GM May 2026: Why Did People Vote the Way They Did?

The time has come once again to reflect on the results of another General Meeting. The May 2026 GM was held during RIPE 92 in Edinburgh, and was certainly a big one with an important vote on the Charging Scheme at its centre.

Next to the existing One LIR Account-One Fee model, we proposed a category model that would have seen 74% of the membership paying less. To many people’s surprise, the option that would have financially benefitted almost three-quarters of the membership fell slightly short of receiving half the votes cast.

So what factors shaped the decisions made by our members at this GM? In this article, we look at turnout, analyse voting patterns, and explore several possible explanations for the outcome.

Overview

Being a Spring GM, our agenda this time involved the usual resolutions on discharging the Board and adopting the Financial Report, as well as an Executive Board election to fill three seats.

More excitingly, a new charging scheme model was proposed for the membership. This proposal was the culmination of nearly two years of work involving a member-driven Charging Scheme Task Force, which worked to come up with principles that the charging scheme should be based on. In many ways, this reflects many more years of (often heated) debate about how the RIPE NCC should charge its members.

Model A (One LIR Account-One Fee) kept the existing model with an adjusted fee of €1,894, and Model B (Category Model) offered a tiered approach, with 21 categories determined by the amount of registered IPv4 and IPv6 allocations on each LIR account. Both models also included identical additional fees for registered ASNs, PI and legacy resources to help focus discussions on how the base fee would be applied. More information on both models was shared in advance by the task force. As stated in the intro, and as has been heavily discussed on the members-discuss list, voting members chose to stay with the flat fee and voted Model A.

Out of 19,415 eligible members, 3,421 registered to vote and 3,049 cast their votes, resulting in a 15.7% turnout. The May 2026 GM achieved one of the highest turnouts ever, coming only slightly behind the record-holder in May 2020.

The high turnout is undeniably driven by the possibility of a completely different charging scheme. This is confirmed through abstained votes. Out of the 3,049 votes cast, only 23 abstained from the charging scheme vote, compared to 1,032 that abstained from the Executive Board election.

By excluding the abstained votes, we can observe which of the two very important decisions were more successful at getting members to turn out. The below graph shows the number of votes cast on the charging scheme and Executive Board elections over the May GMs.

The May 2020, May 2023 and May 2026 GMs were certainly exceptional, with consequential decisions being voted on. Votes cast for the Executive Board surpassed those for the Charging Scheme in May 2020, the highest-participation GM to date. But it was the charging scheme vote that pushed the total participation high in May 2023 and May 2026.

Participation by country

Participation varies across countries, and when overall turnout changes significantly, graphs such as the below also tend to differ more from their predecessors. Turnout increased in almost all countries, but not equally.

Some things don’t really change: German members contributed by far the most to the vote registrations, followed by the other countries with big memberships, as usual. Although among the big countries, Spain had a higher contribution than usual.

Although not visible on this plot, it’s notable that Serbia’s turnout had a bigger jump than most countries, most likely due to the first Serbian candidate standing in the Board elections.

Ukraine is the most impressive, with a 41% turnout. Ukrainian members certainly cared about the charging scheme and created the highest participation from Ukraine since May 2023, when two Ukrainian candidates were running for the Board. The high Ukrainian turnout this time was almost certainly fueled by the discussions on PI resources’ contribution to the overall fees. Ukraine has the highest number of members holding or sponsoring PI relative to its membership, and we saw several contributions to charging discussions on the mailing lists and at the GM from Ukrainian members relating to this topic.

Participation by age of membership

However interesting the country stats may be, we think when a member is from says more than where they are from. People's motivations to become a member have changed as the market has evolved. The time they joined is a great indication of the resources they hold today and has significantly shaped their expectations. This is likely reflected in how they vote, and how likely they are to vote.

The below graph shows the number of eligible members that joined each year, coloured by whether they registered to vote at this GM. Since younger members are more abundant overall, they also make up most of the voters. But in fact, relatively more of the older members actually turned out to vote, which aligns with our observations from the previous GMs.

Age is an indicator of “resourcefulness” (as in, the amount of Internet number resources a member holds), and resources are a factor in how much members would pay under the proposed Model B. Knowing this, we have to break this graph down further, based on the charging scheme model that would result in a lower fee for each member.

In the above graph, the darker coloured bars show the members that would pay less under Model A - more resourceful members. They are clearly accumulated on the left-hand side, reinforcing the fact that older members have more resources.

Exploration. Interpretation. Speculation.

The country stats and ages of members are not surprising, but the peak participation at this GM certainly stands out. The vote on the charging scheme brought members to the ballot, and its result brought them to the members-discuss mailing list afterwards.

