What does platform engineering actually cost?

For a 100-engineer organisation, budget $1.1M to $1.8M per year all-in. About 70 to 85 percent of that is loaded salary for the platform team itself. Tooling usually runs 8 to 15 percent, cloud infrastructure for the platform 3 to 8 percent, and overhead (training, on-call premiums, recruiting amortisation) the remaining 5 to 10 percent. Smaller organisations pay a higher per-engineer rate because the team has fixed minimum cost; larger organisations benefit from specialisation and volume tooling discounts.

Industry consensus from Gartner peer community data and CNCF surveys puts the target at 5 to 10 percent of the engineering organisation, which works out to a platform-to-product ratio of roughly 1:8 to 1:12. Minimum viable team size is two to three engineers, below that you have a single point of failure. Specialisation typically emerges around 80 product engineers, and a dedicated manager becomes necessary around 120.

Typical payback is 18 to 24 months, with a 2x to 5x return on investment when measured against recovered engineering time. The canonical frame is one hour saved per product engineer per day at a loaded hourly rate of around $100, which for a 100-engineer team is $2.2M per year of recovered capacity. DORA research links mature platforms to higher deployment frequency and lower change-failure rates; DX Core 4 benchmarks show 3 to 12 percent efficiency gains. Attribution is harder than vendor blogs suggest, so claims above 5x should be treated sceptically.

Under 20 engineers, almost certainly not yet. Use managed services and conventions; a platform team adds overhead you cannot yet absorb. Between 20 and 50 engineers, one fractional engineer can help with pipeline and cloud. From 50 engineers and up, you are paying the cost of not having a platform team whether you acknowledge it or not. Above 80 engineers, failing to invest is the expensive option.

Neither extreme works well. Pure build means carrying a product team forever to maintain internal tooling with no feature parity to commercial options. Pure buy means integration work that rarely shrinks and licence costs that escalate. Most successful platforms at 200+ engineers are hybrids: commercial core for the capabilities vendors do well (CI/CD, observability, service catalogue) and custom glue for the golden paths unique to your stack. See our /build-vs-buy decision framework for the full treatment.

For a healthy platform at 80 to 200 engineers, expect $11,000 to $18,000 per product engineer per year. That range assumes a standard tooling tier, managed cloud infrastructure, and a platform team compensated at typical US market rates. Numbers below $9k per dev per year often indicate under-investment rather than efficiency; numbers above $25k indicate scale-up growing pains or a team size out of proportion to the engineering org. Our /cost-per-developer page has the full benchmark by org size.

Consulting day rates in the US range from $1,500 to $3,500 depending on seniority and location. A consulting engagement might cost $300k to $600k for a one-off platform bootstrap over six months. A permanent platform team is roughly 1.5x the annual equivalent but delivers ownership, knowledge retention, and iteration that consulting cannot. Most mature organisations use consulting for kickoff or specific migrations, then transition to in-house ownership.

On-call premiums add 5 to 10 percent to loaded salary. Documentation debt costs 3 to 6 months when senior engineers leave. Training is $3,000 to $10,000 per platform engineer per year. Internal advocacy work (roadshows, office hours, scorecard reviews) soaks up half to one FTE once the team exceeds ten engineers. Add 15 to 25 percent to your visible budget to plan for these honestly. Our /hidden-costs page breaks down each line item.