Nextcloud Hosting Comparison 2026: Providers, Prices and the Number Nobody Quotes

Every Nextcloud hosting comparison tends to lose the same line item: the number that decides whether the installation is actually usable day to day. Not the storage price. Not the user count. It is compute — and it never shows up in a plan comparison, because it does not reveal itself when you try the service out.

We measured it on 19 August 2026. A real Nextcloud instance (34.0.3, MariaDB 11.4, Redis), 300 files in one folder, then the same request served with one, two and four processor cores. The headline first, because it reframes the whole comparison:

With eight concurrent users, the same folder request took 4,095 ms on one core and 565 ms on four — a factor of 7.2. And with a single user, the one-core server was just as fast as the four-core server: 276 ms versus 266 ms, a 3.8 % difference.

That is the trap in every Nextcloud hosting comparison: if you test a plan on your own, you are not measuring the server size at all. The cheapest plan feels exactly as fast as the most expensive one during the trial. The difference only arrives once colleagues join — which is after the purchase decision.

This article compares four operating models with live prices retrieved the same day and our own measurements: your own VPS, Hetzner Storage Share, managed providers and Nextcloud Enterprise. Where we had to correct ourselves while measuring, it is stated in the text.

Nextcloud hosting comparison: the four models at a glance

Nextcloud is open source. That means the software costs nothing, but somebody has to run it. The four routes differ almost exclusively in who that somebody is — and what happens when something breaks.

The spread between these rows is wider than it looks. For five users, Storage Share costs €51.48 a year, a managed subscription with the same five users €780 a year — fifteen times as much. Both deliver a working Nextcloud.

Who actually offers Nextcloud hosting?

For this Nextcloud hosting comparison we did not consult review sites. We parsed the official Nextcloud partner directory — 156 entries, each carrying machine-readable service and country attributes.

The result is surprising if you are searching as an individual or a small team:

Method: parsed nextcloud.com/partners on 19 Aug 2026, 156 partner blocks, attributes read from the page’s own filter values (host-own, host-rend, host-home, host-enterprise) and counted. Multiple services per partner possible.

Of 156 official partners, only 29 offer hosting for home users at all — eight of them covering Germany. 43 partners do not host anything; they sell consulting, integration or training. The official partner list is largely an enterprise directory, not a provider comparison for small installations.

That explains something many people experience while searching: you click through the partner page, find dozens of impressive companies, and not a single offer that fits five people. The list is not wrong. It simply answers a different question.

A side finding from the same dataset: only 7 of 156 partners cover more than 50 countries, while 51 partners serve at most three countries. Nextcloud hosting is mostly a regional business — which is good news for data residency and awkward for shortlisting, because comparison sites usually only know the big names.

The measurement: why a solo trial makes every plan look good

Now to the core question of any Nextcloud hosting comparison — compute. We installed Nextcloud 34.0.3, put 300 files into one folder and queried that folder over WebDAV (PROPFIND, exactly what every desktop client does when syncing). Then we gave the container different numbers of cores via cgroup limits.

Method: Nextcloud 34.0.3 (Docker, MariaDB 11.4, Redis, APCu), folder containing 300 files, PROPFIND with Depth: 1, 24 requests over 8 parallel connections per run. CPU limited via docker update --cpus on the application and database containers together. Measured 19 Aug 2026.

And now the same instance with a single user:

A 3.8 % difference. That is measurement noise.

This is the most practically important finding in this Nextcloud hosting comparison, and it is uncomfortable: the usual way of testing a plan is structurally incapable of seeing the difference between the cheapest and the most expensive plan. You book the small option, upload a few files, click around, everything feels smooth — and the decision is made. The bottleneck only appears once several clients sync at the same time, which is to say in production.

So if you use a trial period, do not test alone. Two colleagues with a configured desktop client syncing a larger folder simultaneously tell you more about the plan than a week of solo use.

The bottleneck is PHP, not the database

While measuring we also recorded which service carries the load. The result contradicts the common assumption that Nextcloud is “database-heavy”:

Method: docker stats during 48 requests over 16 parallel connections, no CPU limit (host: 12 cores). Values sampled at full load.

The database is essentially idle for this workload. Almost all of it lands on PHP — and PHP scales through processes, meaning through cores. For a Nextcloud hosting comparison that means something concrete: a plan with lots of RAM and few cores is the wrong shape for Nextcloud. The whole stack needed 265 MiB idle and 422 MiB under load. A 4 GB plan is therefore comfortable on memory while still collapsing on a single core.