A tiered charging scheme has been long requested by many in the membership to make things fairer, especially for the newer members with fewer or no resources. To respond to the demand, the RIPE NCC created Model B as an alternative to the existing flat fee model, using the task force principles as a guide. The total income generated from Model A and Model B would amount to the same total, only shifting it to make members with more resources pay more. This shift offered in Model B meant that 74% of the members would pay less than they would under Model A. Naturally, most of us expected the option that financially benefits a clear majority of the members would pass. Instead, Model B could only gather 48.9% of the votes cast on this resolution and was not chosen.

Many people (including us at the RIPE NCC) are puzzled by the result and have been trying to make sense of it. In this section of the article, we will explore the reasons that could have affected the vote. Below are some factors that are talked over, along with our interpretation.

Not enough participation, and an unrepresentative sample

This GM is the second-highest voted on with 15.7% turnout. Yet some people point to the low participation to explain the curious result. From another perspective, 84.3% of the members didn’t have a say in a decision that would determine how much their bill will be. And what about the 372 members that registered but didn’t cast their votes?

At every GM, some of the registered voters end up not casting their votes. Registration for this GM started almost three months before the meeting took place - earlier than usual due to the importance of the charging scheme vote. People might have registered and later forgot, or decided not to go through with it. For whatever reason, we consistently see that around 10-20% of the registered votes don’t end up being cast.

Below is the same graph from the introduction, focusing on how many of the registered votes were cast. At this GM, 3,049 of the 3,421 registered votes were cast, resulting in the highest cast-to-registered ratio in the last decade at least.

Uncast votes aside, it is fair to wonder why more members wouldn’t register to vote on a financial decision that could significantly reduce their fee. It is certainly not surprising that the majority didn’t vote, as turnout fluctuates around 5-15%. It is difficult to find out why though. If someone would not vote on a decision that determines their fees, how likely is it that they would engage with us to tell us their reasons for not voting? A big part of our membership only seems to be concerned with the resources and services they receive, and don’t want to be involved more than they need to.

Questioning participation levels is ultimately about figuring out how representative the voters are of the whole membership. We can easily answer this when it comes to the charging scheme. The below graph shows the number of members that would fall into each category under Model B. Those who would be in a category below 6, 74% of the membership, would financially benefit from this model. Among the voters, 59% would be in a category below 6.

This stat shows that voters tend to be more resourceful, also related to their age, but still would have mostly benefitted from this model financially. The sample of voters are not highly representative of the overall membership, but representative enough to choose Model B if they were to vote according to their financial interest.

Muddied waters

It seems like a simple enough choice that should have resulted in Model B being approved. So why wasn’t it? Could it be that some people didn't get the memo?

From our side, we did our best to ensure the message got across. We communicated many times with members on various channels, we developed and announced a brand-new charging scheme calculator, we reminded people to register to vote between charging scheme discussions on the members-discuss mailing list, and plenty more besides.

On the last stretch, to reach as many members as we could and to make the alternatives as clear as possible, we sent personalised emails to all members with the info on how much they would pay under both charging scheme models. The emails were sent to Admin, Regular and Billing contacts of all eligible members a week ahead of the GM. This was successful at drawing people to the GM, as we observed a big jump in registrations right after the mails were sent.

Along with more registrations, we also saw increased spam ratings with the mails we sent out. Although unrelated to the voting behaviour we’re trying to analyse in this article, it does illustrate the challenge of balancing conflicting feedback from members. Some say we don’t share enough information in the emails, while others tell us we send too many emails. We take all feedback seriously and work to improve how we communicate with members, especially since emails from the RIPE NCC being marked as spam prevents people from receiving the information they need to make an informed decision.

While we do our part to communicate on the important topics with the members, there was also increased communication among members themselves to try and influence the vote. This GM saw campaigning in the form of printed materials at the RIPE Meeting and a website shared around, pushing for Model A. How impactful these efforts were is difficult to say, but it is a noteworthy change from what we’ve seen in the past.

During the lengthy period of preparing the charging scheme options, we published consultations with different ideas for the base fee and various additional fees. Some earlier models and additional fees that were put forward were later abandoned following member input.

Throughout these consultations, the alternatives were heavily discussed on the members-discuss list, where we encourage you to share your feedback. Among these discussions, the potential impact of PI resources on fees provoked some strong reactions, which we will explore further in the next section. As we moved to the final stage of choosing between models A and B, the criticism over earlier considerations and the RIPE NCC’s budget continued. Even after announcing that most of the early ideas that were floated would not be incorporated into the final proposed models, including increased fees for independent resource assignments, we saw that people continued to argue for no changes to these fees.

