Azure API Management is one of those services that quietly drains budgets. Not because it's expensive by nature, but because the jump between tiers is enormous, and most organisations end up on Premium without ever questioning whether they actually need it.
We regularly find APIM Premium instances running in environments where Standard or even Developer would do the job perfectly well. At roughly £2,500 per month per unit, that's not a rounding error. Across multiple environments, it becomes one of the biggest line items on the Azure bill.
The tier landscape
Azure APIM has five classic tiers plus the newer v2 tiers. The pricing gaps are significant, and the v2 tiers have changed the game entirely.
Consumption is the pay-per-call option. You pay for what you use, there's no standing charge, and it's the cheapest route for low-volume APIs. The trade-offs are real though: no SLA, cold start latency on idle APIs, and limited features. It works well for internal tools or proof-of-concept work, but it's not where you'd run production traffic.
Developer comes in at around £40 per month. This is the tier that most people overlook, and it shouldn't be. Developer has the full portal experience, the full policy engine, all the same configuration capabilities as Premium. The only things missing are the SLA and the scaling. For every non-production environment, Developer is the right answer. There is almost no reason to run Dev, Test, or UAT on anything higher.
The argument for running non-prod on a higher tier is usually "we need to load test against something representative." Fair enough for one specific load-testing phase, but that's a temporary need. You can scale up for a week of testing and scale back down. Running Premium 24/7 in a test environment to cover a load test you run once a quarter is paying £30,000 a year for something you need for a few days.
Basic sits at roughly £120 per month. You get an SLA, which is the big step up from Developer, but no VNet integration of any kind. For straightforward, publicly accessible APIs that don't need to talk to private backends, Basic is a solid option.
Standard costs around £550 per month for one unit. Standard supports private endpoints for inbound traffic. It also gives you higher throughput limits and a production-grade SLA.
Premium is the top tier at approximately £2,500 per month per unit. This is where you get VNet injection, multi-region deployment, and availability zones.
The v2 tiers: the game changer
Microsoft has released v2 versions of these tiers, and they fundamentally change the cost equation for private networking.
Standard v2 now supports both inbound private endpoints AND outbound VNet integration. That means your APIM instance can receive traffic privately and reach backends sitting in your VNet, all without Premium. This is the capability that used to be the sole justification for Premium, and it's now available at roughly a fifth of the price.
You can combine inbound private endpoints with outbound VNet integration on Standard v2 for full end-to-end network isolation. No VNet injection required. No Premium required.
The maths is compelling. Two Standard v2 instances in different regions or availability zones give you redundancy and geographic distribution for roughly £1,100 per month. That's less than half the cost of a single Premium unit, and you get actual redundancy rather than a single compute node.
The single-unit problem
Here's something that doesn't get talked about enough: Premium APIM defaults to a single compute unit. One node. No redundancy. You're paying £2,500 per month for a service that has a single point of failure.
We have seen APIM outages caused by Microsoft platform updates hitting that single unit. Azure performs maintenance on APIM infrastructure, and on a single-unit Premium instance, there is no second node to failover to during the update. These updates can happen during business hours, which means your API gateway goes down during the working day. For a service costing £2,500 per month, that's a poor outcome.
To get actual redundancy on Premium, you need to add a second unit. That's £5,000 per month. For the same money, you could run multiple Standard v2 instances with full private networking and genuine redundancy.
Why everyone ends up on Premium
There's usually one reason organisations land on Premium: VNet integration. If your APIs need to call backend services sitting on private virtual networks, historically the only way to achieve that was VNet injection, and that meant Premium.
This made sense when it was the only option. The problem is that many organisations chose Premium years ago, the requirement landscape has shifted with the v2 tiers, and nobody's gone back to reassess.
The four-environment trap
Here's where the real damage happens. Most organisations run at least four environments: Dev, Test, UAT, and Production. If all four are on Premium, you're looking at £10,000 or more per month just for API Management. That's £120,000 a year before you've processed a single API call.
