According to Axios, citing sources, AI chip startup Groq is seeking a new round of $650 million in funding, primarily to expand its inference cloud business. The company is currently betting on inference infrastructure based on its self-developed chips and systems, providing hosted AI application services to developers and enterprises.
Funds are being invested in reasoning cloud computing.
Groq's current business focus is on inference cloud, which provides a runtime environment for applications that require significant inference computing power. Compared to model training, inference corresponds to the actual processing after a user provides a prompt. The current AI market has a greater demand for this type of computing power and is closer to commercialization.
The report mentions that this round of financing will primarily involve existing investors. Groq hopes to use this funding to further expand its inference cloud business and strengthen its position in the AI infrastructure market.
They had a deal with Nvidia last year.
Groq reached a special deal with Nvidia last December. According to reports, the deal was worth approximately $20 billion, but it wasn't a full acquisition. The deal included the transfer of some senior Groq employees to Nvidia and Groq licensing its hardware technology to Nvidia.
Axios stated that this arrangement provided Groq's investors with a cash return. If calculated on a full acquisition basis, this could have been the largest acquisition in Nvidia's history.
Existing shareholders may fill the financing gap.
Currently, Groq's new direction is being driven by interim CEO Adam Winter and CFO Matt Eng. The company is reorganizing its growth plans around its inference cloud business.
According to reports, Groq's investors Disruptive and Infinitium have agreed to fill the funding gap in this round should other existing shareholders not participate in the investment proportionally. This means that the $650 million in funding is, to some extent, secured.
If the funding round is completed, Groq will continue to focus on inference-side infrastructure rather than competing on the training side. This also reflects the shift in focus of the AI chip and cloud services market towards inference computing power supply.
