n8n Self-Hosted vs n8n Cloud vs Make: TCO at 10k, 100k and 1M Executions

Affiliate disclosure: optimization.systems earns a commission if you sign up for Make through links on this page, at no extra cost to you. We have no commercial relationship with n8n, AWS or Hetzner, and links to them are not paid. The commission does not change the model: where self-hosted n8n is cheaper, this guide says so.

What you are actually comparing

The three options put the cost in different places:

Option	What you pay the vendor	What you pay for yourself
n8n self-hosted (Community)	Nothing	Servers, Postgres, Redis, backups, monitoring and the engineers who run them
n8n Cloud	Subscription by monthly executions	Workflow building only
Make	Subscription by monthly credits	Workflow building only

n8n bills one execution per workflow run, however many nodes it has. Make bills one credit per module that runs, including the trigger. That difference matters as much as hosting does, so we use the same reference workflow as our Make vs n8n vs Zapier TCO guide: one trigger plus seven actions, no loops — 8 Make credits or 1 n8n execution per run.

Executions per month	n8n executions	Make credits
10,000	10,000	80,000
100,000	100,000	800,000
1,000,000	1,000,000	8,000,000

TCO model: assumptions (dated)

Vendor prices change often. Every figure below is a planning estimate, not a quote.

• Make and n8n Cloud subscriptions use the tier prices in our pricing registry, last verified 1 September 2025, billed annually, with ranges consistent with our three-way TCO guide. n8n Cloud is billed in euros. Above the largest published tiers both vendors quote Enterprise contracts, so those cells are planning ranges.
• Infrastructure uses on-demand list prices as of November 2026: Hetzner Cloud in Germany or Finland, and AWS in us-east-1 with no Reserved Instances or Savings Plans. Euro prices are converted at about €1 = $1.15. Check the AWS Pricing Calculator and Hetzner’s price list before you budget.
• Engineering labor is a loaded rate of $55 to $120 an hour: $55 reflects a contractor or a lower-cost market, $120 a senior DevOps or platform engineer in the US, UK or Australia. Labor covers upgrades, security patches, monitoring, backup restore tests, database maintenance and incident response.
• Excluded: building and maintaining the workflows themselves (broadly similar on all three), 24/7 on-call stipends, and the self-hosted n8n Business or Enterprise license you need for SSO, environments with Git, log streaming or advanced RBAC. If you need those features self-hosted, add the license quote to the self-hosted columns.

Self-hosted n8n: infrastructure by volume

Component	10k executions	100k executions	1M executions
Architecture	Single instance	Queue mode	Queue mode, scaled out
n8n processes	1 VM (4 vCPU, 8 GB)	Main + 2–4 workers	Main, 2 webhook processors, 6–10 autoscaled workers
Postgres	Same VM	Dedicated VM or managed, single node	Highly available primary + replica
Redis	Not needed	Single small instance	Replicated, with failover
Extras	Off-server backups	Load balancer, monitoring, log retention	Load balancer, NAT, metrics, logs, execution-data pruning
Hetzner, per month	~$20–$30	~$100–$150	~$300–$500
AWS, per month	~$55–$75	~$350–$450	~$1,200–$2,000

On AWS, Multi-AZ RDS, ElastiCache, the load balancer and NAT gateways carry most of the premium. On Hetzner you run Postgres and Redis yourself, which is cheaper in cash and costs some of the extra hours below.

Engineering hours by volume

Executions per month	Monthly hours	At $55/hr	At $120/hr
10,000	2–4	$110–$220	$240–$480
100,000	6–10	$330–$550	$720–$1,200
1,000,000	16–30	$880–$1,650	$1,920–$3,600

One-time setup adds 20 to 60 hours for hardening, TLS and reverse proxy, backup automation, deployment pipelines and runbooks — $1,100 to $7,200 depending on rate and scope, or about $90 to $600 a month if amortized over the first year. The tables below leave it out, which flatters self-hosting in year one.

Total monthly cost of ownership

Executions per month	Make	n8n Cloud	n8n self-hosted on Hetzner (infra + labor)	n8n self-hosted on AWS (infra + labor)
10,000	~$60–$110	~$60 (Pro tier)	~$130–$510	~$165–$555
100,000	~$400–$700	~$800–$1,500 (Business tier or quote)	~$430–$1,350	~$680–$1,650
1,000,000	~$2,500–$5,000 (Enterprise quote)	Enterprise quote (plan for $3,000+)	~$1,180–$4,100	~$2,080–$5,600

Self-hosted ranges run from the low end of hours, rate and infrastructure to the high end of all three. Your figure sits wherever your team’s actual hours and rate put it, so the break-even table below is more useful than the ranges.

Break-even: how many hours can you afford?

Self-hosted n8n beats Make when infrastructure + (hours × rate) < Make subscription. Using range midpoints, this is how many engineering hours a month self-hosting can absorb before Make is cheaper:

Executions per month	Hetzner at $55/hr	Hetzner at $120/hr	AWS at $55/hr	AWS at $120/hr	Hours you’ll realistically need
10,000	1.1	0.5	0.4	0.2	2–4
100,000	7.7	3.5	2.7	1.3	6–10
1,000,000	61	28	39	18	16–30

Read across each row. Where the hours you’ll need exceed the break-even, Make is cheaper. Where they fall below, self-hosted n8n is.

