Perceptron raises $6.5M to build decentralised AI data network
Tech.eu Cate Lawrence
Perceptron just closed a $6.5M round to build a decentralized network where regular people sell AI training data. It's already got over 700,000 nodes signed up, aiming for 5 million.
Based on reporting by Tech.eu, Cate Lawrence — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Money is flowing again into the unglamorous plumbing of AI: the data itself. Perceptron, a startup building what it calls a decentralised AI data network, announced today it closed a $6.5 million strategic round. The investor list reads like a Web3 who's-who — Sigma Capital, Selini Capital, QCP Capital, P2 Ventures, CoinDCX Ventures, and a dozen others spanning trading firms, ecosystem funds, and infrastructure players like Aethir and Walrus Foundation.
The pitch is straightforward, even if the execution is anything but. AI companies need real-world data, and right now that mostly means scraping the internet at scale. Perceptron wants to replace that with a mesh of contributors — people offering idle bandwidth, niche datasets, or specialised expertise — who can be tapped directly. Contributors keep ownership of what they submit and can monetise or withdraw it whenever they like, rather than handing it off to some third party that controls the terms.
The numbers so far are not trivial. Perceptron says it has more than 700,000 nodes onboarded. Early agent deployments pulled in over 200,000 users through Telegram and Discord communities, and that figure has since climbed to more than 300,000 daily active users across a network that now counts over 807,000 nodes. Peter Anthony, the UK-based co-founder and CEO, frames this growth as proof that the organic model works before the company adds paid commissioning on top of it.
That's where the new funding actually goes to work. Perceptron is launching a data-questing platform that lets AI companies commission specific datasets directly from the contributor base, instead of waiting for data to arrive organically. It's a shift from passive collection to active demand-matching, and the company says it shortens the gap between an AI firm asking for a dataset and actually getting one down to days. Nathan Gurr of P2 Ventures pointed to Perceptron's ability to reach specialists — doctors, lawyers, native speakers — as the real differentiator, calling it a path toward a scalable intelligence layer for AI.
The ambition stretches well past where the network sits today. Perceptron is targeting five million nodes, envisioning a single integrated system where AI companies find everything they need without any mismatch between what's wanted and what's available. Further announcements are expected next quarter, alongside expanded contributor tooling and rewards infrastructure funded by this round.
My take — AI-written commentary, not fact-checked reporting
Decentralised data networks keep promising to cut out the scraping middleman, and the pitch always sounds cleaner than the reality of actually verifying quality at scale. Perceptron's node counts are genuinely impressive, but signing up isn't the same as delivering datasets AI companies will trust for training. The real test starts now, with the questing platform — commissioned data is a much harder problem than organic contribution, and that's where plenty of DeAI projects have stumbled before.
Read more about this at: Tech.eu