The State of Serverless | Datadog
The State of Serverless

/ / / / / / / / /

Updated August 2023

This research builds on the previous edition of this article, which was published in June 2022. Click here to download the graphs for each fact. And click here to download our deck of the report.

Serverless has become a mainstay of modern computing. Today, organizations are making use of a growing set of serverless offerings to build and manage applications in new, exciting ways. Teams continue to expand beyond traditional function-as-a-service (FaaS) solutions with containerized functions and fully managed container-based applications. Major cloud providers such as AWS, Google Cloud, and Azure, as well as emergent platforms such as Vercel and Cloudflare, now offer distinct serverless compute services designed to address unique developer workloads. Serverless is supplanting traditional infrastructure in some places while more tightly integrating with it in many others.

For this report, we looked at usage data from more than 20,000 of our customers who are running in all major clouds and monitoring their serverless workloads with our platform. Here you'll find key insights that show how these customers are using serverless technologies in real-world scenarios.

Continue reading to learn about trends within the serverless landscape.

Fact 1

Major cloud providers continue to see significant serverless adoption

Over the past year, serverless adoption for organizations running in Azure and Google Cloud grew by 6 and 7 percent, respectively, with AWS seeing a 3 percent growth rate. Over 70 percent of our AWS customers and 60 percent of Google Cloud customers currently use one or more serverless solutions, with Azure following closely at 49 percent.

Serverless adoption by major cloud provider

For this year’s report, we have included two new serverless platforms in our data: Azure Container Apps and AWS CloudFront Functions. While FaaS solutions continue to be the primary driver of overall serverless popularity, we see that cloud providers are expanding their suite of serverless tools to better meet the needs of their customers. As cloud providers improve their existing solutions and deliver newer and wider sets of offerings—such as serverless containers and serverless edge computing—more organizations are able to use these tools to deliver value to their customers.

Note: For the purpose of this fact, a serverless organization uses at least one of the following technologies:

  • AWS: AWS Lambda, AWS App Runner, ECS Fargate, EKS Fargate, AWS CloudFront Functions
  • Azure: Azure Functions, Azure Container Apps, Azure Container Instances
  • Google Cloud: Google Cloud Functions, Google App Engine-Flex, Google Cloud Run
Fact 2

Google Cloud leads in fully managed container-based serverless adoption

In 2022, we reported that organizations were extending their use of serverless technologies beyond traditional functions with managed runtimes—such as AWS Lambda, Google Cloud Functions, and Azure Functions—toward functions packaged as containers and fully managed container-based application platforms.

Those trends have continued across all three major clouds—particularly in Google Cloud, where 66 percent of all serverless organizations now use container-based serverless workloads (i.e., Cloud Run and 2nd gen Cloud Functions).

Usage of fully managed container-based serverless platforms by major cloud

While Google Cloud has led in this category since launching Cloud Run in 2019, this year AWS increased to 26 percent of serverless organizations running fully managed container workloads with containerized Lambda functions and AWS App Runner. This is up from roughly 20 percent last year. Close behind is Azure at 22 percent; albeit starting with smaller numbers, Azure’s year-over-year growth in container-based workload adoption has been notably impressive at 76 percent, driven by the release of Azure Container Apps in May 2022.

One reason why container-based serverless compute platforms have grown more popular is likely because they simplify serverless adoption and migration. For instance, organizations can upload their existing container images to cloud provider-hosted registries and seamlessly deploy those containers as microservices. Serverless container products also support a wider breadth of languages and larger total application sizes compared to serverless functions, further aiding workload migration.

“At Google Cloud, we are committed to serving our customer needs for portability and flexibility with our serverless offerings. Cloud Run and 2nd gen Cloud Functions—our container-first approach to Functions-as-a-service (FaaS)—gives customers the ability to run their workloads on any cloud platform, on prem, and any container platform on Google Cloud, making it easy to go from Cloud Functions to Cloud Run and Google Kubernetes Engine. This approach serves our customers by saving them money, reducing their need to rework their workloads between platforms.”

Rachel Tsao
Product Manager, Serverless, Google Cloud
Fact 3

Frontend development is the leading category of emerging serverless platforms

The major cloud providers, such as AWS, Google Cloud, and Azure, are not the only players in the serverless compute game. Today, modern frontend development and content delivery network (CDN) platforms such as Vercel, Netlify, Cloudflare, and Fastly also offer developers specialized serverless compute capabilities that are tightly integrated with their core platform offerings. Seven percent of all organizations that are monitoring serverless workloads in a major cloud are also running workloads using at least one of these emerging cloud platforms. Of these, 62 percent use a frontend development platform like Vercel or Netlify, and 39 percent use edge compute offerings from Cloudflare and Fastly.

