Hub.xyz is building the API for real-world training data, turning a global crowd of contributors into a programmable pipeline for AI labs that have exhausted the public web. The company operates a SuperNetwork of more than 500,000 contributors across 150 countries and 100 languages, collecting original multimodal data including audio, image, and video that simply does not exist in Common Crawl, YouTube, or any other open corpus.
The core product is an API that promises to go from request to delivered dataset in under two minutes for many tasks. Hub owns the full pipeline end to end, from contributor collection and automated processing to human-in-the-loop quality assurance and final delivery. Custom long-tail projects are scoped by modality, volume, geographic coverage, and complexity, with quotes returned within hours rather than weeks. Clients include leading AI labs and Fortune 500 technology firms that need hard-to-source data for post-training, evaluation, and physical AI.
Founded in 2024 by Tim Sprecher and Armin Kiani, Hub.xyz is headquartered in Palo Alto and is part of Y Combinator Spring 2026 P26 batch. It has raised approximately 1.7 million dollars in seed funding led by SwissBorg, with YC participating. The company sits at the intersection of data-labeling marketplaces such as Scale and Surge and decentralized contributor networks, positioning itself as the data layer for the next wave of physical and embodied AI systems.