Groq Raises 650 Million After the Nvidia Deal
The chip startup is pivoting to an inference neocloud business after Nvidia took its hardware and most of its team for 20 billion.
- Groq is raising up to 650 million dollars from existing investors, announced on May 28.
- The round follows a 20 billion licensing deal with Nvidia.
- Nvidia took the chip architecture and most of the leadership, leaving the neocloud platform.
- The new Groq is run by interim CEO Adam Winter and interim CFO Matt Eng.
What is left of a chip startup once a bigger player buys both the technology and the people? For Groq, the answer is: the platform, the customer contracts, and 650 million dollars in fresh cash for a fresh start.
The company is raising up to 650 million dollars from existing investors, as announced on May 28. The money goes toward a pivot to a so called inference neocloud business, meaning renting out compute to run already trained AI models.
What did Nvidia actually buy?
In December the two signed a licensing deal worth roughly 20 billion dollars, the largest in Nvidia’s history. On paper it was not an acquisition, but the result is very close to one.
Nvidia took Groq’s accelerator architecture and most of its leadership. Existing investors were paid out in cash through the transaction, and some of them can now reinvest in the new entity.
What stayed with Groq is not the hardware but the software layer and the business relationships. That includes the neocloud platform and existing contracts with customers who already use Groq for inference.
Why is Groq turning to the inference business?
Inference is the processing that happens after a user sends a prompt to a model. Right now that is a much bigger market need than training new models, because every prompt, every chatbot reply and every search burns that compute again.
The new entity is run by two company veterans. Adam Winter is the interim chief executive, and Matt Eng is the interim chief financial officer.
Existing backers Disruptive and Infinitum are ready to fill the round if other investors do not close it through their pro rata participation. That signals the main money partners still believe in the plan, even if it is not the one they first signed up for.
This structure, where investors get paid out through a deal with a large player and can then reinvest in a new company, could become a template for the AI private market. A big buyout and a continued business are no longer two separate stories.
Would you trust an AI startup more if it sold a chip or if it rented out the power to run the models?