Serverless compute adoption by emerging platform

Platforms like Vercel Functions and Netlify Functions enable frontend developers to easily create and run entire applications from the cloud, making full-stack projects and development accessible to even more people. Meanwhile, distributed edge platforms like Cloudflare Workers and Fastly Compute@Edge enable developers to build serverless applications closer to end users in a tightly integrated way with their existing CDN provider, meaning they can more easily deliver highly performant A/B testing, personalization, authentication, and more.

Cloudflare, Vercel, and others have evolved to embrace both frontend development and edge compute for full-stack development. For example, Cloudflare offers Pages and other tools to enable building full-stack applications in tandem with Workers. And when deploying an app in Vercel, developers can make use of their Edge Network. Overall, the evolution of CDN and frontend development platforms outside of the major cloud providers suggests organizations are embracing alternative options for providing highly performant end-user experiences.

“From a frontend perspective, developers have always understood the power of the server for rendering better end-user experiences and SEO, but historically the barrier of entry to maintaining and provisioning a globally distributed edge network has been prohibitive for everyone but the largest companies. With innovations in the frontend cloud ecosystem—such as streamlined access to the serverless primitives and frontend frameworks that increasingly embrace a server-first model, like Next with React Server Components—we’re seeing the industry shift back to what made the web great to begin with: faster, more personalized, more first-party, and more private experiences.”

Guillermo Rauch
CEO, Vercel
Fact 4

Node.js and Python remain dominant languages for AWS Lambda functions

Python and Node.js are still the most popular languages used among AWS Lambda developers. In fact, well over half of the invocations in our Lambda dataset were from functions written in Python or Node.js. In addition to being among the two earliest languages to be supported by Lambda, both languages have robust tooling for Lambda and large communities of developers. This is why, when organizations are starting out with serverless, they tend to do so with Python and Node.js.

Percent of AWS Lambda invocations by language

Per our research, Java is the third most common Lambda language, closely followed by custom runtimes and Go. Together they account for just under a quarter of all the Lambda invocations by our customers. Java’s status as the third most common Lambda language is likely due to a growing number of large enterprise organizations migrating their existing workloads and applications written in Java over to Lambda.

Additionally, custom runtimes are the fastest-growing types of functions and have seen their share of Lambda function invocations increase by more than 50 percent year over year. It is our belief that custom runtimes have risen in popularity because of the increased interest in serverless containers, which extend the capabilities of serverless by enabling developers to write functions in languages Lambda doesn’t natively support, such as PHP and Rust.

“In 7 years of running production workloads on AWS Lambda, across Node, Python, Java, and custom runtimes, I can't think of one time where our client has regretted the decision. The operational burden is dramatically lower and—after an initial learning curve—developer and DevOps productivity much higher.”

Andy Warzon
CTO, Trek10
Fact 5

AWS Lambda function cold starts in Java are two times longer than Python or Node.js

One of the main concerns of serverless developers today is overcoming the challenge of cold starts, which occur when a serverless compute platform must create a new execution environment in order to serve a request. The impact of cold starts in Lambda varies depending on which runtime a function is written in. For instance, Node.js and Python functions experience the shortest cold start durations whereas Java functions experience the longest. Java’s average cold start duration—which is nearly three times as long as Python’s—is likely due to the time it takes to load the Java Virtual Machine (JVM) and libraries Java requires to run.

Relative Lambda cold start time by language

Our research also revealed that the amount of memory allocated to Lambda functions varies by runtime. This is likely because increasing the memory allocated to a Lambda function also increases the amount of CPU allocated to it, which helps to reduce cold start durations. Runtimes that already have the shortest cold start durations—such as Python and Node.js—typically have less memory allocated to them than Java. AWS continues to develop a suite of tools that help developers combat cold starts and lessen their impact on serverless function performance, such as provisioned concurrency, proactive initialization, and Lambda SnapStart.

Percent of Lambda functions with more than 1,024 MB in allocated memory

“Datadog's 2023 The State of Serverless report unveils insights on how developers accelerate modern app development by adopting the serverless operational model. As we strive to enhance the developer experience, we continue to innovate across many areas, such as Lambda SnapStart for Java, which delivers up to 10x faster function startup performance without developers making changes to their code.”

