Groq has raised $350 million in a Series A round to create a world-leading cloud for artificial intelligence inference. The company aims to halve the cost of running AI inference tasks. The $350 million Series A round was led by Disruptive, with expected participation from NVIDIA.
In the last approximately 60 days, Groq increased its total raised funds to $1 billion, following a $650 million round in June. However, the current company valuation of $3.5 billion is lower than the $6.9 billion valuation it received in September 2025 after raising $750 million.
Initially, in 2025, Groq positioned itself as an independent chip company aiming to challenge NVIDIA's dominance in AI hardware. However, by August 2026, NVIDIA itself plans to become an investor.
Announcing the $350 million raise, Groq explained that its mission is to transform the platform into a global leader in cloud computing for AI inference. Alex Davis, CEO of Groq and founder of Disruptive, stated that 'inference will undoubtedly become the largest and most critical layer of AI infrastructure.'
The decrease in valuation is attributed to Groq losing and being unable to retain certain assets. A turning point was December 2025, when NVIDIA reportedly paid $20 billion for a license to Groq's core inference technology.
As part of this deal, NVIDIA also hired Jonathan Ross, President of Sunny Madre, and about 90% of the engineering team that developed Groq's Language Processing Blocks. Ross, formerly an engineer at Google and involved in creating Google TPUs, founded Groq in 2016.
Despite the changes, Groq continues to operate in 13 data centers, serving over six million developers, and plans to increase capacity from 54 MW to over 200 MW by 2027. The company maintains its status as an NVIDIA Cloud partner and is certified to run NVIDIA accelerated computing according to the NVIDIA reference architecture.
The main idea promoted by Groq is that running AI models in production environments will ultimately become more significant than training them. The company asserts that as models transition from labs to finished products, the costs and computational power required for inference will surpass training costs. While others share this view, Groq no longer holds the most obvious technical advantage.
In July 2026, Together AI closed an $800 million round at an $8.3 billion valuation, targeting the same inference market. Fireworks AI also raised $1.5 billion in a Series D round at a $17.5 billion valuation, with NVIDIA support.
Thus, Groq is now competing not so much on the level of chip technology, but on operational scale. Its core LPU technology is now integrated into NVIDIA. Groq's advantage lies in its existing data center base and developer customer base.
The relationship with NVIDIA has changed: it has shifted from direct competition to a supply agreement with a major investor. Groq remains certified to work with NVIDIA hardware, effectively becoming a large-scale inference operator on top of the NVIDIA stack. The central question for Groq is whether it can build a sustainable business after transferring the license for the technology that made it unique. With $1 billion in fresh capital and NVIDIA as a shareholder, Groq has the resources to scale. Whether this turns into an independent platform or a dependent computing provider sold by NVIDIA will determine if this down round was merely a dump or the start of a long decline.