A complete guide to understanding AWS Lambda pricing

Let’s be honest, staring at an AWS bill can feel like trying to decipher an ancient scroll. One month your Lambda costs are pocket change, the next they’ve skyrocketed, and you’re left scratching your head. You’re not the only one.

The whole "pay-as-you-go" promise sounds great in theory, but the reality of Lambda pricing is packed with variables that can really sneak up on you. This guide is here to translate that scroll. We’re going to pull back the curtain on how AWS calculates your bill, point out the hidden costs that are easy to miss, and give you some real strategies to get your serverless spending under control.

What is AWS Lambda?

At its core, AWS Lambda is a "serverless" compute service. Now, that’s a bit of a misnomer, because there are definitely servers involved. The "serverless" part just means you don’t have to manage them. You just upload your code, and AWS takes care of all the behind-the-scenes work to run and scale it for you.

Here’s a simple way to think about it: imagine you want to make a sandwich. You could rent a whole restaurant kitchen 24/7, paying for it even when you’re sleeping. Or, with Lambda, you can just rent a single toaster for the exact two minutes it takes to toast your bread. You only pay for what you use, down to the millisecond.

This is why it’s so great for event-driven tasks, things like resizing an image the moment it’s uploaded or processing an API call, because you’re not paying for a server to sit idle waiting for something to happen.

The core components of Lambda pricing

Okay, let’s get into the nuts and bolts. Lambda’s pricing model really boils down to two main things: how many times your code runs (requests) and how long it runs for (duration). Let’s look at each one.

Number of requests

Every single time your Lambda function gets triggered to do its job, that’s a "request." It doesn’t matter if it’s an S3 bucket event or a direct API call, each execution counts as one.

The good news? AWS starts you off with a pretty generous freebie. The AWS Free Tier gives you 1 million free requests every single month. For a lot of small projects or for just trying things out, that’s more than enough.

After you cross that million-request mark, you start paying. The standard rate is $0.20 per 1 million requests. It sounds like pennies, and it is, but if you have an app that gets a lot of traffic, those pennies can add up faster than you’d think. For the latest numbers, it’s always best to check the official AWS Lambda pricing page.

Compute duration and memory (GB-seconds)

This is where things get a little more interesting. On top of paying for each request, you also pay for the time your code is actually running, measured down to the millisecond. But here’s the twist: the price you pay for that time depends on how much memory (RAM) you give your function.

When you set up a Lambda, you pick a memory size anywhere from 128 MB to 10,240 MB. Giving it more memory also gives it more CPU power, which can make it run faster, but it also costs more per millisecond. This blend of memory and time is what AWS calls "GB-seconds."

Like with requests, there’s a free tier that includes 400,000 GB-seconds of compute time per month. Once you burn through that, you pay for what you use. To give you a feel for it, here’s how the price per millisecond shifts as you add more memory:

Beyond the basics: Other factors that affect your bill

So you’ve calculated your requests and your GB-seconds, but your bill is still higher than you expected. What gives? It’s probably a few of these "hidden" costs that aren’t always obvious at first glance. Let’s dig into the common culprits.

CPU architecture choices

You get a choice of processor for your functions: the standard x86 or AWS’s own Arm-based Graviton2 chips. For a lot of workloads, just flipping the switch to Graviton2 can get you a better price-to-performance ratio, sometimes saving you a decent chunk of change without having to touch your code.

Provisioned concurrency

If your function absolutely has to respond instantly with no "cold start" delay, you can pay for Provisioned Concurrency. This basically keeps a certain number of your functions "warm" and ready to go at a moment’s notice. It’s a great feature for performance-critical apps, but be aware that it has its own price tag and isn’t covered by the free tier.

Temporary file storage costs

Every Lambda function comes with a little bit of free temporary storage space (512 MB, to be exact) in a "/tmp" directory. If your code needs more room to juggle files or temporary data, you can configure more, but you’ll be charged for any storage you add beyond that initial free amount.

Data transfer: A hidden cost

This is the classic "gotcha" that trips up so many people. Moving data around isn’t free. If your Lambda function pulls data in from the internet, or sends data out to the internet or even to another AWS region, you’ll get billed for it at the standard data transfer rates. For apps that move a lot of data, these charges can be a real surprise.

Other AWS services

Remember, your Lambda functions rarely live on an island. They’re usually talking to other AWS services like API Gateway, S3, or DynamoDB. Every one of those services has its own bill. It’s easy to focus on the Lambda cost and forget that all the services it connects to are also adding to your monthly total.

This video offers a concise breakdown of how AWS Lambda pricing is calculated.

Lambda pricing vs. EC2: When does serverless stop being cheaper?

This is the big question, right? When does it make sense to use Lambda, and when is a good old-fashioned EC2 virtual server actually the cheaper option?

it depends on your traffic

Reddit

For spiky, unpredictable workloads with lots of downtime, Lambda is almost always the winner. You’re not paying for a server to sit idle. But if you have a steady, high-volume workload that keeps a server busy 24/7, the fixed hourly rate of an EC2 instance will eventually be more cost-effective than paying for millions of individual Lambda runs.

But don’t forget the hidden costs of running an EC2 instance. With a virtual server, your team is on the hook for patching the OS, managing security, and setting up scaling. Lambda handles all of that for you, which frees up your developers to do what they do best: build stuff.

The business impact of unpredictable pricing

The tricky thing about Lambda’s pricing isn’t just a technical headache; it’s a business one. How do you set a budget when your core operational costs can swing wildly from one month to the next? A successful marketing campaign or, worse, a malicious bot attack could send your bill through the roof without warning.

This kind of financial uncertainty isn’t just a cloud infrastructure problem. We’re seeing it pop up with a lot of new AI tools, especially in customer support. Many of them charge you per ticket resolved or per AI conversation. In a way, this model punishes you for growing. The more customers you help, the higher your AI bill climbs, making it impossible to forecast your expenses.

How eesel AI delivers predictable pricing for powerful AI

We think that’s a backward way of doing things. At eesel AI, we believe your tools should support your growth, not penalize you for it. That’s why we built our platform around predictable, transparent pricing from day one.

We saw the anxiety that usage-based billing causes and decided to offer a clear alternative. We have simple, flat-rate pricing plans that don’t charge you per resolution. That means you can handle a huge influx of customer questions without worrying about a surprise bill at the end of the month. You can scale your support and your business with confidence.

eesel AI offers predictable, flat-rate pricing plans to avoid surprise costs associated with usage-based billing models like Lambda pricing.

We even built a simulation mode into eesel AI that lets you test out the AI on your past support tickets. You can see a solid estimate of your resolution rates and ROI before you even commit, taking another layer of guesswork out of the equation.

Take control of your Lambda pricing

AWS Lambda is a fantastic tool, but its pricing model has a learning curve. Once you get a solid grip on requests, GB-seconds, and all those sneaky extra costs, you can use it powerfully without breaking the bank.

At the end of the day, whether you’re choosing a cloud service or an AI tool, the lesson is the same: predictability matters. Opting for partners and platforms with clear, transparent pricing models isn’t just about saving money; it’s about being able to grow your business without constant financial anxiety. It’s why we built eesel AI the way we did, to give you one less thing to worry about.