 77 per LIR is a defensible way to express a shared infrastructure cost. It is an allocation, not a price, and the draft should not let it stand as a member’s share of the decision.

What follows is narrow on purpose. A watchdog may ask whether the like-for-like comparison, once a member has requested it in public, is published where members actually deliberate; whether the resilience trade-off and its cost are stated in the document being voted on rather than in minutes; and whether the December version answers the question. What a watchdog may not do is campaign to force or block the programme. That choice belongs to the membership, and the instruments say so.

Nothing in the material reviewed here supports an allegation that a figure was misstated, a charge excessive, or an officer improper. The finding is about disclosure sufficiency — a process finding, and a fixable one. Print the second column.

What this review did not obtain

Six items, named precisely: the full text of the draft read page by page, since only a truncated extraction was retrieved 3; the final approved 2027 plan and budget, which the notice says is due in December 2026 1; the Dutch text of the Articles of Association, which the publisher says legally governs 5; the full context and any dissent around the ‘no optional extras’ recommendation 8; a dated, authored record of the General Meeting item at which the draft was presented, including whether a vote was taken 1; and any published comparison against the roughly EUR 5M alternative other than the minutes excerpt 10.

The consequence

The 2027 plan was published in draft, discussed at the General Meeting alongside RIPE 93 and in the Services Working Group, with the Executive Board to approve a final version for publication in December 2026 1. That timetable is the whole opportunity. A member who wants a like-for-like answer is arguing against a document already written and a decision not yet ratified; a registry that wants the argument settled has a cheap way to settle it, by publishing the comparison it already holds. Until it does, the honest description of the EUR 77 figure is that it tells a member what the chosen design costs them, not what the choice cost.