Prime Intellect raises $130M Series A at $1B valuation

Prime Intellect raised a $130 million Series A at a $1 billion valuation to sell a full stack for enterprise AI agent development. The company, founded in 2024, packages compute access, a reinforcement learning framework and evaluation tools into a hosted, modular marketplace used by customers including Ramp, Zapier and Flapping Airplanes.

What does Prime Intellect’s platform include?

Prime Intellect offers compute access, a reinforcement learning framework and evaluation tools as a single platform, and provides them modularly so customers can pick components rather than adopt an all-or-nothing stack. The company describes the offering as a “full stack” for AI agent development and runs a hosted version that customers pay for.

Beyond that short definition, Prime Intellect positions the platform like a marketplace: customers select the specific tools they need, and the startup supplies hosted compute and software that stitches the pieces together. That approach is intended to reduce the engineering complexity enterprises face when assembling production-ready agentic systems.

Who invested and who’s using it?

The $130 million round was led by Radical Ventures and included Nvidia Ventures, Intel Capital, Dell Technologies Capital, Iconiq and a group of angel investors who are founders at other AI companies. Named angels include Aravind Srinivas (Perplexity), Aaron Levie (Box), Winston Weinberg (Harvey), Jeff Wang (Cognition) and Brendan Foody (Mercor).

Prime Intellect already counts Ramp, Zapier and Flapping Airplanes among paying customers. The startup has reached an annualized revenue run rate of $100 million, driven in part by hosted deployments. Ramp used the platform to build an agent for finding answers inside spreadsheets; Ramp co-founder and co-CEO Karim Atiyeh said the result “beat the frontier models on accuracy while running at faster speeds and a fraction of the cost.”

Why are enterprises choosing this over frontier models?

Prime Intellect was founded in 2024 to let organizations train and refine their own agentic systems without relying on closed frontier AI labs. The company leans on reinforcement learning techniques, which iteratively reward successful task completion and penalize errors, to tailor models for specific business tasks.

Investors and customers cited two practical drivers. First, many companies do not want to provide proprietary information to OpenAI and Anthropic because of the risk of losing control over their data. Second, firms worry about depending on external models that can be turned off, an example being Anthropic’s Fable being shut down last month. David Katz, a partner at Radical Ventures, argued Prime Intellect has “stitched this together and built it in such a way that they’re operating at the frontier in a way that’s affordable.”

Why it matters

Prime Intellect packages tools and compute that historically only top AI labs could assemble, and it has matched real commercial demand: the company is already at a $100 million annualized run rate and attracted high-profile backers. If enterprises prefer locally tailored agentic systems over reliance on a few frontier labs, the market for on-premise or hosted enterprise agent tooling will expand and create a new set of infrastructure winners.

What to watch

Watch whether Prime Intellect can sustain and grow its $100 million ARR beyond early customers and whether more enterprises choose self-hosted or proprietary-agent development over frontier models. Also watch product comparisons: Ramp’s claim that its agent beat frontier models on accuracy sets a concrete benchmark the company will need to replicate at scale.

Prime Intellect CEO Vincent Weisser framed the company’s mission simply: “It shouldn’t just be a few nerds in a glass tower in San Francisco that have the capability to train AI models,” he said, arguing that enterprises and nation states should also be able to build models.