Data Center Protests Reach Europe: What Buyers Should Do

Data Center Protests Reach Europe: What Buyers Should Do

Short answer (checked 2026-11-07): On Friday 6 November 2026, several hundred activists entered the construction site of a Microsoft data center on Plimsollweg in Amsterdam's Sloterdijk district. Dutch police said they detained 150-200 people on suspicion of trespass. It was part of "Stop Dirty Data Centres", two days of actions in 15 European countries. The fight in Europe is mostly about grid power, and the Dutch figures show why: data centers used 4.6% of the country's electricity at the end of 2024, and grid operators expect about 15% by 2030. For anyone who rents servers, the lesson is the same as in the US: capacity that is already built and powered is worth more than capacity that is planned. We also measured how far the big hubs are from our Prague site.

Sources. Al Jazeera with AFP and AP (2026-11-06), AP via ABC News (Mike Corder and Suman Naishadham, 2026-11-06), NL Times (2026-11-06, quoting AT5 and NOS), NL Times of 2026-01-28 and 2026-07-14 for the Dutch figures, and a Hacker News thread. Crowd counts differ between outlets; we give the range and who said what.

What happened in Amsterdam and around Europe

The action was organised by Geef Tegengas, a Dutch direct-action group. Extinction Rebellion and pro-Palestinian activists joined; Fossielvrij NL held a separate protest next to the site. A Geef Tegengas spokesperson told AT5 that about 500 people went in; Al Jazeera, AFP and AP reported "several hundred"; NL Times wrote that dozens climbed the fence while hundreds stayed outside. Some protesters climbed a crane and lit flares, and police used pepper spray. "We don't want it to be built. Nobody needs it," said Nathan Timmer of Geef Tegengas. Microsoft said it respects the right to protest peacefully but does "not condone any actions that are unlawful or that put people's safety at risk".

Amsterdam was the loudest point of a wider campaign. AP counts actions in 15 countries, including Ireland, Poland, Denmark, Finland, Spain, Italy, France and the UK. About 100 people protested in front of the Polish parliament and a dozen at a site near Berlin. In Milan, more than 1,000 people from about 40 local committees had marched on 10 October. The common demand is a moratorium on large new developments. The money at stake is large: PwC's Global Data Centre Outlook 2026-50, quoted by AP, projects $5.6 trillion of data center capital spending in Europe over that period, about 18% of the world total.

This is the European side of what we saw in the US, where voters backed restrictions in 45 of 46 counted ballot measures and 9 in 10 council comments opposed new sites. The arguments differ in one way: in Europe, grid capacity comes up first.

The Dutch numbers behind the anger

Fact	Figure	Source
Microsoft's Sloterdijk project	78 MW grid connection, three towers of 85 m, 23,000 square metres	NL Times, 2026-01-28
National hyperscale ban (2022)	Applies only to sites with BOTH >= 70 MW AND >= 10 hectares	NL Times, 2026-01-28
Microsoft Middenmeer, 2025	1.17 TWh, about 1% of the 116 TWh the Netherlands used	NRC via NL Times, 2026-07-14 (RVO data under the EU EED)
Data centers' share of Dutch power	4.6% at end of 2024 (CBS); about 15% expected by 2030	NL Times, 2026-07-14
Grid connection waiting list	More than 14,000 businesses and about 7,300 households	NL Times, 2026-07-14

Two things explain why the protest happened at a site that is legal. First, the permit is old. "This permit was granted five years ago. So there's nothing I can do about it anymore," caretaker minister Mona Keijzer told parliament in January. Second, the ban's two thresholds must both be met. A 78 MW site on 2.3 hectares passes under it, which is why ChristenUnie MP Pieter Grinwis asked to change "and" into "or". The comparison that sticks is his: "there are currently 10,000 homes that aren't being built", in a country where thousands of businesses wait for a grid connection.

For a hosting buyer, the table says one thing: in Western Europe's hubs, the scarce input is not land or buildings but megawatts. A site that already has its power contracted and energised is in a different position from one that is still waiting. Our guide to Tier III data centres in Europe explains what to ask about power and redundancy.