A second finding from the same series, with no CPU limit: up to 8 concurrent requests the median stayed flat at roughly 260 ms; only at 16 did it rise to 393 ms and p95 from 325 to 724 ms. The degradation does not creep in, it arrives as a knee — typical of a system that queues processes.

Model 1: your own VPS — full control, full responsibility

The classic route. You rent a virtual server and install Nextcloud yourself. All prices retrieved from Hetzner’s official pricing API on 19 Aug 2026 (net, Nuremberg location):

Source: hetzner.com/_resources/app/data/bench/cloud_data.json, retrieved 19 Aug 2026, location NBG1, net euro prices. Add €0.50/month for the IPv4 address — it is not included in the server price.

⚠️ A trap we have now hit twice: the same file contains a field with the value 0.50 sitting right next to the server data. That is not a plan, it is the IPv4 fee. Anyone parsing the file automatically without checking what a value means ends up with a “€0.50 server” in their table. A number from the right source can still be the wrong number if you do not verify its meaning.

For larger data volumes, a Storage Box usually joins the bill, either as a backup target or as mounted third-party storage:

The provider comparison that usually goes wrong

A frequent error in Nextcloud hosting comparisons is comparing providers who advertise different things. Take IONOS, as of 19 Aug 2026:

The advertised VPS M+ (4 vCores, 4 GB RAM, 120 GB NVMe) is listed at €3 per month. The small print on the same page says: “for 3 months”, then €9, plus a €10 setup fee — and all of those figures include VAT. Hetzner quotes net prices.

Computed properly, first year, same basis (net):

Method: IONOS gross prices divided by 1.19 (product page figures retrieved 19 Aug 2026: promo €3, then €9, setup €10, all gross). Hetzner net prices from the pricing API plus €0.50 IPv4.

IONOS really is considerably cheaper in year one, and it offers more cores while doing so. That is a genuine advantage given our measurement. The point is not that a provider is cheating — it is that a comparison placing €9 gross next to €11.49 net is systematically wrong, and wrong in either direction depending on which price happens to be the promotional one.

What self-hosting really costs: your time

The server price is the smaller part. So we measured what a clean backup costs in downtime — the task that recurs with self-hosting and does not exist with managed offerings:

Method: test instance with 376 file entries, a 6.4 MB database (131 tables) and 65 MB of user data. Measured 19 Aug 2026.

We are publishing this even though it argues against our own thesis. Backups are technically cheap on small installations — 1.3 seconds of downtime, one script, done. Anyone claiming in a Nextcloud hosting comparison that self-hosting fails on backups is arguing past the reality of small instances.

The real effort lies elsewhere: major version upgrades. Nextcloud maintains each major version for roughly a year and does not allow skipping versions — miss one and you have to walk the chain. On top of that, any installed app can block an upgrade if its developer has not adopted the new version yet. We counted how big that problem is in our article on Nextcloud for schools: 23.5 % of apps in the official store are effectively dead.

Two hours of maintenance a month is realistic. At €60 an hour that is €1,440 a year — more than any server plan in this comparison. Do the work yourself and you do not pay it in money, but it still has to be paid.

Model 2: Hetzner Storage Share — the underrated middle ground

Storage Share is a Nextcloud instance operated by Hetzner. You get Nextcloud, but not the server underneath. Prices from the pricing API on 19 Aug 2026:

Source: product keys KONSOLEH_136/137/138 from hetzner.com/_resources/app/data/app/live_data_prices.json, matched against the rendered product page, retrieved 19 Aug 2026. Net prices. Falkenstein location.

The user count is unlimited — the concurrent connections are not. That column sits on the product page and appears in practically no Nextcloud hosting comparison, even though it is the actual capacity ceiling.

This is where the price list meets our measurement: a desktop client holds several connections open while syncing, and so does every mobile app running in the background. “Unlimited users” therefore does not mean “unlimited working users”. Our measurement shows that just 16 parallel requests push p95 from 325 to 724 ms — on an instance that had 12 cores available.

For most small and mid-sized cases Storage Share is nonetheless by far the most attractive model on price:

Method: own VPS = CPX22/CPX32/CPX42 + €0.50 IPv4 + a Storage Box at twice the data volume (versions, trash, backup). Storage Share = smallest plan covering the assumed data volume (1/2/5 TB). Nextcloud One = €13/user/month on annual billing from 5 users. All figures net or as published by the provider, as of 19 Aug 2026. Labour not included.

