Groq built its reputation on a chip it called a language processing unit, designed specifically to run AI models fast and cheaply. That chip was the reason Nvidia came calling. In December 2025, Nvidia paid $20 billion to license that technology, and the founder, Jonathan Ross, the president, and much of the core engineering team all left with the deal to join Nvidia.
What remained was a cloud business, a network of data centers that rents out AI processing capacity, and a company that needed new leadership and a new direction. On June 22, Groq confirmed a $650 million funding round led by existing investors Disruptive and Infinitum, the same firms that had already collected a substantial payout from the Nvidia deal.
That investor behaviour is worth noting. These backers already got paid. The fact that they are reinvesting into the rebuilt company means they see a real commercial opportunity in the cloud business, not just loyalty to the brand. The round was effectively guaranteed from the start: Disruptive and Infinitum agreed to cover the full $650 million if other existing investors passed.
Groq's new focus is on selling AI processing as a service. Every time a company or developer runs an AI model, that is called inference, the part that happens after the AI has been trained. It is the piece that runs millions of times a day in real products, and it requires fast, affordable computing. Groq built its data center network around exactly this use case, and it now operates across North America, Europe, the Middle East, and Asia-Pacific.
The competitive picture is not gentle. CoreWeave, which went public earlier this year and has a revenue backlog approaching $100 billion, is the largest pure-play competitor in this space. Lambda Labs, Together AI, and the major cloud providers from Amazon, Google, and Microsoft all compete for the same enterprise contracts. Groq is entering this market without the hardware differentiation that once made it special, now that Nvidia holds the key chip technology.
The parallel to watch is Scale AI. After Meta paid $14.3 billion in a structurally similar deal in mid-2025, taking CEO Alexandr Wang and a 49% stake, Scale was written off in some quarters. It has since grown to an estimated $2 billion in revenue for 2025, with its data and applications businesses both expanding. The lesson is that these not-acqui-hire deals do not necessarily kill the remaining company.
For business operators, the practical takeaway is simpler than the corporate drama. Groq's cloud service is a real option for any organization that needs fast, affordable AI processing without building its own infrastructure. Companies across insurance, manufacturing, logistics, and retail are increasingly running AI models in their day-to-day operations, and the cost and speed of that processing matters. Services like Groq, CoreWeave, and their competitors exist to serve that demand at commercial scale.
What Groq's story also illustrates is how quickly AI company structures can shift. A firm that raises $750 million at a $6.9 billion valuation in September can license away its core technology for $20 billion by December and be raising again by the following June. For anyone evaluating AI vendors or planning multi-year technology relationships, that pace of change is worth factoring in.