What if the most expensive part of your cloud automation stack is not the tool at all, but the way you’re paying for it?
That question trips up a lot of teams. They compare licence fees, pick the cheaper number, and then discover six months later that per-resource metering, per-seat charges, and support contracts have quietly tripled the bill. Cloud infrastructure automation pricing is genuinely confusing, because different vendors charge for completely different things. Some bill per user. Some bill per managed resource. Some bill per resource-hour. Some charge nothing for the software and everything for the support.
This comparison breaks down how the major tools actually structure their pricing, where the hidden costs hide, and what to check before you sign anything.
Why Cloud Automation Pricing Is Harder Than It Looks
Most software pricing is simple: a number per seat, per month. Infrastructure automation resists that model because the work itself is variable. One team might manage forty resources. Another manages forty thousand. A flat per-seat price either overcharges the small team or bankrupts the vendor serving the large one.
So vendors invented metering. And metering is where budgets go to die.
There is a second complication. Many of these tools have open-source cores that are free to run, with paid layers wrapped around them. That means the “price” of a tool depends entirely on which version you adopt and how much operational work you’re willing to do yourself. Self-hosting Terraform or OpenTofu costs nothing in licence fees, but someone has to run the state backend, manage locking, handle upgrades, and babysit the pipeline. That labour is real money even when it never appears on an invoice.
The Three Pricing Models You’ll Actually Encounter

Almost every tool in this space falls into one of three buckets.
| Pricing model | How it works | Typical examples | Where costs surprise you |
|---|---|---|---|
| Free core, paid support | Software is open source; you pay for subscriptions or enterprise support | Ansible, Chef, Puppet, Crossplane | Support tiers are quote-based, so budgets are hard to forecast |
| Per-user plus per-resource | Seat fee on top of a metered charge for managed resources | Terraform Cloud, Pulumi | Resource counts grow silently as teams add environments |
| Consumption-based | You pay for underlying cloud resources, not the tool | AWS CloudFormation | Costs hide inside your general cloud bill |
The first model looks cheapest and often is, provided you have the engineering capacity to operate it. The second is the most common source of bill shock. The third is the easiest to overlook because the tool itself never sends you an invoice.
Terraform and OpenTofu: The Licensing Split That Changed Budgets
In August 2023, HashiCorp announced it was moving Terraform from the Mozilla Public License to the Business Source License 1.1. The BSL restricts use in competing commercial products, and the licence converts to MPL 2.0 after four years. That announcement reshaped the economics of the category almost overnight.
The community response was OpenTofu, an open-source fork released under MPL 2.0 and generally available in 2024. Because OpenTofu is free to self-host, its price is infrastructure and operational cost rather than licence fees. For teams with strong platform engineering, that’s a genuine saving. For teams without it, the operational burden can outweigh the licence saving.
Terraform Cloud and Enterprise remain tiered products, with per-user and per-managed-resource charges applying on paid tiers. The important detail is not the number but the structure: you are billed on two axes at once. Adding a contractor increases the seat count. Adding a staging environment increases the resource count. Both move at the same time, which is why forecasts drift.
One practical warning: exact per-unit prices in this category have changed repeatedly since 2022. Any figure you read in a blog post, including this one, should be treated as indicative and dated. Check the vendor’s current documentation on billing and usage before you build a budget around it.
Pulumi: Usage-Based Pricing With a Free Individual Tier
Pulumi takes a different route. The core engine is open source under Apache 2.0, so running the CLI yourself costs nothing. The managed service uses usage-based pricing with a free Individual tier and paid Team and Enterprise tiers, charged per user and per resource-hour.
That resource-hour metric deserves attention. It is not the same as a resource count. A resource that exists for a full month accrues far more hours than one spun up for a test. Teams that leave ephemeral environments running overnight, or forget to tear down preview stacks, pay for that forgetfulness directly. It’s a model that rewards disciplined environment hygiene and punishes sprawl.
Ansible, Chef and Puppet: Open Source With Paid Support
Ansible’s core is open source under GPLv3, with paid support delivered through Red Hat Ansible Automation Platform subscriptions. Pricing is subscription-based and not publicly listed per unit; quotes come through sales. Chef’s Infra Client moved to a permissive Apache 2.0 licence in 2019, and Puppet’s core followed in 2023 with Puppet 8. Commercial support remains paid in both cases.
For budgeting purposes, the honest framing is this: the software is free, and the support contract is a negotiation. That’s good news if you have leverage and bad news if you need a predictable number on a spreadsheet by Friday. If you’re building a broader tooling budget, the same flat-fee-versus-metered tension shows up in affiliate marketing tools pricing comparisons, where flat fees and usage charges create the same forecasting headaches.
CloudFormation and Crossplane: Free Runners, Different Costs
AWS CloudFormation carries no separate service charge. You pay for the underlying resources it provisions, and AWS states this plainly in its documentation. That makes CloudFormation deceptively cheap to adopt and easy to underestimate, because the cost simply migrates into your general cloud bill where nobody attributes it to automation.
Crossplane is open source under Apache 2.0 and free to self-host, with commercial offerings adding paid tiers. It takes a control-plane approach, which means your Kubernetes cluster becomes the automation engine. That’s elegant if you already run Kubernetes and an extra operational surface if you don’t.
Neither tool charges you for the automation itself. Both charge you in architectural commitment.
