product · marketplace
subscribe to data, not files
Petabytes of analysis-ready scientific data, instantly queryable. Subscribe and the dataset appears in your organization — versioned, updated, and ready to read.
data from trusted providers
the bottleneck
File-based distribution hasn't kept pace
Weather, climate, and Earth-system data are among the most valuable inputs to modern analytics, modeling, and AI workflows. But most of it is still published the way it was thirty years ago: as massive collections of GRIB and NetCDF files.
To make those files usable, every team downloads, ingests, transforms, rechunks, versions, and maintains a custom pipeline — often for exactly the same datasets that dozens of other organizations are processing in parallel. For a forecast that updates four times a day, those costs compound with every cycle.
-
Duplicated work — the same ingestion problem solved independently across an entire industry.
-
Latency — hours between a forecast being published and it being usable in your stack.
-
Ongoing cost — egress, storage, and an engineer whose job is keeping the pipeline alive.
for data users
Analysis-ready, cloud-optimized, already ingested
Marketplace datasets are delivered as Icechunk data cubes: all timesteps, variables, and members in one logical object, chunked and compressed for streaming directly out of object storage. Instead of a directory of files, you get a dataset you can slice.
No ingestion pipeline
Subscribed datasets show up as read-only repositories in your Arraylake organization. Open them with Xarray, serve them with Compute, or query them from an agent — the same way you use your own data.
Incremental updates
High-velocity datasets like operational forecasts update multiple times a day. Each cycle arrives as a new commit, so you get the latest data without re-downloading anything.
Nothing is copied
Subscriptions read chunks directly from the provider’s object storage. There is no duplicated archive to pay for and nothing to keep in sync.
Free and paid, side by side
Free datasets can be subscribed to instantly by anyone with an Arraylake account. Paid datasets are granted by the provider once terms are agreed.
for data providers
Share or monetize your data without building a portal
Publish free or open data to reach the community, or offer paid subscriptions on your own terms. Either way, distribution, access control, and delivery are handled by the platform.
List from your own repo
Your data stays in your Icechunk repository and your bucket. A listing is a view onto it — name, description, README, license, thumbnail, and pricing.
Control what ships
Paid listings let you choose exactly which groups and variables subscribers receive, so you can build several products from one source repository.
Read-only by construction
Subscribers can only read, never list, write, or delete, and only within the paths your listing defines. Icechunk’s cryptographically random keys mean nothing outside the listing is even discoverable.
Publish on your schedule
Draft a listing privately, announce it as coming soon to gauge interest, then publish when the data is ready.