David Nasi
Head of Product Management, AWS Lambda
Fact 6

AWS Lambda use on ARM has doubled over the past year

In 2021, AWS announced support for Lambda functions on ARM-based Graviton2 processors, promising faster execution times and up to 34 percent lower costs compared to x86-based processors. The portion of serverless organizations using Lambda that have adopted ARM has doubled in the past year, with 11 percent of organizations now using ARM in some capacity to invoke Lambda functions.

Percent of serverless orgs using Lambda on ARM

Among these organizations that have begun adopting ARM, 29 percent of their functions using ARM-eligible runtimes are now running on ARM. These adoption trends suggest that these organizations are likely getting the combined performance and cost benefits that were proposed.

But these numbers also show that there is still opportunity for even more eligible functions to be moved to ARM. One reason why we haven’t seen more AWS Lambda functions moved to ARM could be that Graviton2 is not yet available in all AWS regions (see here for the latest) and is only compatible with a subset of Node, Python, Java, .NET, Ruby, and custom runtimes. Because they are not universally available, deployment and automation tools like the Serverless Framework have not yet begun setting them as the default configuration.

Based on conversations with our customers, we expect to see adoption continue to grow as regional and runtime availability continue to increase and as ARM becomes a default configuration for serverless tooling.

“Lambda functions powered by ARM-based Graviton2 processors offer better performance at a lower cost. We expect to see them become the default choice across deployment frameworks as they become available in more regions.”

Jeremy Daly
AWS Serverless Hero and CEO, Ampt
Fact 7

Terraform is the preferred AWS Lambda deployment tool among larger organizations

Infrastructure as code (IaC) tools like the Serverless Framework and Terraform help developers overcome the challenge of manually deploying and configuring Lambda functions and other resources at scale. AWS offers CDK, SAM, and Chalice as frameworks on top of their IaC platform, CloudFormation. These tools enable teams to better define their infrastructure and manage all of their serverless resources in a programmatic way. Our observations reveal that as organizations mature and expand, the preferred IaC tool for deploying and managing their Lambda functions shifts in predictable ways.

For instance, the Serverless Framework is the most popular IaC tool for managing AWS Lambda functions among Datadog customers. It provides streamlined opinionated workflows that make it straightforward to deploy applications and serverless infrastructure together. This, along with the strong community support it has among developers, has made it a popular and inviting entry point for newcomers and serverless-focused teams.

With that said, we’ve seen that as organizations’ host counts increase, they tend to use Terraform for managing AWS Lambda. This suggests that Terraform’s flexibility, multi-cloud support, and wide adoption and use by DevOps teams are preferred by larger organizations operating within a diverse cloud infrastructure.

Lambda deployment tool adoption by host count

“Startup and serverless-only teams have the freedom to choose the purpose-built tools for building serverless apps, like Serverless Framework, SST, and others. As serverless technologies become a key piece of an organization's infrastructure alongside traditional workloads, those teams need to prioritize a solution that allows them to manage all of their workloads centrally.”

Filip Pyrek
AWS Serverless Hero and CEO,
Fact 8

65 percent of organizations have connected AWS Lambda functions to a VPC

As organizations adapt and replace parts of their existing services with Lambda, it’s important to make sure every serverless function remains integrated across their entire infrastructure. To do this, many teams connect Lambda functions directly to the virtual private clouds (VPCs) that contain their broader compute and database resources in isolated networks. In addition to providing granular network accessibility and security, VPCs are increasingly required for organizations to connect serverless workloads for larger applications to existing infrastructure services, such as Elastic Cloud Compute (EC2) or Relational Data Store (RDS) instances in standard deployments.

Over the last year, 65 percent of Datadog customers using Lambda have deployed at least five of their Lambda functions connected to a dedicated VPC in their own AWS account. Even more customers—80 percent—have at least one Lambda function connected to a dedicated VPC, while 10 percent run all of their functions with their own VPC.

Lambda functions connecting to VPCs

There are risks and tradeoffs to consider when connecting Lambda functions to your account’s VPC. By not using the default AWS-monitored and secured VPC, your organization may be taking on a greater footprint of responsibility within the shared responsibility model. Additionally, while AWS has made significant improvements to ENI connection performance in recent years, there is still an additional cold start cost to VPCs.

