Measured, not advertised

Hosting and uptime across the servers we track

We ping every server on this list around the clock, resolve each one to the company actually running it, and compare them on the thing that matters to players: whether the server answers when they try to join. This is a report on our own sample — read the note below before treating any figure as a statement about Minecraft servers generally.

Based on 2,953,845 status checks across 153 days · Updated 5 Nov 2026
1,067
Self-hosted servers
The segment reported on below
37
Minecraft hosts identified
Among self-hosted servers
8%
Behind DDoS protection
Self-hosted only — see note
153 days
Continuous measurement
What this is a sample of

These figures describe the 2,358 servers we track, not Minecraft servers as a whole. We find servers by crawling public listings and platform APIs, so the sample reflects that method rather than the game's true population.

1,291 of them (55%) run on free hosting platforms such as Minehut and Aternos, because those platforms publish machine-readable server lists we can import in bulk. Those servers are typically small, short-lived, and all sit behind the same shared infrastructure — so mixing them with the 1,067 servers on their own domains would distort every figure below.

Everything on this page is therefore reported separately for the two groups. Where a single number is shown, it refers to self-hosted servers.

Uptime by hosting provider

Every listed server is checked continuously. This is the share of those checks where the server answered, averaged across all servers on each provider over the last 30 days. Servers that never answered once, or that went dark and never came back, are excluded — an owner who stops paying is not an outage, and counting it as one penalises whichever provider hosts the most servers. Providers need at least 3 measured servers to appear.

Provider Servers Average uptime Uptime Above 95% uptime Set aside
AWS datacenter 3 100.0% 3 of 3 1
netcup datacenter 3 100.0% 3 of 3 2
InterServer datacenter 7 99.9% 7 of 7 5
IPXO datacenter 3 99.9% 3 of 3 2
GameServerKings MC host 3 99.9% 3 of 3 1
Apex Hosting MC host 5 99.8% 5 of 5 3
DigitalOcean datacenter 3 99.7% 3 of 3 1
FY Technology datacenter 3 99.7% 3 of 3 2
Pufferfish Host MC host 14 99.6% 14 of 14 7
Contabo datacenter 4 99.5% 4 of 4 2
Bloom.host MC host 39 99.4% 37 of 39 13
Sparked Host MC host 8 99.4% 8 of 8 5
BisectHosting MC host 9 99.3% 8 of 9 8
ReliableSite datacenter 7 99.1% 6 of 7 —
HyperPacket datacenter 3 99.0% 3 of 3 —
Oracle Cloud datacenter 11 98.7% 10 of 11 6
CloudExa Hosting MC host 3 98.5% 3 of 3 2
OVH datacenter 153 98.3% 148 of 153 73
WebNX datacenter 3 97.7% 2 of 3 1
BB Host MC host 3 97.5% 2 of 3 3
Servers.com datacenter 10 96.4% 9 of 10 6
SlashN Services datacenter 7 96.3% 5 of 7 2
Hostinger datacenter 8 96.2% 7 of 8 7
PebbleHost MC host 18 94.8% 15 of 18 14
Azure datacenter 3 93.6% 2 of 3 2
AME Hosting datacenter 13 92.8% 12 of 13 9
Hetzner datacenter 43 92.6% 37 of 43 29
How abandoned servers are handled, and why it matters. We ping the Minecraft server, not the provider's hardware, so a server that has been switched off or abandoned is indistinguishable from one whose host is down. That is not a small effect: 91% of all the downtime we recorded over these 30 days came from 84 servers that never answered a single check and 784 that went dark and never returned — out of 1,440 measured. Only the remaining tenth was servers that drop out and recover, which is the instability actually worth reporting. Those listings are therefore set aside rather than charged to the host, and the count is shown per provider in the last column so you can see what was removed. Without that correction the table ranked providers by how many stale listings they had collected, which quietly punished the largest ones. Small samples still move a lot, so a provider with three servers is an observation, not a verdict.

What this table leaves out. Of the 1,067 self-hosted servers we have checked, 243 (23%) could not be attributed to any identifiable operator, and a further 89 sit behind a protection layer that hides their origin. Those servers are absent from every row above rather than counted against anyone. We identify a host from the address players connect to, from reverse DNS, and from the organisation registered to the network block, in that order — so a host that brands none of those is invisible to us, and its absence here says nothing about its size or quality.

Not every address belongs to a hosting company

This is worth understanding before reading any ranking. A Minecraft server's address can resolve to a specialist host, to the raw datacenter that host rents from, to a protection layer, or to address space the network owns outright. They are not comparable, so they are never mixed below.