The fix is straightforward: every non-production environment should be on Developer tier. Developer has the same policy engine, the same portal, the same configuration format, and the same feature set as Premium. The only difference is the SLA and the scaling, neither of which matter for non-production workloads.
Your CI/CD pipeline deploys the same configurations regardless of tier. Your developers test the same policies, the same products, the same subscriptions. Moving Dev, Test, and UAT from Premium to Developer saves £7,500 per month immediately. That's £90,000 a year, and the change takes five minutes per environment.
Alternative patterns that avoid Premium
Even for production, Premium is rarely necessary with the v2 tiers available.
Standard v2 with VNet integration and private endpoints. This is the big one. Standard v2 supports outbound VNet integration (your APIM can reach private backends) and inbound private endpoints (only traffic from your VNet can reach APIM). Combined, this gives you the same end-to-end network isolation that used to require Premium. The cost saving is roughly £2,000 per month.
Two Standard v2 instances for redundancy. Instead of paying £2,500 for a single Premium unit with no redundancy, deploy two Standard v2 instances. You get genuine redundancy (two independent compute nodes), full private networking on both, and you still come in cheaper than a single Premium unit. Front them with Application Gateway or Traffic Manager for failover.
Self-hosted gateway for hybrid scenarios. If you have a specific on-premises or hybrid requirement, the APIM self-hosted gateway lets you run a containerised gateway component wherever you need it. You keep your management plane on Standard and deploy lightweight gateways into your private networks. This is particularly useful when the VNet injection requirement is driven by a handful of legacy backends rather than a wholesale private networking need.
When Premium actually makes sense
We're not saying Premium is never justified. There are a small number of scenarios where it's genuinely the right choice.
Genuine multi-region deployment with built-in failover. If you need active-active API Management across multiple Azure regions with automatic failover managed by the platform, that's a Premium feature. Though the two-Standard-v2-instances pattern achieves something similar at lower cost if you're willing to manage the failover yourself.
Availability zones within a single region. If your compliance requirements specifically mandate zone-redundant deployment for API Management, that's Premium. But remember, a single Premium unit is still a single node. You need multiple units to actually benefit from zone distribution.
The key question is whether you genuinely need these capabilities, or whether they were selected years ago when the alternatives didn't exist.
Quick wins to action today
If any of this sounds familiar, here's what to do about it.
Move all non-production environments to Developer tier. This is the single biggest win, and it's low risk. Developer has every feature of Premium except the SLA and scaling. Your configurations, policies, and APIs work identically. Do this first, save immediately.
Evaluate Standard v2 for production. If your current Premium justification is "we need private networking," Standard v2 now covers both inbound and outbound. Review whether v2 meets your requirements. In most cases, it will.
Check your unit count. If you're on Premium and genuinely need to stay there, check how many units you're running. We've seen instances running two or three units when traffic analysis shows one would suffice. Each unnecessary unit is another £2,500 per month. Though if you're running a single unit, consider whether the lack of redundancy is an acceptable risk.
Consider the two-Standard-v2 pattern. Two Standard v2 instances with VNet integration cost less than one Premium unit and give you actual redundancy. If you're currently running single-unit Premium, this is a strictly better architecture at a lower price.
The broader pattern
APIM is a textbook example of a broader FinOps problem: infrastructure decisions made once and never revisited. The original architect chose Premium for good reasons at the time. But Azure evolves, new tiers appear, and what was the only option three years ago might now be one of several, and the most expensive one at that.
The organisations that control their cloud costs aren't the ones that never overspend. They're the ones that regularly look back at decisions and ask, "Is this still the right choice?" For API Management, with the v2 tiers now available, that question is worth asking today.
Not sure if your APIM tiers are right-sized? Our free Azure cost assessment will identify exactly where you're overspending, including API Management, reserved instances, and orphaned resources. No commitment, just clarity.