Similarly, the topic of voting rights was one that we needed to jump in on to correct misconceptions. We saw several times that people assumed a category model would mean increased voting rights for those who paid more, or that multiple LIR accounts come with multiple voting rights. To be clear, this was never up for debate. In fact, no matter the model chosen or the fees paid, a membership organisation in the Netherlands can legally only allow one vote per member. While the current charging model is “one LIR account-one fee”, the voting model will always be “one member-one vote”.

All of these and similar discussions might have muddied the image of the new model more than the existing one. Some members might have wrongly associated the additional fees only with the new Model B. Some might have even assumed the category model was Model A, since it was the option recommended by the Executive Board. Normally, such factors would not make a difference to the outcome. But given the result was so tight, a swing of unintentional votes could have meant a different outcome.

Protecting future interests

When a vote results in a decision that is seemingly against the interests of the majority, it is easy to attribute that to misinformation. But members could have preferred to stay with the existing model in order to protect their future interests.

Stats like “74% would pay less under Model B” only show the dynamics based on the resources held by each member today under the exact categories proposed today. This could change over time.

Perhaps some members foresee expansion of their networks and the need to acquire resources from the market, and therefore jumping to a higher category in the future. They could also expect further changes to the exact fees and cut-offs that could shift the balance. A tiered model opens the door for potentially lower fees for some in future iterations, but that could come with increased costs for others.

These concerns about a model based on resources are possibly even stronger for members holding/sponsoring PI and Legacy resources. If we start considering PI and Legacy the same as PA, and members are eventually charged for all IPv4 resources in the same way, then the category model could see higher charges for some members who hold/sponsor PI and Legacy resources. Although, in other cases, they would end up with even lower fees with no additional charge for PI and Legacy, provided these resources weren’t enough to move them to a higher category.

Some members with PI assignments in particular were vocal during early discussions about increased fees for PI, despite the final proposed models not taking PI into account. They stressed it would be a disproportionate burden on some communities that rely heavily on PI resources. Ukraine leads in this area with 59% of their members holding or sponsoring PI resources, followed by Romania (47%) and Poland (46%).

34.7% of the registered votes were from members with PI, 37.2% when we include members with legacy resources in the group. Given the strength of opinion we heard regarding PI, this may be another case of some members preferring the perceived certainty of the existing model.

Equality over equity

We’ve been receiving a lot of feedback on tiered model proposals over the last few years. Along with some good pointers, they also show there is no model everyone would agree on. We don’t know whether a different category model would be able to gather the required amount of votes to be selected. We can only contemplate how this one could have disappointed those in favour of a tiered model.

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. A membership association means all members contributing equally, having equal say in matters and having equal access to the services provided. Some members don’t want to give up on that ideal.

Who casts the votes for our voters?

We wonder why members don’t turn out, or why they wouldn’t vote in the financial interest of their organisation, but we don't consider the overall interests of the organisation nor that the member’s representative authorised to cast the vote might have different priorities.

It is generally in the interest of companies to reduce their costs, but is that true for all organisations? Is that true for the authorised representative that votes, who might well be more interested in the technical health of their company rather than its financial health? While we do our best to provide all relevant information to people who cast their votes, the people with financial concerns might not be the ones staying on top of RIPE NCC governance matters.

We tend to assume that we understand companies’ interests and we see votes as a perfect representation of such interests. In reality, we don't know the overall motivations or interests of the organisations or of the individuals who are authorised to vote for our members.

Final thoughts

Had either Model A or Model B won in a landslide vote, this article would be much shorter. However, when a swing of 35 votes would have created a different outcome for all members, it’s worth considering the factors that might have tipped the balance and resulted in this outcome.

The charging scheme has been discussed for years, and since the GM, it feels like everyone is talking about the result and trying to make sense of it. But we forget that what feels like everyone is the members who are among the most active in the community. The membership is not only far bigger but also much more diverse than what we see at RIPE Meetings and on the mailing lists.

Just as it is difficult to get a perfect representation of the membership on the ballot, it is also difficult to get a perfect representation of the membership through events and mailing lists. And it is wrong to assume we know what the membership wants based on our limited interaction. This is a group of 20,000 members, in various industries, spread across vastly different countries and cultures, having different needs, expectations and opinions, often even differing within their own organisations. Why should we expect the membership as a whole to have a single idea of how they should be charged?

All this leads us to ponder other questions: Would the result change if we asked again? Would the membership change their mind in the upcoming years? Are we going to keep receiving requests for a category model? And will any of this matter once everyone has deployed IPv6? ;)