What the address belongs toServersShare
A datacenter / VPS provider 479 45%
The network's own address space 213 20%
A Minecraft hosting company 191 18%
A DDoS protection layer 89 8%
A home internet connection 65 6%
Not identifiable 30 3%
Why "the network's own address space" is its own row. Some large networks hold their own IP allocation, so ownership records name the network rather than any host — and a handful of records name an individual person. Neither is a hosting company, so those servers are counted here and excluded from every provider table. An earlier version of this page listed them as providers; that was wrong.

Minecraft hosting companies

Companies that sell Minecraft server hosting specifically. Most rent their hardware from the datacenters in the next table, so they are identified by how their servers identify themselves in reverse DNS rather than by who owns the address.

HostServers ServersOnlinePlayers
Bloom.host 51 46 1,108
PebbleHost 27 18 278
Pufferfish Host 19 18 3,465
Sparked Host 12 11 203
BisectHosting 12 11 28
Apex Hosting 11 7 1
BB Host 5 2 22
CloudExa Hosting 4 4 1,023
Meteor Cloud 4 4 165
DatHost 4 4 0
GameServerKings 3 2 150
Freakhosting 3 2 103
Folium 3 2 25
MCProHosting 3 1 11

Datacenters and VPS providers

Raw infrastructure. Servers here are either self-managed by their owner, or run by a Minecraft host whose reverse DNS does not identify it — so treat this as "where the metal is", not "who the customer bought from".

ProviderServers ServersOnlinePlayers
OVH 217 204 16,282
Hetzner 65 56 559
AME Hosting 25 19 399
Oracle Cloud 21 18 24
Servers.com 15 11 41
Hostinger 12 7 21
ReliableSite 12 12 1
InterServer 11 10 165
SlashN Services 11 9 23
Contabo 7 6 5
netcup 6 6 138
HyperPacket 6 6 69

DDoS protection: 8% of self-hosted servers, 68% of platform-hosted

An earlier version of this page reported a single blended figure of 41%. That number was an artefact of how we collect data, not a fact about Minecraft: free hosting platforms route every customer through the same protection layer, and we import those servers in bulk from platform APIs. Separated, the two groups behave completely differently — and only the self-hosted figure reflects a decision an operator actually made.

Protection layerServers
Minehut 872
TCPShield 74
Cloudflare 17
Rabisu Anti-DDoS 1
MineGuard 1
PebbleHost 1
Servers behind a protection layer are excluded from every provider figure on this page — counting them would attribute hundreds of servers to Cloudflare, which hosts none of them. The provider and country tables above cover self-hosted servers only, for the same reason: a free platform's servers all resolve to the platform, which tells you nothing about any operator's hosting choice.

Where the machines are

Physical location of directly-hosted servers. Worth knowing if latency matters to you — a server three continents away feels slow no matter how good its uptime is.

CountryServersPlayers
United States 299 12,702
Canada 101 11,212
Germany 87 1,641
France 45 2,061
United Kingdom 36 1,320
The Netherlands 18 24
Brazil 16 2,271
Singapore 15 151
Finland 15 46
India 11 3
Running a server on one of these?

List it free and we'll track its uptime the same way — you'll get the numbers for your own server, not just the averages.

Add your server Server trends →

How this was measured

Every server on this list is pinged continuously using Minecraft's own server-list protocol — the same request your game client makes. Each check records whether the server answered. Uptime is the share of checks answered over the stated window; a server needs at least 200 checks before it counts towards any average.

Identifying the host takes two steps, because address-ownership records alone give the wrong answer. Minecraft hosts are usually resellers: Bloom.host rents from AME Hosting, PebbleHost from ColoCrossing, so ownership records name the upstream and the actual host disappears. We therefore resolve each server's advertised address — following its _minecraft._tcp SRV record where one exists — and read the reverse DNS of the result, which is where hosts brand themselves (bloom.host, hosted-by.sparkedhost.com, *.your-server.de). Ownership records are used only as a fallback, and only for names we recognise as companies.

Each server is then assigned a category — Minecraft host, datacenter, protection layer, own address space, or home connection — and the tables never mix them. Brand names are normalised because providers operate under many legal entities: OVH alone appears under five in raw ownership records.

Servers behind Cloudflare, TCPShield and similar protection layers are excluded from provider and country figures, since the observable address belongs to the protection network rather than the host. So are servers on address space the network owns itself, and the small number whose ownership record names an individual rather than a company. All are counted in the breakdown above.

Figures refresh every six hours. This page reports what we can observe from outside; it is not based on any information supplied by hosting companies, and no provider pays to appear or rank here.