At 10,000 executions a month

Self-hosting cannot win on cost: no team maintains a production n8n instance in under an hour a month. Choose between Make (opens in a new tab, affiliate link) and n8n Cloud on fit, not price — they cost about the same. Note that n8n Cloud’s Pro tier includes exactly 10,000 executions, so retries, test runs or growth push you to the next tier, while Make’s 80,000 credits for the reference workflow sit inside its published tiers. Self-host at this volume only for data residency or air-gapped requirements, and budget it as a compliance cost.

At 100,000 executions a month

This is where Make’s managed cloud most clearly wins on total cost. Running n8n in queue mode with Redis, workers and a separate Postgres takes 6 to 10 hours a month, and the break-even is 1.3 to 7.7 hours. On AWS, self-hosting loses at any realistic rate. On Hetzner, it can only come out ahead if an existing platform team runs it as one more service at a low effective rate. n8n Cloud’s Business tier also typically costs more than Make for this workload.

If your team would be standing up servers, databases and on-call just for automation, Make (opens in a new tab, affiliate link) is the lower-TCO choice and removes an operational risk as well.

At 1,000,000 executions a month

Self-hosted n8n wins. Make’s 8 million credits a month land in Enterprise pricing, while n8n’s cost stops growing with step count. On Hetzner, self-hosting is cheaper at any realistic hours and rate. On AWS it is cheaper for most teams, but at $120 an hour and more than 18 hours a month — common for a highly available setup — Make Enterprise comes close or wins. n8n Cloud Enterprise is the middle path when you want n8n’s billing model without running Postgres and Redis yourself; get a quote and compare it to the AWS column.

What moves the numbers

• Loops and long workflows. Make bills every module on every iteration; n8n bills the run once. A workflow that loops over 50 line items can cost 250+ Make credits per run, which moves the Make break-even down by roughly one volume tier. If your workflows loop, self-hosted n8n wins earlier than this model shows.
• Polling triggers. Make spends a credit every time a scheduled trigger checks for new data, even when there is nothing to process. Use webhooks wherever you can.
• Execution data retention. Self-hosted n8n stores execution history in Postgres. At 1M executions, without pruning (EXECUTIONS_DATA_PRUNE, EXECUTIONS_DATA_MAX_AGE) the database grows by tens of gigabytes a month and slows down.
• Reserved capacity. One-year AWS Savings Plans or Reserved Instances cut compute and RDS costs by roughly 30–40%, which helps the AWS column at 1M.
• Existing platform. If n8n runs on a Kubernetes cluster, Postgres fleet and monitoring stack you already operate, its incremental hours fall sharply and self-hosting wins earlier.
• Enterprise features. SSO, Git-based environments and audit logs on self-hosted n8n need a paid license. If you need them, the “free” license column is no longer free.

Comparison at a glance

Make vs n8n compared

Tool	Best for	Starting price	Free plan	Deployment	AI capabilities	Score	Link
Make Top pick	Ops teams that need complex, multi-branch logic at low cost	$9/mo		cloud	AI Agents, OpenAI/Anthropic modules, AI-assisted scenario building	4.6	Visit Make (affiliate link, opens in new tab)
n8n	Engineering-led teams with data-residency or volume needs	$24/mo		cloud, self-hosted	LangChain-based AI Agent nodes, vector stores, tool calling, MCP	4.6	Visit n8n (affiliate link, opens in new tab)

Beyond cost: what else should drive the decision?

Factor	n8n self-hosted	n8n Cloud	Make
Who operates it	Your engineers	n8n	Make
Data residency	Any region or on-premises	n8n’s hosting region (EU)	Make’s US or EU hosting
Uptime responsibility	Yours, including on-call	Vendor	Vendor
Security patching	Yours	Vendor	Vendor
Custom code	JS and Python Code nodes, custom nodes	JS and Python Code nodes	Limited custom functions
Billing unit	None (infrastructure only)	Execution	Credit per module
Best for	Platform teams, compliance, very high volume	Engineering teams that don’t want to run servers	Ops and RevOps teams scaling to six-figure volumes

A decision rule you can apply this week

1. Count executions and Make credits per run for your three highest-volume workflows.
2. Project volume 12 months out, not today’s run rate.
3. Pick your hosting — Hetzner or AWS — and price the infrastructure tier from the table above.
4. Estimate monthly engineering hours honestly, including on-call, and multiply by your loaded rate between $55 and $120.
5. Compare against the break-even table. If your hours are above the break-even, choose the managed option; if below, self-host.

As a rule of thumb: up to about 100,000 executions, or no platform team → Make; engineers who want n8n without servers → n8n Cloud; 1,000,000+ executions, loops, or strict data residency → self-hosted n8n. Use the ROI estimator to put a value on the hours automation frees up, then subtract the hours self-hosting consumes.

The bottom line

n8n’s Community edition costs nothing to license and real money to run. Below about 100,000 executions a month, those running costs exceed a managed subscription, and Make (opens in a new tab, affiliate link) delivers the lowest total cost of ownership. At 1,000,000 executions, self-hosted n8n wins — decisively on Hetzner, narrowly on AWS — for teams that can staff it. Re-run the break-even table whenever your volume grows tenfold or your engineering rate changes, and check the ROI estimator before committing either way.