Nvidia, Intel, and Dell Just Funded a Startup Whose Pitch Is 'Stop Renting Your Intelligence From OpenAI.'

Three companies that manufacture the chips frontier AI labs depend on just wrote checks into a startup whose entire pitch is that enterprises should stop depending on frontier AI labs.

Prime Intellect closed a $130 million Series A on July 8 at a $1 billion valuation, led by Radical Ventures with Nvidia Ventures, Intel Capital, and Dell Technologies Capital all participating, alongside Iconiq and a roster of operator-angels that includes Perplexity's Aravind Srinivas and Box's Aaron Levie. The company's product, which it calls the "Open Superintelligence Stack," gives enterprises the compute, reinforcement-learning tooling, and evaluation infrastructure to train and continuously improve their own models — rather than calling an API and hoping the vendor's next price increase or capability change doesn't break their product.

The traction isn't hypothetical. Prime Intellect already runs at over $100 million in annualized revenue across roughly 6,000 customers, including Ramp and Zapier. Ramp used the platform to train a 35-billion-parameter model that beat Anthropic's Opus at searching spreadsheets — 27% faster, and at a fraction of the inference cost of a much smaller model. Zapier co-founder Mike Knoop and Ramp co-CEO Karim Atiyeh both appear on the round as individual angel investors, which is its own signal: the customers with the clearest view of Prime Intellect's product are also betting personal capital that the category survives.

The obvious reading is that Nvidia, Intel, and Dell are simply chasing another hot AI round. The more precise reading is a hedge. If enterprises rent intelligence through two or three frontier labs' APIs, compute demand concentrates behind whichever cloud or custom-silicon deal those labs have already signed. If instead enterprises train and host their own models, the number of independent buyers of raw GPU-hours multiplies, and none of them can renegotiate compute pricing with the leverage of an OpenAI or an Anthropic. Model independence, for a chip vendor, means demand fragmentation — and fragmented demand is demand a supplier doesn't have to discount.

This lands in a specific moment. OpenAI and Anthropic together absorbed 43% of a record $510 billion in global H1 2026 venture funding, by Crunchbase's count — concentration extreme enough that every other player in the AI stack has a direct financial interest in a countervailing force existing. Prime Intellect is a small check by megaround standards, but it's aimed exactly at that concentration, funded by companies with no upside in the current arrangement continuing indefinitely.

The bet isn't that Prime Intellect's customers will out-build OpenAI's frontier models. It's that "good enough, and owned" beats "best-in-class, and rented" for a wide swath of enterprise use cases — and that whoever sells the picks and shovels for that shift doesn't much care which side of the API call wins, as long as the compute meter keeps running either way.