Prime Intellect closed a $130 million Series A on July 8 at a $1 billion valuation, led by Radical Ventures with participation from Nvidia Ventures, Intel Capital, Dell Technologies Capital and Iconiq, plus a notable angel roster including Perplexity CEO Aravind Srinivas, Box CEO Aaron Levie, Harvey CEO Winston Weinberg and other operator-investors.
Founded in 2024, Prime Intellect's pitch is that enterprises shouldn't have to depend entirely on a frontier lab's API to build sophisticated AI agents -- it provides the compute, a reinforcement-learning framework, and evaluation tooling needed to train and deploy custom agents in-house. That's a direct response to a growing concern among large enterprises: vendor lock-in and cost unpredictability from building mission-critical workflows on top of OpenAI, Anthropic or Google's constantly shifting pricing and model versions.
The company already counts Ramp and Zapier among its customers and claims roughly $100 million in annualized revenue, a fast ramp for a two-year-old infrastructure startup and evidence that the "build your own agents" pitch is resonating with technically sophisticated buyers rather than staying theoretical.
Prime Intellect sits in a crowded but still-forming category alongside Together AI and Fireworks AI, both of which offer infrastructure for running and fine-tuning open models, though Prime Intellect's specific focus on reinforcement-learning tooling for agent-building is a narrower and more technical wedge. The $1 billion valuation on $100 million of claimed revenue is a roughly 10x multiple -- rich, but not out of line with other infrastructure plays this year given the growth rate implied by a Series A this size.
For founders building AI products, Prime Intellect's traction is a signal that enterprise buyers increasingly want optionality and control over their agent stack rather than full dependency on a single lab -- a trend that should worry any AI-application startup whose entire moat is a thin wrapper around one frontier model's API. For infrastructure-focused GPs, the deep bench of strategic investors -- Nvidia, Intel and Dell all writing checks -- signals hardware vendors see agent-training infrastructure as a category worth being adjacent to, not just compute customers of.