Registry of Open Data on AWS review: the best AI datasets at cloud scale

Answer in brief: The Registry of Open Data on AWS is the place to find huge datasets you want to process where they are stored. It lists Common Crawl (over 300 billion web pages), Sentinel-2 satellite imagery, The Cancer Genome Atlas, the Human Sleep Project and much more, from partners such as NASA, NOAA, NIH, EPA, AI2 and Data for Good at Meta. Browsing is free and compute is billed by AWS. Unless a dataset says otherwise, AWS doesn't provide or maintain it, and each one carries its own third-party license.

Best for: training on terabyte-to-petabyte scientific, geospatial, genomics and web data without moving it.

Independent research from the official registry, checked 2026-11-09. Not hands-on testing.

What it does for AI datasets

The registry helps people find and share datasets that are available through AWS resources. Each entry has tags, a description, a details page and published usage examples. Many have tutorials with SageMaker Studio Lab notebooks. Anyone can add a dataset or usage example through the registry's GitHub repository.

The datasets worth grabbing here, by task

• LLM pretraining: Common Crawl; its usage examples include the LAION-5B paper.
• Earth observation and vision: Sentinel-2 imagery, with a SageMaker Studio Lab deforestation-detection tutorial, and the CCRS MODIS albedo series over Canada.
• Medical AI: The Cancer Genome Atlas (tumor and normal tissue from 11,000 patients across 33 cancer types) and the Human Sleep Project, whose sleep recordings started with about 15K patients at Massachusetts General Hospital.
• Science: Folding@home COVID-19 simulation datasets.

Who it's built for

ML teams already on AWS, scientific computing groups, and anyone whose dataset is too big to download comfortably.

Where it shines

• Scale: data that would be hard to move anywhere else.
• Real usage examples and papers listed for each dataset (37 for the Human Sleep Project, 36 for Common Crawl).
• Reputable partners such as NASA, NOAA, NIH STRIDES and AI2.

Honest drawbacks

• AWS doesn't provide or maintain most datasets, so quality and upkeep depend on each provider.
• Licenses are third-party and vary, so check each one.
• It is built around AWS, so processing costs money.

Pricing and licenses

The registry is free to browse. What you pay depends on the compute and services you use. Each dataset's license is set by its provider.

Best alternatives

• BigQuery public datasets: query in SQL with 1TB/month free.
• Azure Open Datasets: curated enrichment data.
• Common Crawl: the registry's flagship web corpus.

FAQ

Are AWS open datasets free?

The registry is free to browse. You pay for the AWS compute you use to process the data.

Does AWS maintain these datasets?

Unless a dataset's documentation says otherwise, no. Third parties provide and maintain them under their own licenses.

Is Common Crawl on AWS?

Yes. Common Crawl is listed in the registry and stored on AWS Public Data Sets.

Sources

• Registry of Open Data on AWS
• Common Crawl on AWS
• Sentinel-2 on AWS

← Back to the ranked list of the best AI datasets

Named on this page as a provider? Join the network for a linked profile, Visit buttons and featured placement.