Prime Intellect raises $130m to help firms build AI agents

Prime Intellect has raised a $130 million Series A at a $1 billion valuation, led by Radical Ventures with participation from Nvidia Ventures, Intel Capital, Dell Technologies Capital, and Iconiq. Founded in 2024, the startup has already reached an annualized revenue run rate of $100 million.

The round also included angel investors who are founders of other companies:

• Aravind Srinivas (Perplexity)
• Aaron Levie (Box)
• Winston Weinberg (Harvey)
• Jeff Wang (Cognition)
• Brendan Foody (Mercor)

Prime Intellect provides what it describes as a full-stack platform for building AI agents: compute access, a reinforcement learning framework — a training method that iteratively rewards successful task completion and penalizes errors — and evaluation tools, sold in a modular, marketplace style that lets customers select specific pieces rather than commit to an all-or-nothing system. That combination, enabled by the rise of reinforcement learning techniques, can allow companies to refine models for their own business tasks and effectively become their own AI labs. According to David Katz, a partner at Radical Ventures, the company has combined these tools in a way that lets it operate at the frontier while remaining affordable, unlike rivals that offer only parts of the stack.

Customers including Ramp, Zapier, and Flapping Airplanes use Prime Intellect’s platform to build custom agents. Ramp built an agent using Prime Intellect that outperformed frontier models on accuracy while running faster and at a fraction of the cost, according to Karim Atiyeh, Ramp’s co-founder and co-CEO.

This growth is also driven by enterprise wariness of closed frontier labs. Companies are increasingly hesitant to share proprietary data with OpenAI and Anthropic due to the risk of losing control over that data, and they are wary of depending on models that can be suddenly turned off, as happened with Anthropic’s Fable last month. Katz said that uncertainty is pushing companies to ask whether a model provider might use their own data to replace them, and to look for ways to own their enterprise intelligence rather than carry those risks. Vincent Weisser, Prime Intellect’s co-founder and CEO, said: “It shouldn’t just be a few nerds in a glass tower in San Francisco that have the capability to train AI models.” He added that it should extend to every enterprise and nation state instead.

Why it matters

The funding round highlights a growing enterprise demand to build proprietary AI agents independently of closed frontier labs like OpenAI and Anthropic, driven by concerns over data control and platform dependency.