Groq is raising up to $650 million from existing investors as it converts itself into an AI inference neocloud business, Axios reported Thursday. Disruptive and Infinitum are backstopping the round, so it closes regardless of broader appetite.
The pitch follows December’s $20 billion Nvidia licensing deal that emptied the house: founder Jonathan Ross, president Sunny Madra, and most senior leadership left to join Nvidia. That’s when CFO Simon Edwards took the CEO chair. For Groq 2.0, Adam Winter is serving as CEO and Matt Eng as CFO. Before reinvesting in “Groq 2.0,” existing shareholders will first receive cash distributions from the Nvidia proceeds, then invest pro rata into the successor vehicle.
What do investors actually get? GroqCloud, with 3.5 million developers on the platform as of February, and the inference infrastructure behind it. No chip IP, no founding team. The structure is comparable to Inflection in 2024: after that company’s leadership moved to Microsoft, what remained was a service layer absent the people and IP that built it. Groq’s last standalone valuation was $6.9 billion in September, when it still had both.
Strip the IP and talent via a licensing deal, return capital to shareholders, re-raise into the service-layer residual. Axios flagged this as a potential template for AI private markets, and the framing is apt. If GroqCloud can sustain developer adoption competing against CoreWeave and Lambda Labs without a proprietary chip, the template works. If it can’t, $650 million is a steep price to learn that.
The $650M closes. Disruptive and Infinitum ensure that. The open question: what multiple do investors assign to an inference cloud that surrendered its chip IP and founding team to Nvidia, because that number becomes the template for every AI company that follows.
— Diana Kowalski