The Best VPS for n8n of 2026
If you’re arriving from shared hosting: it won’t run there, and Node.js support doesn’t change that — n8n needs a resident daemon and a bound port, and shared accounts get neither. The full reason and the three working paths are here, including why a Raspberry Pi behind your router can’t receive webhooks.
Most people who self-host n8n pick a server based on the wrong number. They look at how many workflows they plan to run — we measured what actually consumes the memory, and workflow count barely registers — buy 4GB, and it works fine — until a few months in, when the box is out of disk and nobody can explain why. The answer is almost always the same: n8n saves the full data of every execution by default, forever, and nothing prunes it.
Three decisions determine what you actually need, and none of them is “how many workflows”:
- Regular mode or queue mode. Regular mode runs everything in one process. Queue mode adds Redis and separate worker processes, and it’s what you need for real concurrency — it also roughly doubles your memory floor.
- SQLite or PostgreSQL. n8n defaults to SQLite. It is fine for testing and a genuine liability in production, and migrating after the fact is the kind of job you only do once.
- How long you keep execution data. The single most common cause of a full disk on a self-hosted n8n box.
Get those right and a $6 server runs a serious automation stack. Get them wrong and no amount of RAM saves you.
Quick comparison
| Provider | From | One-click n8n | Money-back | Best for |
|---|---|---|---|---|
| Hostinger | $5.84/mo | ✅ Yes | 30 days | Fastest working setup |
| Bluehost | $2.09/mo | ✅ Yes | 30 days | Cheapest one-click route |
| Liquid Web | $3.50/mo | ❌ No | ❌ No | Raw VPS power, managed option |
| Kamatera | $20.00/mo | ❌ No | 30-day trial | High-concurrency scaling |
| IONOS | $4/mo → $11 | ❌ No | 30 days | European, unlimited traffic |
| InterServer | $3.00/mo | ❌ No | ❌ No | Budget, scales to 64GB |
| Hetzner | $3.49/mo | ❌ No | ❌ No | Best price-to-performance |
Entry prices are current promotional rates and several are tied to longer terms.
Not sure you should self-host at all? Self-hosted n8n vs n8n Cloud puts both on the same curve: Cloud genuinely wins below 10,000 executions a month once your own time counts, and loses badly above it, because the next tier up is 13× the price for 4× the executions.
What n8n actually needs
If you’re arriving from n8n Cloud rather than starting fresh, read the migration guide first — credentials don’t come across with the workflow export, and the Cloud plan you’re leaving caps memory far below anything below. Building from scratch instead? The Docker Compose stack starts on Postgres, which matters because switching to it later doesn’t migrate your data.
The published baseline is straightforward:
| Resource | Minimum | Recommended |
|---|---|---|
| CPU | 2 vCPU | 4+ vCPU for high-concurrency or AI workflows |
| RAM | 4GB | 16GB+ for long-running or parallel automations |
| Storage | 50GB SSD | NVMe, and more than you think — see below |
| Database | SQLite (testing only) | PostgreSQL |
| OS | Linux (Ubuntu, Debian, CentOS) | Same, with Docker |
Docker is the recommended deployment path — it handles dependencies, isolates workflows, and is effectively required if you want queue mode. You also need a stable public IP, because webhooks are inbound.
That table is the easy part. The three sections below are what actually decides whether your server is correctly sized.
Decision 1: regular mode or queue mode
This is the fork that most sizing guides skip entirely, and it changes your memory requirement more than anything else.
Regular mode is the default. One n8n process handles the UI, the scheduler and every workflow execution. It is simple, it is what you get out of the box, and it is genuinely fine for a personal automation stack or a handful of scheduled jobs. The limitation is that executions compete for one process — a long-running workflow blocks others, and a burst of simultaneous webhooks queues up behind whatever is already running.
Queue mode splits n8n into a main process plus separate worker processes, coordinated through Redis. Workers pull executions off a queue and run them in parallel, so concurrency stops being a bottleneck and a single slow workflow no longer holds everything else up.
The cost is that you are now running three things instead of one: n8n itself, Redis, and however many workers you configure. Each worker is its own process with its own memory footprint. This is why the recommended ceiling jumps from 4GB to 16GB+ — it is not that workflows got heavier, it is that you are running several copies of the runtime.
When to switch: if executions are backing up, if you have workflows that run for minutes at a time, or if you are handling webhook bursts. If you are running twenty scheduled jobs a day, regular mode on 4GB is correct and queue mode is over-engineering.
