Prime Intellect just raised $130 million in a Series A that values the two-year-old startup at $1 billion, according to TechCrunch AI. The round was led by Radical Ventures, with Nvidia Ventures, Intel Capital, Dell Technologies Capital, and Iconiq joining in. What stands out is the founder list on the cap table: Aravind Srinivas of Perplexity, Aaron Levie of Box, Winston Weinberg of Harvey, Jeff Wang of Cognition, and Brendan Foody of Mercor all wrote checks.
The pitch is simple and pointed. Prime Intellect wants to let any company train its own AI agents without leaning on frontier labs like OpenAI or Anthropic. Founded in 2024, the startup sells what it calls a “full stack” for agent development, and TechCrunch AI reports it’s already at a $100 million annualized revenue run rate.
What they actually built
Most companies can’t assemble a production AI system on their own. The pieces exist, but wiring them together takes expertise that’s rare and expensive. Prime Intellect packages that work into three parts:
- Compute access to run training and inference at scale
- A reinforcement learning framework that rewards successful task completion and penalizes errors, so models sharpen on specific business jobs
- Evaluation tools to measure whether the agent actually works
The platform runs like a marketplace. Customers pick the tools they need instead of buying into an all-or-nothing bundle. “They’ve stitched this together and built it in such a way that they’re operating at the frontier in a way that’s affordable,” said David Katz, a partner at Radical Ventures. He framed Prime Intellect as a “one-stop shop” that hands smaller teams the capabilities of a top-tier lab.
Why this matters now
A couple of years ago, this business wouldn’t have worked. Training a capable model meant frontier-lab budgets and talent. Reinforcement learning changed the math. Companies can now refine open models for narrow tasks and, in Prime Intellect’s framing, become their “own AI lab.”
The results are landing. Ramp used the platform to build an agent that finds answers inside spreadsheets. “The result beat the frontier models on accuracy while running at faster speeds and a fraction of the cost,” said Ramp co-founder and co-CEO Karim Atiyeh. Zapier and Flapping Airplanes are also paying customers, per TechCrunch AI.
There’s a second force driving demand, and it’s about risk. Enterprises are getting nervous about feeding proprietary data into closed models they don’t control. They’re also wary of building on models that can vanish overnight, the way Anthropic’s Fable was shut off last month. “How do I know that I’m not working with a company that is going to try to replace me and generalize to what I’m doing,” Katz said. That fear of dependence is pushing buyers toward owning their intelligence rather than renting it.
The bigger bet
This is significant because it challenges the assumption that serious AI capability has to flow through a handful of labs. If companies can train competitive agents in-house for less money and more control, the frontier labs lose some of their gravitational pull on enterprise budgets.
CEO Vincent Weisser put the mission bluntly. “It shouldn’t just be a few nerds in a glass tower in San Francisco that have the capability to train AI models,” he told TechCrunch. “It should be every enterprise, every nation state.”
What to watch next: whether more enterprises shift real workloads off closed models, and whether the “own your AI” argument holds up as frontier labs answer with their own privacy and control guarantees. For now, Prime Intellect has $130 million and a billion-dollar valuation to press the case. Full details are in the original TechCrunch AI report.