At 100 users, Storage Share costs €170 a year and a managed subscription €15,600 — a factor of 92. That number needs framing or it becomes misleading: the €15,600 buys 24/7 support, a contractual partner carrying liability and a dedicated instance. The €170 buys storage with a connection ceiling. These are different products, not differently priced versions of the same product.

The honest downside of Storage Share: you have no access to the execution environment. No custom app repository, no configuration file, no way to cap preview generation — precisely the setting that saves 85.4 % of preview storage according to our Nextcloud cost measurement. On Storage Share you pay that overhead as storage.

Model 3: managed providers — you are buying a contact

Nextcloud One is the official managed offering (operated by epiKshare, data centre in Germany). Prices as of 19 Aug 2026:

Included, per the provider: 24/7 support, 30 days of backup retention, virus scanning of every uploaded file, multi-factor authentication and custom branding.

Purely arithmetically that is expensive. But the comparison is only fair if you say what you get: a contractual partner. If the instance is unreachable at 3 a.m., somebody else gets the call. With self-hosting that somebody is you, and with Storage Share the Nextcloud operation sits with Hetzner but there is no SLA in the sense of a business contract with response times.

For organisations with compliance requirements this, rather than price, is often the deciding factor. If you need a data processing agreement under Article 28 GDPR with clear responsibilities, you cannot sign one with yourself.

Model 4: Nextcloud Enterprise — and what is not in it

The subscription directly from Nextcloud, prices from the official pricing page (19 Aug 2026):

Two constraints stated on that same page that comparisons routinely omit:

1. Below 100 users Enterprise cannot be bought at all. The pricing page explicitly redirects smaller interested parties to Nextcloud One.
2. It is not hosting. Verbatim from the pricing page: “Nextcloud employees never gain access to your data as we do not offer hosting.” You are buying support and maintenance commitments — the server remains yours to provide.

That is the most common misconception in enterprise-level Nextcloud hosting comparisons: Enterprise is not an operating model, it is a support contract layered on top of one of the other three.

The criteria that actually decide

Compute, not storage

Based on our measurement this is by far the most underestimated criterion. As a rule of thumb derived from the numbers:

• 1 core: only for a single person who does not sync in parallel
• 2 cores: a family or micro-team of around five people
• 4 cores: a team of up to roughly 25 people with desktop clients
• 8 cores and up: from roughly 50 active people, or once Office/Talk join

These thresholds are experience-based readings of our measurement series, not a measured capacity per user count — more on that in the honesty section below.

Backups: who takes them, and where to?

On Storage Share, Hetzner states it backs up the database and storage backend several times daily. Nextcloud One commits to 30 days of retention. On your own VPS: nobody, until you set it up.

Crucially — and this is a common mistake: a backup on the same server is not a backup. It survives software faults but neither a host failure nor a compromised account. We learned how quickly that becomes relevant when a server of ours was hijacked for five days without anyone noticing. That is exactly why the Storage Box is part of our calculation: separate system, separate credentials.

Location and GDPR

Every provider compared here operates in Germany. That is the norm in the Nextcloud space — partly because, as our partner analysis shows, most providers work regionally.

Server location alone is not the whole answer though: you need a data processing agreement with the provider. On your own VPS that agreement exists with the data centre operator (for the infrastructure, not for Nextcloud); with managed offerings it covers the service itself. Run Nextcloud yourself and you are legally responsible for the application — including security updates. We described what can go wrong there in our overview of IT security vulnerabilities.

Update responsibility

The question that makes the biggest difference in daily life:

Our own practice — and where it argues against us

We run our infrastructure ourselves, on our own servers under Debian. That is a deliberate choice for control. But an honest Nextcloud hosting comparison has to say this:

Price cannot justify self-hosting on small installations. At five users the own VPS with backup costs €370.68 a year, Storage Share €51.48 — seven times as much, before a single working hour is counted. Self-hosting buys control and adaptability, not savings.

The reverse holds too: as soon as you need custom apps, custom configuration, third-party storage or integration with an existing identity system, there is no alternative to your own server — at which point the question is no longer “whether” but only “how large”.

What we did not measure

For the numbers above to stay trustworthy, what is not in them belongs here too. An estimated number looks identical to a measured one inside a table.