Decision 2: SQLite or PostgreSQL
n8n ships with SQLite as the default database. It requires no setup, which is exactly why so many production instances are still running on it.
SQLite is a single file with a single writer. For a test instance or a couple of workflows it is completely fine. Under concurrent load — several executions writing at once, which is the normal state of any busy instance and the guaranteed state in queue mode — you start seeing lock contention and, in the worst cases, corruption. Queue mode with SQLite is not a supported configuration in any meaningful sense.
PostgreSQL is the production answer, and the important thing is to choose it before you have months of execution history. Migrating a live n8n instance from SQLite to Postgres means exporting workflows and credentials, standing up the new database, re-importing and re-verifying — and credentials in particular are encrypted against your instance key, so this is a job with real failure modes rather than a config change.
The practical advice: if this instance will ever matter, start on Postgres. It is one extra container in your compose file on day one and a weekend on day two hundred.
Decision 3: execution data, or why your disk filled up
Here is the failure mode that catches almost everyone, and that no competing guide on this keyword mentions.
By default, n8n saves the full input and output data of every node in every execution. A workflow that processes a modest JSON payload might store a few kilobytes per run. One that pulls a large API response, or handles files, or loops over a few hundred records, can store megabytes — per execution.
Now multiply by a workflow running every five minutes, indefinitely. Nothing prunes this automatically unless you tell it to. The database grows without bound, the disk fills, and n8n stops — usually with an error that points at the database rather than at the actual cause.
Three settings solve it, and you should set them on day one:
- Turn on pruning so old execution data is deleted rather than kept forever.
- Set a maximum age — 7 to 30 days covers virtually every real debugging need.
- Consider saving less per execution. You can configure n8n to skip saving data for successful runs and keep it only for failures, which is usually what you actually want. Successful executions are rarely inspected; failed ones always are.
If you take one thing from this article, take this one. It costs nothing, it takes two minutes, and it is the difference between a server that runs for years and one that dies in March.
The 7 best VPS providers for n8n
1. Hostinger — best for one-click setup
Hostinger is the shortest path from nothing to a working n8n instance. Its VPS template ships n8n preconfigured on Docker, so deployment is a few clicks rather than a compose file, and you still get full root access afterwards for reverse proxies, custom Docker configuration and queue-mode setups.
Entry pricing is $5.84/month, and the plan that matters gives you 8GB of RAM — enough headroom to run n8n alongside Postgres and Redis without thinking about it. There are no artificial workflow or execution limits; performance is bounded only by the resources you bought, which is the entire point of self-hosting.
The Docker-based template is worth calling out specifically because it makes redeployment portable. If you outgrow the box, you are moving a container rather than rebuilding an environment.
Pros: one-click n8n template on Docker; full root access retained; 8GB at the practical tier; no execution limits; 30-day money-back guarantee.
Cons: promotional pricing is tied to a long prepay term and renews higher; no GPU plans if you later want local AI models in your workflows.
Who it’s for: anyone who wants a working instance today, and anyone who would rather spend their evening building workflows than debugging a compose file.
Check Hostinger’s current VPS pricing →
2. Bluehost — best budget-friendly n8n VPS
Bluehost offers the same one-click convenience at a lower entry price, with a one-click n8n install that gets you running in minutes and full VPS-level control afterwards. Unlimited workflows, no per-execution fees, and complete data ownership for automations touching APIs, webhooks, databases and AI models.
The infrastructure is worth noting — Bluehost’s VPS runs on Oracle Cloud, which gives it a more solid foundation than the price suggests.
The catch is the same one as always with Bluehost: the headline $2.09/month is a long-commitment rate, and the entry tier is under-specified for anything serious. Size up from the advertised plan and check what you are actually paying in year two before committing.
Pros: one-click n8n install; cheapest one-click route; unlimited workflows with no per-execution charges; 30-day money-back guarantee; solid underlying infrastructure.
Cons: headline price needs a multi-year term; entry specification is thin for production use; renewal pricing is materially higher.
Who it’s for: cost-sensitive users who still want the one-click path rather than a manual Docker install.
Check Bluehost’s current VPS pricing →
3. Liquid Web — best managed or self-managed VPS
Liquid Web is the pick when you want infrastructure quality rather than convenience. There is no preconfigured n8n template and no one-click deployment — you get a clean VPS with full root access and you build the stack yourself.