Where the Hidden Costs Actually Hide
Across every model, the same five cost drivers recur. Watch for them.
- State and backend infrastructure. Remote state storage, locking services, and the compute to run plans all cost money, even when the automation tool is free.
- Seat creep. Per-user pricing punishes success. Every new hire, contractor, and auditor who needs read access adds a line item.
- Resource sprawl. Metered resource counts grow with every new environment, region, and ephemeral stack that nobody cleans up.
- Support tier jumps. Enterprise support is often priced in bands, and crossing a threshold can cost more than the extra seats.
- Engineering time for self-hosting. The invisible line item. A free tool maintained by two engineers is not free.
If you want a sense of how per-unit metering behaves across other software categories, the pattern in AI automation tools pricing comparisons is strikingly similar: the headline number is rarely the number you pay.
A Practical Comparison Framework
Rather than chase current prices, compare tools on structure. Here’s a framework that survives price changes.
| Question to ask | Why it matters |
|---|---|
| What is the billing unit? | Users, resources, resource-hours, or nothing at all |
| Does the free tier cover production? | Many free tiers are for individuals, not teams |
| How does cost scale with environments? | Staging and preview environments are the usual culprits |
| Is support included or quoted separately? | Quote-based support breaks annual budgeting |
| What does self-hosting cost in labour? | Two engineers is a real, recurring expense |
| Can you forecast a 12-month bill? | If not, the pricing model is fighting you |
Run every candidate through these six questions before you run a demo. The answers will eliminate options faster than any feature comparison.
Common Mistakes Teams Make
The first mistake is comparing licence fees instead of total cost of ownership. A free tool with a dedicated maintainer is often more expensive than a paid tool without one.
The second is ignoring the resource axis. Teams budget for seats, then get blindsided by managed-resource charges after a reorganisation adds three environments.
The third is assuming free tiers scale. They usually don’t, and the jump from free to paid is frequently steeper than the jump between paid tiers.
The fourth is forgetting that licensing changes happen. The Terraform BSL shift in 2023 caught plenty of organisations off guard. Build a licensing review into your annual planning, and keep an eye on how similar shifts affect adjacent categories, much as proposal tools pricing comparisons now factor licence risk into vendor selection.
The fifth, and most expensive, is treating automation tooling as an IT cost rather than a platform investment. The teams that get this right budget for the operating model, not the invoice alone.
What Actually Works in Practice
The pattern that holds up across organisations is unglamorous: pick a billing model you can forecast, then invest in the discipline the model rewards.
If you choose per-resource metering, automate environment teardown from day one. Tag every stack with an owner and an expiry. If you choose per-seat pricing, consolidate service accounts and audit access quarterly. If you self-host, budget the engineering hours explicitly and name the person accountable for upgrades.
And date every number. Pricing pages change without notice, and a comparison written eighteen months ago may describe a product that no longer exists in that form. When you present a budget, state the month and year each figure was checked. It’s the single habit that separates a credible forecast from a guess.
Conclusion
Cloud infrastructure automation pricing rewards teams who read the billing structure rather than the headline number. Open-source cores keep licence costs near zero but shift spend into engineering time. Managed platforms bundle convenience and charge for it on two axes at once. Consumption-based tools hide their cost inside your cloud bill entirely.
None of these models is inherently better. The right choice depends on whether you can forecast the bill, staff the operations, and absorb the risk of a licensing change. Answer those three questions honestly, and the pricing comparison becomes much simpler: you’re no longer shopping for the cheapest tool, you’re choosing the cost structure you can actually manage.
Frequently Asked Questions
Is Terraform still free to use?
The Terraform CLI remains free for most individual and internal business use, but the licence changed to the Business Source License 1.1 in August 2023, which restricts use in competing commercial products. The licence converts to MPL 2.0 after four years. Organisations with commercial products built on Terraform should review the terms carefully, and many have migrated to the OpenTofu fork instead.
What is the cheapest way to run infrastructure automation?
Self-hosting an open-source tool such as OpenTofu, Ansible, or Crossplane carries no licence fee, so it is usually the cheapest option on paper. The catch is operational labour: state management, upgrades, security patching, and on-call coverage all cost real money. For small teams without dedicated platform engineers, a paid managed tier is often cheaper in total.
Why do automation tool bills grow without new users?
Because many tools meter on resources as well as seats. Every new environment, region, preview stack, or ephemeral test cluster adds to the managed-resource count or resource-hours consumed. Growth in infrastructure footprint shows up on the bill even when headcount stays flat. Automated teardown and expiry tagging are the most effective controls.
Does AWS CloudFormation cost extra?
No. CloudFormation itself carries no separate service charge; you pay only for the underlying AWS resources it provisions. The practical implication is that CloudFormation costs are invisible as a line item and get absorbed into your general cloud spend, which makes them easy to overlook during budget reviews.
How often should I re-check automation tool pricing?
At least annually, and immediately after any licensing announcement from a major vendor. Per-unit prices in this category changed repeatedly between 2022 and 2024, and the Terraform licence shift in 2023 reshuffled the entire market. Record the month and year of every figure you rely on so your forecast stays honest.
Is per-user or per-resource pricing better?
It depends on which axis grows faster in your organisation. Per-user pricing suits teams with large, stable infrastructure and few operators. Per-resource pricing suits small teams managing sprawling estates, provided they enforce teardown discipline. The safest approach is to model both against your actual headcount and resource counts before committing.