DataVendor opened its public beta as a marketplace where teams license private data, code, and task data to AI companies that need training and evaluation supply. Sellers list private codebases, raw datasets, tasksets, and reinforcement learning environments they already hold. Buyers, mostly AI labs, purchase a listing, try a sample, make an offer, or post a request when the catalog does not have what they need. The catalog is for supply models have not already seen. Public repositories cannot be estimated, because models are often trained on public code already. Before paying, a buyer sees metadata, a redacted file tree, quality-check results, and taskset previews, never the raw files. Sellers keep ownership and grant a commercial license, and they set the price. DataVendor keeps 20 percent of vendor earnings, charged only when those earnings are withdrawn to a bank. Converting earnings to HUD credits is free. Vendor teams can list as soon as setup is finished. Buyer organizations join a waitlist and are invited. DataVendor is the marketplace from HUD, the San Francisco company in Y Combinator's Winter 2025 batch, founded by Jay Ram, Lorenss Martinsons, and Parth Patel. Purchased tasksets and environments are the same objects the HUD platform runs evaluations against, so a sale can go straight into training and eval work. DataVendor says current demand is strongest for enterprise software codebases and for cybersecurity, legal, and finance data. The beta gives teams that already hold that private supply a direct way to price it and license it to labs.