• No real plans over time. Our CPU measurement ran on a 12-core host constrained via cgroups. That isolates the effect of core count cleanly, but it does not reproduce steal time — the variance that genuinely occurs on shared virtual servers. On an actual single-core VPS, expect worse numbers, not better.
• No capacity claim per user count. We measured concurrent requests, not “how many humans fit on 4 cores”. How many requests a human generates depends on usage patterns (a desktop client differs from occasional web use by a large multiple).
• No test of Talk or Office. Video calls and collaborative editing have entirely different load profiles. For Talk, the bandwidth ceiling we documented from the vendor’s own documentation in the schools article applies.
• No availability measurement. We collected no uptime statistics from the providers and therefore quote none.
• No quotes requested. Prices for managed providers beyond their published lists are not included.

Conclusion: which Nextcloud hosting for whom?

After all the measurements and price retrievals, this Nextcloud hosting comparison boils down to a few clear recommendations:

Private use or a small team, plain file storage: Storage Share, NX11 at €4.29 a month. Unbeatable on price, unlimited users, no operational work. The 50-connection ceiling is irrelevant at that size.

A team with specific requirements (apps, configuration, third-party integration): your own VPS — but not the smallest one. Our measurement puts four cores at the threshold where eight concurrent users stop seeing second-long latencies. Moving from CPX12 to CPX32 costs €24 more per month; the difference in behaviour is a factor of seven.

An organisation with compliance and support needs: a managed provider. The price is high, but you are not buying storage, you are buying accountability.

From 100 users with in-house IT: an Enterprise subscription on your own infrastructure — knowing that the server remains your job.

And the one thing to do before any purchase: do not test alone. Our measurement shows a single user cannot perceive the difference between an €11 and a €35 plan. Two colleagues with a syncing desktop client will show it in five minutes.

Frequently asked questions about Nextcloud hosting

What is the cheapest way to host Nextcloud?

Hetzner Storage Share from €4.29 net per month for 1 TB with unlimited users is the cheapest serious offer in our comparison. Your own VPS starts at €11.49 plus €0.50 for the IPv4 address, so more than double — and your own time for updates and backups comes on top.

How many CPU cores does Nextcloud need?

Based on our measurement of 19 Aug 2026: a single user is fine on one core (276 ms response time). With eight concurrent users, the same query took 4,095 ms on one core and stayed at 565 ms on four. For a team using desktop clients, four cores is the practical minimum. Memory is rarely the constraint: the entire stack needed only 422 MiB under load.

Is Storage Share better than running your own server?

For plain file storage almost always, because it is cheaper and creates no operational work. The drawback: you have no access to the configuration, so you can neither install custom apps nor cap preview generation — a setting that saves up to 85 % of storage. If you need custom adaptations, there is no way around your own server.

What does “unlimited users” mean on Storage Share?

It means you may create any number of accounts. It does not mean any number of people working at once: the plans cap simultaneous connections at 50 (NX11), 100 (NX21) and 200 (NX31) respectively. Since every desktop client uses several connections at once while syncing, that is the real capacity ceiling.

How many Nextcloud hosting providers are there?

Of the 156 official Nextcloud partners (as of 19 Aug 2026), 86 cover Germany. For private and small users the field shrinks sharply: only 29 partners worldwide offer hosting for home users, eight of them covering Germany. 43 of the 156 partners offer no hosting at all, only consulting and integration.

Do I need Nextcloud Enterprise?

Only from 100 users — below that the subscription cannot be purchased at all according to the pricing page. And it is important to know that Enterprise is not hosting: Nextcloud states plainly that it does not offer hosting services. You are buying support and maintenance commitments; running the server stays your responsibility.

Does self-hosting pay off financially?

Not on small installations. At five users, your own VPS with backup storage costs €370.68 a year against €51.48 for Storage Share — seven times as much, before labour. Add two hours of maintenance a month at €60 and another €1,440 a year joins the bill. Self-hosting pays off for control and adaptability, not for price.

How should I test a hosting plan properly?

Not alone. Our measurement shows a single user sees practically identical response times on one core (276 ms) and four cores (266 ms) — a 3.8 % difference. The difference between plans only becomes visible under concurrent use. During a trial, have two or three people sync a larger folder with the desktop client at the same time.

Who is responsible for updates?

On your own VPS, you are — including the rule that Nextcloud major versions cannot be skipped and each version is maintained for roughly a year. On Storage Share, Hetzner performs the updates; on Nextcloud One, the provider does. With an Enterprise subscription you remain responsible but receive support.

Is a backup on the same server enough?

No. A backup on the same machine survives software faults but neither a hardware failure nor a compromised account. That is why our cost calculation includes a separate Storage Box with its own credentials. Technically the backup itself is cheap: in our test, total downtime for maintenance mode, database dump and file archive was 1.3 seconds.