What Datadog’s 2022 Serverless Report Reveals About Lambda Usage and Costs

Overview

Datadog’s 2022 Serverless State report updates the 2021 findings and quantifies the adoption of Function‑as‑a‑Service (FaaS) platforms, especially AWS Lambda, across global enterprises. The report measures usage depth by invocation frequency and execution duration, which directly drive billing.

Insight 1 – Lambda Call Frequency and Runtime

Enterprises that have been using AWS Lambda since 2019 invoke functions 3.5 × more often than two years earlier. The average Lambda runtime per company is reported as 900 hours per day. Assuming a 1 GB memory allocation, the daily cost on Alibaba Cloud Function Compute (0.00003167 CNY per GB‑second) is:

Cost per second = 1 GB × 0.00003167 CNY = 0.00003167 CNY
Daily seconds = 900 h × 3600 s = 3,240,000 s
Daily cost = 3,240,000 s × 0.00003167 CNY ≈ 102.6 CNY
Annual cost ≈ 3.7 万 CNY

A typical AI‑modeling function with 1 million daily invocations adds roughly 13.3 CNY per day.

Insight 2 – Execution Time Shrinks

The median Lambda execution time fell to 60 ms, half of the value a year earlier. Because AWS charges per millisecond, halving the duration reduces compute cost by ~50 % for comparable workloads. The report attributes the reduction to best‑practice optimizations such as provisioned concurrency, smaller package sizes, and language‑runtime tuning, enabling both short‑lived tasks and compute‑intensive jobs (e.g., video transcoding, AI inference).

Insight 3 – Growth of Azure Functions and Google Cloud Functions

Azure Functions usage grew from 20 % to 36 % of Azure customers, while 25 % of Google Cloud customers now run Cloud Functions. This diversification indicates a maturing multi‑vendor serverless ecosystem.

Insight 4 – Vercel Serverless Scale

Vercel’s monthly serverless invocations increased from 262 million to 7.4 billion (≈28×) over the past year, demonstrating that serverless can sustain high‑volume personal‑project and small‑business workloads.

Insight 5 – Lambda@Edge Adoption

One‑quarter of Amazon CloudFront customers use Lambda@Edge. Among those, 67 % of functions complete in under 20 ms, highlighting the importance of sub‑20 ms latency for edge‑centric use cases such as IoT video analytics and real‑time monitoring.

Insight 6 – Reserved Function Instances Under‑utilization

57 % of customers with reserved function instances consume less than 80 % of the allocated capacity; over 30 % consume less than 40 %. While reserved instances mitigate cold‑start latency, the low utilization reveals a tension between pay‑as‑you‑go economics and pre‑provisioned capacity.

Insight 7 – Open‑Source Serverless Frameworks Dominate Deployments

More than 80 % of surveyed organizations use the Serverless Framework for deployment. AWS‑native tools (CloudFormation, CDK, SAM) together account for ~50 % of usage. In China, projects such as Serverless Devs and Midway are gaining traction, providing IaC‑style deployment, monitoring, and lifecycle management.

Insight 8 – Python Becomes the Dominant Runtime

Python accounts for 58 % of Lambda runtimes, surpassing Node.js (31 %). In large‑scale enterprises, Python usage is four times higher than Node.js, driven by data‑intensive and AI workloads. Runtime version distribution (descending) is: Python 3.x, Node.js 12, Node.js 10, Python 2.7, Java 8, Go 1.x, .NET Core 2.1, .NET Core 3.1.

Conclusion

Serverless adoption is accelerating globally: invocation frequency and total runtime are rising, median execution times are halving, and language support is expanding. Multi‑vendor offerings (Azure, Google) are gaining market share, and edge‑focused services (Lambda@Edge) are proving critical for low‑latency scenarios. Open‑source deployment frameworks dominate the tooling landscape, while Python solidifies its position as the preferred runtime for enterprise‑grade workloads.