Backstage cost in 2026: $200k to $1.2M all-in for a self-hosted IDP

What "Backstage cost" actually means

Spotify open-sourced Backstage in 2020 and the project moved to CNCF incubation in 2022. Today it is the de facto reference implementation of an Internal Developer Platform: a unified front-end for the service catalogue, software templates, documentation, tech-radar, and a long tail of plugins.

Because the licence is free, the search query Backstage cost is almost always asking the wrong thing. The right question is: what does it cost to run Backstage well enough that 100 product engineers will actually use it on a Tuesday afternoon? That number is heavily a headcount number, modestly a cloud-infrastructure number, and a small but real upgrade-tax number on top of those. Get those three right and you have your annual.

The headcount line, broken down

Across the public adoption write-ups that have meaningful detail (Spotify's own engineering blog, Roadie's adoption case studies, the CNCF Platforms White Paper, and the Team Topologies community), the consistent pattern is that Backstage at year one needs the equivalent of two to four platform engineers working largely full-time. After the initial scaffolding, the first three to five plugins, and the integration with your authentication and source-control systems, you can usually drop to one to two full-time-equivalents of ongoing effort, with the rest of the platform team contributing intermittently.

At the senior platform-engineer loaded rate of about $234,000 per year (see /salary for the multi-source reconciliation), that translates to a year-one personnel line of $470k to $940k for two to four engineers, and a steady-state line of $234k to $468k for one to two. Most of the visible platform cost for a 100-developer organisation is the platform team itself; that pattern holds for the Backstage portion of the platform.

Where the year-1 range comes from

The low end of $200k a year corresponds to an org that already has a mature platform team running other tooling and is adding Backstage as a unified front-end. In that case Backstage is one engineer's primary focus and a fraction of two others' time, total infrastructure is in the low five figures, and most plugin work is wrapping existing internal APIs rather than building from scratch.

The high end of $1.2M a year corresponds to an org standing Backstage up from zero with no existing service-catalogue source of truth, needing to author its own scaffolder templates from scratch, integrate with a fragmented cloud footprint, and run a small adoption push (office hours, internal docs, scorecard pilots). That is two to three engineers full-time plus a fraction of a manager plus visible cloud infrastructure plus the first batch of plugin development.

Most organisations land somewhere in the middle: about $450k to $700k for year one, dropping to about $250k to $400k for years two and three.

Cloud and infrastructure

Backstage itself is a Node application that scales horizontally with quite modest resource requirements. A typical production deployment for a 100-engineer organisation is a small Kubernetes deployment (two to four pods), a small Postgres database, an object store for documentation artefacts, and an authentication integration. Total cloud cost for the Backstage app itself is usually in the $15k to $35k a year range, comfortably under 10 percent of the headcount line.

The catch is that Backstage is rarely the only thing you are running. If Backstage is fronting workflows that themselves spin up cloud resources (the scaffolder provisioning a new repository, a Helm chart, a database, an environment) then the second-order cloud cost runs through whatever budget owns those resources. That is usually accounted for elsewhere, but a clean platform-cost picture should flag it.

The plugin economy

Backstage ships with a useful default plugin set (catalogue, scaffolder, TechDocs, search, kubernetes, cost-insights, GitHub-actions) and a long tail of community plugins of variable quality. Most organisations of any size end up authoring at least three to five custom plugins by year two: typically a deeper integration with their identity provider, a plugin wrapping an internal API (deployments, environments, on-call rotations), and one or two integration plugins with third-party SaaS that the stock plugin does not cover.

Plugin development cost compounds quickly. A simple read-only catalogue plugin runs about two to three engineer-weeks, or $20k to $35k loaded. A scaffolder template (a working golden path) is four to six engineer-weeks, or $40k to $65k. A deep integration plugin with auth, error handling, and tests is closer to two to three engineer-months, or $50k to $90k. By year three a typical mid-sized organisation has spent $300k to $600k cumulatively on plugin development, embedded inside the platform team's annual operating cost.

The upgrade tax

The Backstage project releases a minor version every two to three weeks. Most upgrades are mechanical and consume an hour or two of a platform engineer's time. Breaking-change cycles arrive every few months and individually consume one to two engineer-weeks across all your plugins, sometimes more if you maintain ten or more first-party plugins. The realistic budget is about 15 to 20 percent of one platform engineer's time devoted long-term to keeping Backstage core current.

The expensive failure mode is falling more than a year behind upstream. The path back to current then exceeds the path forward would have been, typically two to three engineer-months consumed in a single catch-up sprint. The discipline that prevents this is committing to a regular upgrade cadence (monthly or bi-monthly) and treating upgrade time as non-negotiable roadmap time, not as something to slip when product pressure rises.

Build versus hosted Backstage

Hosted Backstage offerings (Roadie, Frontside, Liatrio, others) remove the operating cost of running Backstage itself: install, upgrades, hosting infrastructure, identity-provider integration, and the on-call burden. They do not remove plugin authoring, content debt, or adoption work. For an organisation with a small platform team (under five engineers) or limited Kubernetes operations experience, hosted Backstage typically wins on cost in years one and two, costing $30k to $300k a year depending on seats and entities.

At larger platform-team sizes (eight or more engineers) and at scales where you start to author significant numbers of plugins, the operating cost of self-hosted Backstage falls below the hosted licence, and the crossover usually happens around 300 product engineers. For the full apples-to-apples comparison see /backstage-hosted-cost.

How Backstage compares to commercial IDPs

Commercial Internal Developer Platforms (Port, Cortex, OpsLevel, Compass, Humanitec) are usually cheaper than self-hosted Backstage at year one because you skip the platform-engineer staffing needed for installation and plugin authoring. At year three, when a self-hosted Backstage deployment is mature and the operating cost is amortised across a larger platform team, the cost-per-developer-served can be lower than commercial IDP licences, especially at organisations over 300 engineers.

For dedicated cost framings of each commercial option see /port-cost, /cortex-cost, /opslevel-cost, /compass-cost, and /humanitec-cost. For the broader question see /build-vs-buy.

Honest caveats

A handful of things that get glossed over in Backstage adoption writing and that tend to bite teams in year two:

• Service-catalogue data is only valuable if it is fresh. Without a discipline of automating entity sync from source-of-truth systems, the catalogue drifts and engineers stop trusting it within six months.
• TechDocs adoption needs writers, not just a publishing pipeline. The platform team rarely owns the writing; if no-one does, the docs section sits empty.
• The scaffolder templates need a maintenance owner. Templates rot faster than service code because they accumulate references to whichever frameworks were modern when the template was written.
• The plugin contribution model needs to be defined. If only the platform team can write plugins, plugin authoring becomes a bottleneck; if anyone can, the plugin quality varies wildly.

Bottom line

For a 100-engineer organisation, budget about $450k to $700k for year-one self-hosted Backstage all-in, dropping to about $250k to $400k a year steady state from year three onward. About 85 to 92 percent of that is platform-team headcount. Cloud infrastructure is a small line. The upgrade tax is small but non-zero and best treated as scheduled roadmap time.

If the dollar figures above feel high and the alternative looks like one of the commercial IDPs, see the per-vendor pages (linked above) and run the numbers in /calculator. If the figures feel low, the most likely missing line is plugin development, which compounds across years and is easy to under-budget.

All figures verified as of 2026-05-11. Source links: Backstage project site, CNCF Platforms Working Group, Spotify Engineering Blog, BLS OEWS wage data.