Cloud-native geospatial has made big progress on how we access data in the cloud. Formats like COG, Zarr, GeoParquet, STAC, COPC, FlatGeobuf, and PMTiles let us read just what we need in place instead of downloading everything first. But, as the Cloud-Native Geospatial Guide itself puts it, storing data in the cloud does not, on its own, solve geospatial's data problems. It still leaves open a harder set of questions: where do you find that data, access it reliably, and trust that it will still be there and still be well-governed next year? For terabyte-to-petabyte-scale open datasets, the answer is too often a one-off bucket, a bespoke API, or a single vendor's walled garden.

This talk comes from Radiant Earth, which supports the open geospatial community in two complementary ways. First, Radiant Earth convenes the Cloud-Native Geospatial community — it runs the CNG Forum in-person events and development sprints, online webinars and meetings, the CNG blog and newsletter, a members' Slack, and fiscal sponsorship for community projects — helping practitioners converge on the shared formats and best practices above. Second, it builds and operates Source Cooperative, a service.

Source is a vendor-neutral, open-source "data publishing utility" for data of any kind at any scale — not a geospatial-only tool. Cloud-native geospatial is a big part of what it hosts, but so is a lot that isn't: Harvard Library Innovation Lab's Archive of Data.gov and the Smithsonian Open Access Archive (hundreds of terabytes of public-domain cultural-heritage data), among others. Read access is open to everyone and whitelisted organizations can publish data with a current focus on scientific data publishers. The result is a platform that is open to consume and curated to publish.

On the backend, Source stores data on commodity object storage, using a variety of cloud providers. Although Source is format- and domain-agnostic, it's a natural home for cloud-native geospatial data — and that intersection is the part most relevant to this audience. I'll frame it around FAIR — Findable, Accessible, Interoperable, Reusable — at scale:

• Findable: every product has a stable URL, is indexed by Source Search and Google Dataset Search, can carry a DOI, and is described in a self-describing, queryable metadata catalog (Source publishes its own catalog as a product on Source).
• Accessible: data is reachable over plain HTTP, with metadata kept separate from the data itself, so clients read only what they need without bespoke tooling.
• Interoperable & Reusable: Source aligns with cloud-native geospatial standards and guides publishers toward best practices, while leaving format and structure choices to the data product owner.

I'll briefly cover the Source data proxy, the piece that makes "accessible at scale" real, and focus on why it matters: people read data with the standard cloud tooling they already use without the need to store long-lived credentials. Behind a familiar S3-style interface, short-lived, scoped access is brokered on the fly (using OIDC and STS), so access is both easy and safe by default. And because the data proxy sits in front of S3-compatible storage, the platform isn't captive to any single vendor and access performance is good worldwide.

Finally, I'll argue that the real differentiator is governance. Source's nonprofit incentives, vendor-neutral architecture, and avoidance of lock-in are what let publishers and users trust that large-scale public data will stay accessible and durable. This is where cloud-native standards (the how) meet a sustainable home (the where, and for how long).

Attendees will leave understanding how Radiant Earth supports cloud-native geospatial as both a community of practice and as infrastructure, how to find and consume data on Source today, and why governance determines whether large open datasets survive. The takeaway: Radiant Earth supports cloud-native geospatial both as a community and as durable infrastructure and Source Cooperative is a vendor-neutral, well-governed home for open data at scale, geospatial and beyond.

This is a Technical-track session and assumes general familiarity with cloud storage and geospatial data; no prior knowledge of Source Cooperative is required.