Measured: how far are the hubs from Prague? eu_rtt.js

If hub capacity gets harder to buy, how much do you lose by running outside Amsterdam or Frankfurt? We measured it. The script opens TCP connections to public ping endpoints in seven European cities and times the handshake, which is one network round trip. It needs no root and no ICMP. Save it as eu_rtt.js:

// eu_rtt.js - TCP connect time (handshake RTT) from this server to public
// ping endpoints in European hub cities. Usage: bun eu_rtt.js [samples]
import net from "node:net";

const TARGETS = [
  ["Amsterdam", "ams-nl-ping.vultr.com"],
  ["Frankfurt", "fra-de-ping.vultr.com"],
  ["London", "lon-gb-ping.vultr.com"],
  ["Paris", "par-fr-ping.vultr.com"],
  ["Warsaw", "waw-pl-ping.vultr.com"],
  ["Madrid", "mad-es-ping.vultr.com"],
  ["Stockholm", "sto-se-ping.vultr.com"],
];
const N = Number(process.argv[2] || 15);

function connectMs(host) {
  return new Promise((resolve) => {
    const t0 = performance.now();
    const s = net.connect({ host, port: 80, timeout: 3000 });
    s.once("connect", () => { resolve(performance.now() - t0); s.destroy(); });
    s.once("timeout", () => { resolve(null); s.destroy(); });
    s.once("error", () => resolve(null));
  });
}

const pct = (a, p) => a[Math.min(a.length - 1, Math.floor(p * a.length))];
console.log("city        host                      ok  min    median p90");
for (const [city, host] of TARGETS) {
  await connectMs(host); // warm DNS, not counted
  const xs = [];
  for (let i = 0; i < N; i++) {
    const ms = await connectMs(host);
    if (ms != null) xs.push(ms);
    await Bun.sleep(200);
  }
  xs.sort((a, b) => a - b);
  const f = (v) => (v == null ? "-" : v.toFixed(1)).padEnd(7);
  console.log(
    city.padEnd(12) + host.padEnd(26) + String(xs.length).padEnd(4) +
    f(xs[0]) + f(pct(xs, 0.5)) + f(pct(xs, 0.9))
  );
}
console.log(`samples=${N} bun=${Bun.version} date=${new Date().toISOString()}`);

Output, run on a DCXV Cloud server in our Prague location with Bun 1.4.4 on 2026-11-07 (30 samples per city, times in ms):

city        host                      ok  min    median p90
Amsterdam   ams-nl-ping.vultr.com     30  15.3   23.5   24.3
Frankfurt   fra-de-ping.vultr.com     30  9.2    17.2   20.0
London      lon-gb-ping.vultr.com     30  20.2   28.2   29.1
Paris       par-fr-ping.vultr.com     30  17.6   26.2   27.0
Warsaw      waw-pl-ping.vultr.com     30  25.0   26.1   27.6
Madrid      mad-es-ping.vultr.com     30  35.6   43.6   44.5
Stockholm   sto-se-ping.vultr.com     30  39.1   46.9   48.4
samples=30 bun=1.4.4 date=2026-11-07T07:04:22.516Z

How to read it:

• Use the minimum for the path. Six of the seven medians sit about 8 ms above their minimum; Warsaw's does not. A first run with 15 samples a few minutes earlier gave the same picture (Frankfurt 9.8 / 17.7 ms). We did not find where those 8 ms come from, so compare cities with each other and do not read the median as the network distance.
• Frankfurt and Amsterdam are close. 9.2 ms and 15.3 ms at the minimum. For a web shop, an API or a remote desktop, that is a small share of what a user waits for anyway. Our guide to choosing a cloud server location in Europe shows how to weigh this against where the users actually are.
• Routes follow peering, not maps. Warsaw is nearer to Prague than Amsterdam is, yet it came out at 25.0 ms against 15.3 ms. Measure from where you will run; do not guess from distance. We did the same for memory in our measured memory latency test.
• What we did not measure. These are one provider's endpoints, from one site, on one morning. We did not measure from our Covilha site.

What it means if you rent servers

1. Ask about megawatts, not square metres. The Dutch waiting list shows what a missing grid connection means. If a provider sells you capacity in a new building, ask whether its power is contracted and energised today, not "planned for next year".
2. The hub is not the only answer. A site with existing power that is 10-15 ms from Frankfurt or Amsterdam serves most European workloads without a visible difference. Test with the script above before you decide.
3. Stay portable. Protests, moratoria and grid queues can delay a new region by years. Keep images, backups and DNS in a form you can restore elsewhere, as in our cloud migration checklist.
4. Expect more disclosure. The Middenmeer figure exists because the EU Energy Efficiency Directive makes large sites report their consumption. Similar per-site numbers will appear across the EU, and they will feed the next round of local debates.

Existing capacity in Europe

Our cloud servers and dedicated servers run in Tier III data centres in Prague and Covilha that are already built, powered and in service. Write to sales@dcxv.com with the size of the workload and where its users are, and we will tell you which site fits and what it costs.