What you are buying is the option of a managed tier. For a team where n8n is running business-critical automation, having someone else own OS patching, monitoring and incident response is worth real money, and Liquid Web is one of the few providers on this list that offers it properly rather than as an upsell.
Pricing starts at $3.50/month for self-managed. There is no money-back guarantee, which is a genuine negative when you are evaluating.
Pros: strong infrastructure and support reputation; managed and self-managed options; full root access and complete configuration freedom; scales well.
Cons: no n8n template — manual Docker setup required; no money-back guarantee; managed tiers cost considerably more than the entry price implies.
Who it’s for: teams running n8n as production infrastructure, particularly those who want to hand over operations.
Check Liquid Web’s current pricing →
4. Kamatera — best for scaling complex workflows
Kamatera is the configurable option: you compose the server rather than picking a plan, and you can adjust CPU, RAM, storage and networking as your automation architecture changes. For a queue-mode deployment with Postgres, Redis and several workers, that granularity is genuinely useful — you can add memory to the workers without over-buying everything else.
No one-click n8n setup, so this is a manual Docker build. What you get in exchange is the flexibility to run the full production topology exactly as you want it, across data centres in North America, Europe, Asia and the Middle East.
Entry pricing is around $20.00/month, which is the highest starting point here, and there is no money-back guarantee — though the 30-day free trial is arguably better for evaluation purposes.
Pros: fully configurable server specification; hourly billing and instant resizing; global data centre footprint; 30-day free trial; well suited to queue mode.
Cons: highest entry price on this list; manual setup; no money-back guarantee; the console assumes you know what you want.
Who it’s for: high-concurrency deployments and teams whose resource needs move around.
5. IONOS — most affordable starting plan in Europe
IONOS competes on price, though less sharply than it used to: $4/month for three months on a twelve-month term, then $11 — its old flat-rate $2 VPS XS is discontinued. What you get is 4 vCores and 4GB, with a clean VPS, full root access and a control panel that is noticeably less intimidating than Kamatera’s.
There is no n8n template, so you are doing a Docker install from general documentation. For European deployments the data centre footprint is the draw — if your workflows are hitting EU services or you have data residency requirements, keeping the automation layer in-region matters for both latency and compliance.
At $2.00/month you are getting a small machine. Size up before you enable queue mode.
Pros: cheapest entry price; European data centres; approachable control panel; unlimited traffic; 30-day money-back guarantee; 24/7 phone support.
Cons: no n8n-specific documentation or template; entry tier is small; backups and snapshots are extra.
Who it’s for: European users and anyone optimising hard on monthly cost.
Check IONOS’s current pricing →
6. InterServer — best value for budget-conscious users
InterServer rounds out the list on price-to-headroom. $3.00/month gets a Linux VPS with root access, and the configurations scale up to 16 vCores and 64GB of RAM — which is more ceiling than most of the cheaper options here offer, and relevant if you expect a memory-hungry queue-mode setup later.
No prepackaged n8n plans, so Docker, PostgreSQL and queue mode are all manual. That is fine; the stack is well documented and the provider stays out of the way.
No money-back guarantee is the notable omission.
Pros: low entry price with genuine scaling headroom to 64GB; full root access; flexible configurations; straightforward, no-nonsense provider.
Cons: no money-back guarantee; no n8n template or documentation; fewer managed conveniences than the pricier options.
Who it’s for: budget-conscious self-hosters who expect to scale memory later.
7. Hetzner — best price-to-performance
Hetzner is the value pick for a self-hosted n8n box, and it’s the one the self-hosting community reaches for by default. Roughly $3.49–4.49/month buys 2 vCPU and 4GB on NVMe, which is the working configuration for regular mode rather than a stripped entry tier — and the larger plans stay cheap enough that stepping up to queue mode with Redis and workers doesn’t sting.
Hourly billing with a settable monthly cap makes it a sensible place to build the stack before committing. Attachable volumes are the underrated feature here: when the execution database grows, you add storage rather than migrate the server.
No n8n template, so this is a manual Docker Compose build. Backups and snapshots are charged separately, and account verification at signup is slower than elsewhere.
Pros: best RAM and CPU per dollar here; expandable volumes for a growing database; hourly billing with a spend cap; GDPR-compliant locations; excellent documentation.
Cons: manual setup with no template; backups cost extra; signup verification friction; no Asia-Pacific region.
Who it’s for: anyone comfortable with Docker Compose who wants the most headroom per pound — particularly for queue mode.
