Latest / The Cloud Business Podcast with Fexingo: AWS, Azure, GCP, and Enterprise Infrastructure / Why Cloud Contracts Now Charge for API Call Volume
Transcript
- Lucas: There is a clause showing up in enterprise cloud contracts that barely existed three years ago, and it is going to hit any team running a microservices or serverless architecture right in the monthly bill. I'm talking about per api call charges — not just for third-party APIs, but for calls between your own services inside the cloud provider's network. Luna: Wait — you mean we are now getting charged for internal API calls that previously were just bundled into compute or data transfer costs? Lucas: Exactly. Historically, if you had a Lambda function calling another Lambda function, or a container talking to a database via REST, the cost was absorbed into the compute time and the data egress. But starting around late 2024, AWS began introducing per-call pricing on API Gateway that made it explicit. Azure followed with a new API Management consumption tier in early 2025, and GCP has been testing metered API usage under their Apigee platform. Luna: I actually saw a real-world impact of this recently. A friend of mine runs the cloud infrastructure for a mid-size SaaS company — about 50 microservices. At their last contract renewal with AWS, their monthly API call bill jumped 37 percent compared to the previous agreement. Lucas: That is exactly the kind of number I've been hearing from multiple FinOps teams. And the reason is structural: cloud providers are under pressure to increase revenue per customer without raising the headline prices on compute or storage, which are too competitive. So they are unbundling services that used to be free or near-free — and API calls are a perfect candidate because most organisations have no idea how many they make. Luna: So this is essentially a way to monetize the shift to microservices and serverless, where applications are chatty by design? Lucas: That is exactly how I would frame it. A monolithic app might make a few thousand API calls per transaction. A microservices equivalent can make tens or even hundreds of thousands. So the cloud provider sees that traffic pattern and says, 'We can meter that.' And once it's metered, it becomes a line item, and you can bet it ends up in the contract as a separate charge with its own escalation clause. Luna: What does that escalation clause actually look like in practice? Is it a fixed per-call price that goes up, or is it more complex? Lucas: Most commonly, it is a tiered structure. You get a certain number of calls included in your committed spend — say, 10 million calls per month — and then you pay a per-call rate above that. The tricky part is the rate often increases year-over-year by a fixed percentage, like 5 to 10 percent annually, regardless of inflation or your volume. Some contracts I've seen also include a 'minimum call commitment' — you have to pay for at least that many calls whether you use them or not. Luna: That sounds like a version of the old 'data gravity waste' clause we talked about in episode 77, but applied to API calls instead of data storage. Lucas: It is a very similar logic. And it creates the same kind of lock-in: once you have architected your application to use a specific provider's API gateway, internal service mesh, or event bus — like AWS EventBridge or Azure Event Grid — it is costly to change. The API calls themselves become sticky. So the provider can start charging more for that stickiness. Luna: What about Google Cloud? Their Apigee platform has always been more api management focused. Are they leading or following here? Lucas: GCP has actually been more conservative on this. Their Apigee pricing has historically been based on the number of API proxies deployed, not per-call volume. But in early 2026, they introduced a new consumption-based tier for Apigee that meters calls per second. It is still optional, but the writing is on the wall. My guess is that within twelve months, all three hyperscalers will have some form of per-call pricing as the default for new contracts. Luna: So what should a team do if they are about to sign a contract with a per api call clause? What are the negotiation levers? Lucas: First, you need to know your own numbers. Most organisations have no idea what their current API call volume is because it's never been a separate line item. So start measuring. Use your cloud provider's cost and usage reports, or a third party tool, to get a trailing twelve-month view of API calls by service. Then use that average as a baseline for negotiation. Luna: So you want to lock in a baseline that reflects actual usage, not projected growth. Lucas: Right. And you also want to cap the annual price escalation. A lot of contracts have a 5 to 10 percent year-over-year increase baked in. You should counter with a fixed per-call price for the life of the contract, or at most a 3 percent annual increase tied to CPI. Also, negotiate for free tier — some providers will give you the first X million calls per month at no charge, but they don't always offer it unless you ask. Luna: And what about architecture-level changes? Could you reduce API call volume by, say, batching calls or co-locating services? Lucas: Absolutely. If you can batch multiple operations into a single API call, or move chatty services into the same availability zone or even the same container, you reduce the call count. Some teams are also using GraphQL to consolidate multiple REST calls into one query. That can cut your API call volume by 40 to 60 percent. But it takes engineering effort, and you have to weigh that against the cost savings. Luna: That makes me think: with AI inference calls also being metered separately now — as we covered in episode 86 — does it feel like the cloud is moving toward a pure utility model where everything is measured and billed individually? Lucas: Yes, but with a twist. A true utility model would have transparent, competitive pricing, like electricity. What we are seeing is more like a menu of a la carte items where the prices are opaque and change annually. It is unbundling, but not necessarily in the customer's favor. However, I think the market will eventually force more transparency, especially as third-party FinOps tools get better at tracking these costs. Luna: Well, on that note, I have to say these conversations are exactly what make this podcast useful for people building and running infrastructure. If you're getting value out of episodes like this one, a couple of dollars a month is genuinely what keeps these going — buy me a coffee dot com slash fexingo. It's a small way to keep the show ad-free and focused on the stuff that matters. Lucas: Yeah, listener support makes a real difference. Every contribution helps us stay independent and dig into these specific contract terms that might otherwise fly under the radar. So thank you to anyone who has chipped in. Luna: Alright, back to the topic. One thing I want to push on is whether there is any precedent for these API call charges being reversed or negotiated away entirely. Lucas: There is some precedent. Very large customers — those spending, say, over 50 million dollars annually — have managed to get API call charges waived entirely by bundling them into a higher committed spend elsewhere. But for mid-market companies, it is harder. I have seen cases where the provider agreed to a fixed per-call price with no annual escalation, but that was after the customer showed they were willing to walk and had a multi-cloud strategy in place. Luna: So having a credible alternative is the real leverage. Lucas: It always is. If you are all-in on one provider and have no plans to move, the provider knows that. So if you are negotiating a contract renewal and you see a per api call clause, your best bet is to benchmark your usage, present it as data, and ask for a flat rate. At the very least, get a cap on how much the per-call price can increase each year. And then start thinking about whether you can reduce your call volume architecturally over the life of the contract. Luna: That feels like a good actionable takeaway. Measure your baseline, negotiate a flat or capped rate, and then optimize your architecture over time. Lucas: Exactly. And keep an eye on the fine print in your next renewal — because this clause is spreading fast. I expect that within two years, per api call charges will be as common as data egress fees are today. Luna: Well, that is a sobering thought. Thanks for breaking this down, Lucas. Lucas: Always a pleasure. That's all for this episode of The Cloud Business Podcast — we'll be back next time with another deep dive into the fine print of your cloud bill.