“Teams adopting Lambda still need to make informed decisions based on their security, performance, and functional requirements, which won’t be one-size-fits-all. Today, those commonly involve the need to connect to VPCs to integrate with existing applications, infrastructure, and compliance requirements, though VPCs are not a universal fit for all applications or cases. As applications and teams evolve, there is an opportunity to move toward more cloud-native and serverless-first solutions for networking and security, such as RDS proxies and VPC endpoints.”

Mike Stemle
Principal Architect, Arc XP, A Washington Post Company

Detect and resolve performance issues in your serverless applications with Datadog Serverless Monitoring.



For this report, we compiled usage data from thousands of companies in Datadog's customer base. But while Datadog customers cover the spectrum of company size and industry, they do share some common traits. First, they tend to be serious about software infrastructure and application performance. They also skew toward adoption of cloud platforms and services more than the general population. All the results in this article are biased on the fact that the data comes from our customer base, a large but imperfect sample of the entire global market.

Serverless definition

AWS, Azure, Google Cloud, and other cloud platforms offer a variety of compute services to help developers solve problems for their customers. Serverless has evolved as a category label and marketing term for a subset of those compute services that share at least some of a common set of principles:

  • Improve time to delivery and time to value for developers
  • Reduce operational costs
  • Abstract infrastructure configuration and management

The majority of these services also provide the opportunity for—but no guarantee of—cost savings with granular pricing models that only charge for the resources applications consume in a multi-tenant model, and scale down to zero billable consumption when resources are not in use.

The precise level of infrastructure abstraction and the granularity of billing behavior exist on a spectrum. We see five common categories that have emerged across clouds, with significant functional overlap and blurring of these lines between categories across and within clouds:

  • Containers with serverless orchestrators
  • Application platform as a service (PaaS)
  • Fully managed container applications
  • Functions as a Service (FaaS)
  • Edge functions

For the purposes of this report, we are focused primarily on the adoption of serverless functions, fully managed container applications, and edge functions. We have also considered PaaS and containers with serverless orchestrators in certain facts to provide a more complete view of the technology choices that our customer base is making to deliver value for their customers quickly, despite those services not fitting in with all possible definitions of serverless.

Fact 1

In our 2022 The State of Serverless report, we found that over half of all organizations running in each major cloud (AWS, Google Cloud, and Azure) had adopted serverless. We consider an organization to be a customer of a given cloud if they run at least five hosts per month in that particular cloud. They are also considered a customer of a cloud if they run at least five functions or one serverless application per month in that cloud.

In 2023, we expanded our definition of serverless workloads and the set of metrics we used to detect host-based usage within each cloud. This resulted in a higher number of total detected customers of each cloud but a lower detected percentage of serverless organizations within Azure and AWS in 2022. However, each cloud continued to show steady growth year over year in serverless adoption.

Note: In order to determine what percentage of organizations have adopted serverless in each cloud, we included customers monitoring the following technologies:

  • AWS: AWS Lambda, AWS App Runner, ECS Fargate, EKS Fargate, AWS CloudFront Functions
  • Azure: Azure Functions, Azure Container Apps, Azure Container Instances
  • Google Cloud: Google Cloud Functions, Google App Engine-Flex, Google Cloud Run

Some organizations meet both sets of criteria, whereas others meet only one. For the purposes of this fact, organizations that met the latter set of criteria as of May 2023 are considered to have adopted serverless compute, and we make year-over-year comparisons with May 2022.

Fact 2

We share the same definition of serverless organizations from Fact 1.

Fact 3

We share the same definition of serverless organizations from Fact 1. To identify customers using emerging cloud platforms, we include organizations submitting metrics and/or logs from Cloudflare Workers, Fastly Compute@Edge, Vercel Functions, and Netlify Functions.

Fact 4

We broke down invocations from monitored Lambda functions in May 2023 using the runtime metadata associated with invocation metrics. We combined the runtime versions for each language to aggregate invocations at the language level.

Fact 5

To determine relative cold start times, we examined the Datadog enhanced init_duration metric in May 2023. We combined the runtime versions for each language to aggregate initialization duration times at the language level. The relative cold start times are based on the median cold start duration in each language. Percent of Lambda functions with more than 1,024 MB in allocated memory is based on metadata from invoked Lambda functions in May 2023.

Fact 6

Based on metadata from invoked Lambda functions in May 2023. We define ARM-eligible runtimes as any runtimes compatible with ARM as of May 2023.

Fact 7

Based on metadata from invoked Lambda functions in May 2023.

Fact 8

Based on a sample of monitored Lambda functions from May 2023.