Prime Intellect raises $130 million, hits $1 billion valuation for AI tools

Prime Intellect raised a $130 million Series A at a $1 billion valuation, drawing fresh backing from Radical Ventures, Nvidia Ventures, Intel Capital, Dell Technologies Capital and Iconiq. The startup, founded in 2024, said the new financing lifts total funding past $150 million and underscores how quickly capital is moving toward the infrastructure layer of enterprise AI.

The company is pitching a full stack built around compute access, large-scale reinforcement learning, environments, sandboxes, evaluations and deployment tools. Prime Intellect says more than 6,000 customers are already using the stack, and that annualized revenue run rate reached $100 million in under a year. Its hosted tools are being used by customers including Ramp, Zapier and Flapping Airplanes, while the company’s modular setup is designed to let buyers pick only the pieces they need instead of assembling an internal AI lab from scratch.

Ramp has become one of the clearest examples of what Prime Intellect is selling. Karim Atiyeh, Ramp’s co-founder and co-CEO, said the company’s agent beat frontier models on accuracy while running faster and at a fraction of the cost. Prime Intellect said Ramp trained a 35B model on its Lab product that beat Opus at spreadsheet search, running 27% faster and far cheaper than Haiku. That is the kind of result enterprise buyers are chasing: less dependence on a frontier lab’s API, more control over the optimization loop, and lower cost for a narrow business task that can be measured in spreadsheet searches and accuracy scores.

Intel Capital framed the bet around reinforcement learning, saying it is becoming more important as training on existing data nears saturation. The firm said the RL market was worth about $2.8 billion in 2022 and is projected to reach $88.7 billion by 2032. Intel also said Prime Intellect’s architecture can aggregate idle data-center compute globally and run RL training across fragmented GPU supply, a technical claim that fits the broader push to make enterprise AI less dependent on a few dominant model makers.

Prime Intellect describes itself as an open stack for self-improving agents, and the new round suggests investors see demand rising for the compute, orchestration and evaluation layers underneath them. The larger test is whether companies want to build and control those agents fast enough to justify the spend.