Redpine raises €6.8M seed

AI infrastructure is starting to look less like bigger models and more like better pipes. Redpine, a Stockholm startup founded in 2024, just raised €6.8 million in seed funding to build one of those pipes — an API that gives AI companies and agents access to licensed, non-public datasets instead of forcing them to rely on the open web. The round was led by NordicNinja, with Luminar Ventures and node.vc joining in, plus angels tied to OpenAI, Perplexity, Spotify, Sana, Validio, and Silo AI. (tech.eu) ### What is Redpine actually selling? Basically, a data access layer. Redpine says developers can query licensed, high-quality, multimodal datasets from a single entry point, and use them inside AI products and autonomous agents. The company frames itself as “the knowledge layer for agentic AI” — which sounds buzz(tech.eu)c website waiting to be scraped. (redpine.ai) ### Why is non-public data the whole point? Because public internet data is the easy part, not the valuable part. Redpine’s founders keep coming back to the same stat — only around 1% to 2% of the world’s data is actually available to train or feed AI systems, while the rest lives in archives, databases, journals, enterprise systems, and other closed stores. If you w(redpine.ai)nce, finance, legal research, or enterprise workflows, public web search stops being enough very quickly. (nordicninja.com) ### Why does this show up now? Because the market is moving from models that answer questions to agents that have to complete tasks. A chatbot can bluff its way through thin information. An agent can’t — it has to fetch the right record, use current data, and stay inside legal (nordicninja.com)trieval layer on top of scraped content. That is the real timing behind the round. (thenextweb.com) ### What makes the model different from scraping? The clean version is licensing and usage-based access. Redpine says rights holders get paid, customers get a compliant way to use premium datasets, and AI builders avoid the grey zone that has defined a lot of model training and retrieval so far. One investor described it(thenextweb.com)because the pitch is licensed access instead of extraction without compensation. (thenextweb.com) ### Who backed it? NordicNinja led the round, and partner Marek Kiisa is joining the board. Luminar Ventures and node.vc also participated. The angel list is part of the signal here — people tied to OpenAI, Perplexity, Spotify, Sana, Validio, and AMD’s Silo AI showed up, which suggests investors see data rights and data access as a real infrastructure layer, not a side feature. (nordicninja.com) ### What will the money do? Redpine says the seed funding will go into international expansion, more exclusive data partnerships, and further development of the platform. Some coverage also pegs total funding at about €9 million after this round. So this is not just a product p(nordicninja.com)e datasets before somebody else does. (tech.eu) ### Where’s the catch? The catch is that data businesses are two-sided marketplaces, and those are hard. Redpine has to convince rights holders to join, developers to integrate, and both sides to trust the economics. But if it works, the company becomes useful in the boring, durable way infrastructure companies do(tech.eu)th anything. (redpine.ai) ### Bottom line? Redpine is not trying to build a better model. It is trying to control one of the scarcest inputs around those models — lawful access to better data. If AI agents become real software workers, that could end up being the more important layer.