Databricks data marketplace: alternatives for licensing business data

What is the Databricks data marketplace, and who buys there?#

The Databricks data marketplace (officially Databricks Marketplace) is an open exchange inside the Databricks platform where providers publish datasets, AI models, notebooks, apps and MCP servers that Databricks customers find and request from their own workspace. Its buyers are, first of all, teams that already run analytics or AI work on Databricks.

Listings can be public, visible to everyone browsing the marketplace, or offered through a private exchange that only invited member consumers can see. Consumers on a Unity Catalog-enabled workspace receive an approved product as a catalog in their workspace in near real time. Consumers on other platforms can also take part, but only for the data set product type, and they still need a Databricks workspace (a free trial works) to request it; they then read it with a credential file.

For a business that owns operational records, such as ten years of Zendesk tickets, ServiceTitan job histories or NetSuite transactions, that audience shapes the decision. Databricks Marketplace is a distribution channel to data teams on one platform. It is not a broker that finds an AI developer for your records, reviews your rights or writes the license.

What does a supplier need before listing on Databricks Marketplace?#

You need a Databricks account on the Premium tier or above and at least one workspace enabled for Unity Catalog; Databricks notes you can create one workspace just to manage listings. That means your data has to live in, or be copied into, Databricks-governed storage before anyone can receive it.

The route then splits. Public listings require an application to the Databricks Data Partner Program, after which the Provider console appears in your Unity Catalog-enabled workspaces. Private-exchange-only providers can use a self-service signup in the Provider console, which must be done by a Databricks account admin. Either way, a user needs the Marketplace admin role to manage listings.

Every listing also needs a terms of service URL and a privacy policy URL, both publicly accessible without a login. Databricks' provider policies say listings must describe the license terms and any restrictions. In practice, you need a written data license before you need a listing.

• 1Platform: Premium-tier Databricks account plus a Unity Catalog-enabled workspace.
• 2Approval: Data Partner Program application for public listings; account-admin self-signup for private exchanges only.
• 3Roles: Marketplace admin to manage listings; CREATE RECIPIENT and USE RECIPIENT privileges to fulfill requests.
• 4Documents: public terms of service and privacy policy pages linked from the listing.
• 5Data: a Unity Catalog share containing exactly the tables or assets you intend to release.

How do instant-access and by-request listings differ?#

An instant-access listing gives consumers the data after they accept terms, with no provider approval; a by-request listing makes you approve each consumer and handle the business agreement outside Databricks. You choose the access type when you create the listing, and Databricks states it cannot be changed once the listing is saved as a draft or published.

For licensing business records, by-request is the realistic setting. It lets you see who is asking (the request carries the consumer's name, company and intended use), keep a request pending while you negotiate, and only then attach a share. Instant access suits public or sample data that you would be comfortable giving to any Databricks customer who clicks through.

What does OpenSharing (the next evolution of Delta Sharing) solve, and what does it leave to you?#

Databricks announced OpenSharing on June 10, 2026 as the next evolution of Delta Sharing. It solves secure access to shared tables: the recipient reads what you share without you shipping files, and you can revoke or rotate that access. It does not decide whether you may share the data, remove personal information, set a price or limit what the recipient does after reading it.

The open protocol is now hosted by the Linux Foundation, and Databricks OpenSharing is its enterprise implementation, built on Unity Catalog. Some Databricks pages and tools still use the Delta Sharing name. Between Unity Catalog workspaces, Databricks-to-Databricks sharing needs no credential file. For recipients without Unity Catalog, the provider issues a credential file holding a bearer token, which the recipient can download only once; the provider can rotate the token and send a new file.

The gap that matters for AI licensing is what happens after a read. A training pipeline usually materializes shared tables into its own storage, and revoking a share stops new reads but does not reach those copies. That boundary is covered in detail in the zero-copy warehouse data sharing guide linked below; the short version is that your license, not the sharing protocol, has to name permitted copies, retention and deletion at the end of the term.

Who handles rights, personal data and pricing on Databricks Marketplace?#

The provider does. Databricks' provider policies require that you hold all rights needed to sell or share each product, that you are legally authorized to share any personal data it contains, and that a product never includes sensitive personal data. Listings themselves may not include personal data.

Databricks positions itself as host and introducer. Its guide for legal departments says it hosts each provider's listing and facilitates the introduction, and that for by-request listings the provider and consumer negotiate terms outside the marketplace. The product documentation adds that no commercial transactions are handled directly on Databricks Marketplace.

On money, A Databricks marketplace FAQ blog post says there is no fee for providers to list and that Databricks does not take a cut of what providers receive from paid listings, because those transactions happen directly between provider and consumer. Any other commercial term, and any change to these, is set by Databricks' current provider terms, including the Data Partner Program Agreement that the marketplace policy forms part of. Databricks has said it is adding transaction capabilities, so read that agreement rather than relying on a blog summary.

Keep four questions separate: access (can a person technically read the table), privacy permission (may the records be disclosed for this purpose), ownership (who holds the data and any intellectual property in it) and licensing rights (what your contracts let you grant a third party). De-identifying records helps with privacy permission but does not answer the licensing question. The full framework is in the guide to data privacy and the right to license business records.

Databricks Marketplace vs managed licensing: side-by-side#

The two paths differ less in technology than in who does the work around the data. Use the matrix below to see which tasks stay with you on a Databricks listing and which a managed licensing service takes on.

Which alternative fits which kind of supplier?#

Pick by where your likely buyer already works and how much of the deal you can run in-house. A platform marketplace widens reach among that platform's customers; a managed service or direct deal fits when the buyer is an AI developer who needs a negotiated license for sensitive records.

What should you settle before any listing or deal?#

Whichever path you choose, the same groundwork decides whether a deal is safe and worth doing. Most of it is document review, not engineering.

• 1Inventory the records: system of origin (for example Salesforce, Zendesk, Jira), years covered, record counts and the fields you would release.
• 2Read the contracts that touch the data: customer agreements, SaaS vendor terms and any data processing agreements, looking for limits on use or disclosure.
• 3Map personal data: which fields identify people, which could be sensitive, and what your privacy notices said at collection.
• 4Decide the license shape: training or evaluation use, term, exclusivity, territory, and what happens to copies at the end.
• 5Name an internal approver who signs off each release, separate from the person preparing the data.
• 6If you plan a Databricks listing, draft the public terms of service and privacy policy pages first, because the listing cannot go live without them.
• 7Confirm the current Databricks provider terms with your Databricks account team before you price anything.

How this applies if you license records through SourceX#

SourceX is the enterprise data transaction layer for AI: it runs sourcing and qualification, rights review, contracting, data preparation, delivery and payment for businesses licensing operational data to AI developers. It works with U.S., Canadian and EU companies that had 50 or more full-time employees at their peak, contractors excluded.

The first step is a short company assessment, and no data uploads are needed to see whether a company qualifies. The company approves exactly what is released, once per deal, and sees what would leave before anything is shared. De-identification is performed by SourceX directly or through its processing partner, delivery is 30 calendar days after signing unless a deal agrees otherwise, and exclusivity is set per deal rather than defaulted.

Large deliveries stay in the seller's own storage or ship on encrypted drives; SourceX does not host multi-terabyte datasets. That makes it a complement rather than a rival to lakehouse sharing for some deals: the delivery mechanism is agreed per deal, while the rights, preparation and contract work happen before any data moves.