If you’re running n8n alongside other self-hosted apps, a PaaS layer like Coolify can manage them together — see our guide to the best VPS for Coolify.
Check Hetzner’s current pricing →
What self-hosting actually saves you
Considering a platform instead of a server? Railway vs a VPS for n8n covers why usage-based billing suits n8n poorly — it’s billed for residency, and n8n never idles.
If you’re weighing two specific hosts, we compare them directly in DigitalOcean vs Hostinger for n8n — including the n8n Cloud break-even and what happens to in-flight executions when you resize.
The argument for self-hosting n8n is usually framed as control, but the money is a large part of it.
n8n’s managed cloud is priced by execution. Self-hosting removes that entirely — you pay for a server, and the number of executions is bounded only by what the hardware can do. For a workflow firing every five minutes, that is roughly 8,600 executions a month from one automation. On a per-execution plan that adds up quickly; on a $6 VPS it is free.
The real total for a self-hosted instance:
| Component | Monthly |
|---|---|
| VPS | $3–12 |
| Backups | $0–3 |
| Your time | The actual cost |
That last row is not a joke. Self-hosting means you own updates, database maintenance, TLS renewal and incident response. If n8n is running something a business depends on, budget for that honestly — either your attention or a managed tier like Liquid Web’s. The hosting is cheap; the operations are not free.
How we picked
We ranked providers on the things that determine whether a self-hosted n8n instance stays healthy: whether a preconfigured template exists, whether you keep full root access for Docker and queue mode, memory headroom at the price points people actually buy, backup and snapshot handling, and data centre locations. Pricing and specifications are the providers’ published rates at the time of writing, with promotional terms flagged rather than quoted as the ongoing cost.
The requirements and configuration guidance come from n8n’s documented behaviour and from the failure modes that recur consistently in self-hosted deployments — the execution-data growth problem in particular is the most common cause of a self-hosted instance falling over, and it is a configuration default rather than a resource limit. Where a provider is a poor fit we say so, including providers we have a commercial relationship with; two of the seven here earn us nothing.
How to choose
- Will you need queue mode? If yes, buy 8GB minimum and plan for Redis plus workers. If no, 4GB is genuinely enough.
- Do you want to install Docker yourself? If not, Hostinger or Bluehost have one-click paths. If you don’t mind, IONOS and InterServer are cheaper.
- Is this business-critical? Then Liquid Web’s managed option is worth the premium over saving four dollars a month.
- Where do your integrations live? EU services and data residency point to IONOS; a global footprint points to Kamatera.
- Configure pruning on day one, whichever you pick. It is the difference between a server that lasts and one that doesn’t.
For most people the answer is Hostinger — the template removes the step where people give up, and the plan is correctly sized. Bluehost if you want that same convenience cheaper and will size up. Liquid Web if it’s production. Hetzner if you want the best price-to-performance and don’t mind Docker Compose. IONOS or InterServer if you’re optimising hard on price. Kamatera if you’re building a genuinely complex, high-concurrency topology.
FAQ
What are the server requirements for self-hosting n8n?
2 vCPU and 4GB of RAM is the working minimum for standard workflows, with 50GB of SSD storage. Move to 4+ vCPU and 16GB if you are running queue mode, high concurrency, or AI-heavy workflows — queue mode in particular runs multiple worker processes plus Redis, so the memory requirement roughly doubles. Use PostgreSQL rather than the default SQLite for anything production.
Why does my n8n database keep growing?
Because n8n saves the complete input and output data of every node in every execution by default, and nothing removes it unless you configure pruning. A workflow running every few minutes will fill a disk over months. Enable execution pruning, set a maximum age of 7–30 days, and consider saving execution data only for failed runs — successful executions are rarely inspected and are the bulk of the volume.
Do I need Redis to self-host n8n?
Only for queue mode. In regular mode n8n runs as a single process and needs no Redis. Queue mode uses Redis to distribute executions across worker processes, which is what you want once executions start backing up or you are handling webhook bursts. For a personal instance running scheduled jobs, regular mode without Redis is the right call.
Is self-hosting n8n cheaper than n8n Cloud?
On execution volume, substantially — self-hosting removes per-execution pricing entirely, and a single workflow firing every five minutes is around 8,600 executions a month. A $6 VPS handles that without counting. What self-hosting adds is operational work: updates, database maintenance, backups and incident response are yours. The hosting is cheap; whether the total is cheaper depends on what your time costs.
Related Blog Posts
