Executive Summary
Adoption is not a business case. Most organizations already run an internal developer platform, and that number says nothing about return. A platform with no line on the budget becomes a cost center nobody defends when cuts arrive. The argument, not the build, is the job.
DORA’s 2025 research found that 90 percent of organizations run an internal developer platform and 76 percent keep a dedicated platform team. A 2025 VMware survey found that 29.6 percent of teams do not measure platform success at all, and another quarter admit to no product mindset. Backstage hands over a catalog and a scaffolder, Port sells a hosted catalog, and Humanitec sells orchestration behind whichever portal you run. The question that does not make it onto the slide is whether the platform pays for itself. A team of five supporting three hundred developers can cut onboarding from five days to one and return four days for every new hire. Halve lead time and revenue moves sooner. Build the case on three numbers, cost per service per month at the namespace, lead time from commit to production, and support load from tickets the golden path failed to prevent. Take the baseline before you ship.
Platform engineering won the argument. The proof is a number. DORA’s 2025 research found that 90 percent of organizations now run an internal developer platform, and 76 percent keep a dedicated platform team. That is adoption, and adoption is not a business case. A platform with no line on the budget is a cost center nobody defends when the next round of cuts lands.
The gap shows up in measurement. A 2025 platform engineering survey from VMware found that 29.6 percent of teams do not measure platform success at all. A further quarter say they have no product mindset. Teams that ship a portal and never price it end up running a service with no owner and no return.
Velocity Is the Product, Not the Portal
Internal developer platform ROI comes from hours handed back to developers. DORA ties platform quality to delivery performance. When the platform works, lead time falls and deployment frequency rises. When it does not, developers route around it. A portal nobody opens is worse than no portal, because you paid for the hosting and the people.
The tooling market makes the confusion easy. Backstage, a CNCF incubating project that Spotify started, hands you a catalog, a scaffolder, and a plugin surface. It does not hand you a checkout path or a cost model. Port sells a hosted catalog with scorecards and self-service actions. Humanitec sells orchestration that sits behind whichever portal you already run. Each one answers a different question. None of them answers whether the platform pays for itself.
Here is the shape of a real case. A platform team of five supports three hundred developers. If a golden path cuts onboarding from five days to one, that is four days for every new hire, every quarter. If it halves lead time, features reach customers sooner and revenue moves with them. Neither number needs a vendor slide. Both need a baseline you took before the platform shipped.
DORA’s own advice is to treat the platform as a product and developers as its customers. That framing only works if you know what the customers cost to serve. A team with no price per service sets its roadmap against requests rather than value. Requests come from the loudest team, and the loudest team is rarely the one with the biggest problem.
Kubernetes Economics Are Where the Case Lives
Most platforms are really Kubernetes platforms, and that is where the money moves. Kubernetes, a CNCF graduated project, schedules workloads well and tracks cost not at all. Clusters multiply for good reasons and get left running for none. Every one of them keeps billing. OpenCost, a CNCF incubating project, reads spend straight from the cluster API. That gets you a namespace-level number instead of a monthly surprise.
Cost accountability is the harder half. Kubernetes bills arrive as one number from the cloud provider and one from the data center. Nobody can say which team caused either. Showback fixes that. Publish a monthly namespace bill, split by team, and watch how fast requests for oversized clusters disappear. A platform team that does this stops arguing about its budget and starts defending a number that beats it.
Build the case on three numbers. Cost per service per month, attributed at the namespace. Lead time from commit to production, measured the way DORA measures it. Support load, meaning the tickets the golden path failed to prevent. Teams that skip the measurement step usually report higher spend in year two rather than lower.
Put the Platform on the Books
The vendor market rewards the opposite habit. Portals ship in weeks and demo well. The measurement layer ships in quarters and shows up as a spreadsheet. One of those wins the budget meeting when nobody asks for the return.
Ask for the numbers before you ask for the headcount. A platform plan with a cost baseline, a velocity target, and a named owner signs itself. A plan that promises better developer experience and nothing else loses to the next pitch that promises the same thing. The build is the easy part. The argument is the job.
Platform engineering is one of the forces in the 2026 State of Enterprise Infrastructure report, which covers how the operating layer changed and why maturity tracks with outcomes.
Related reading. Your Coding Agent’s Biggest Bill Is the Data You Feed It. Your AI Pilot Died From Infrastructure, Not From the Model.. Cortex Builds an Internal Developer Portal That Answers Your Own Questions. The 2026 State of Enterprise Infrastructure.
Get the next one before it is old news
Independent analysis of cloud-native infrastructure, Kubernetes and data centre economics. No vendor spin.