Groq spent years selling an alternative to GPU-based AI inference. Its next phase looks very different.
The company announced a $350 million Series A led by Disruptive, valuing the inference provider at $3.5 billion. NVIDIA is expected to participate, subject to customary closing conditions, while the funding will support an expansion that includes NVIDIA accelerated computing systems.
Fresh capital gives the company more room to compete for enterprise AI workloads as its infrastructure footprint grows.
New funding backs a larger compute footprint
Groq says it currently operates 13 data centers with 54 MW of capacity and plans to exceed 200 MW during 2027.
The new financing will support medium and larger clusters of NVIDIA accelerated computing for training and inference, according to the company. Added capacity could broaden the workloads Groq can pursue and give enterprises another source of compute outside the largest hyperscalers.
Groq executive chairman Alex Davis called inference “the largest and most critical layer of AI infrastructure.” His forecast assumes demand will keep moving toward the compute needed to run deployed AI models, giving large-scale inference providers a bigger piece of enterprise infrastructure spending.
A licensing deal changed Groq’s direction
December brought the first major turn. A non-exclusive licensing agreement gave NVIDIA access to Groq’s inference technology while the company remained independent and kept GroqCloud running.
By June, cloud infrastructure had become central to the business as neocloud providers expanded their role in enterprise AI. A $650 million financing round went toward expanding its footprint, including work with infrastructure and data center operators around the world.
August brought another step closer to the GPU giant. Joining the NVIDIA Cloud Partner program allows the provider to deploy its partner’s accelerated computing systems in existing facilities. Customers will also be able to gain more capacity and newer inference technology without changing application code, according to Groq, reducing one potential hurdle as the underlying infrastructure changes.
Partners should separate capacity from channel opportunity
MSPs and systems integrators advising customers on AI infrastructure can consider the provider when comparing alternative sources of GPU capacity. Inference-heavy deployments remain an obvious fit, while larger training clusters broaden the workloads worth evaluating. Partners can benchmark representative customer workloads against hyperscalers and other neoclouds before committing capacity.
Resellers and VARs face a different question. No new reseller terms, marketplace route, or channel program accompanied the financing, so NVIDIA’s planned participation does not automatically create something partners can transact today. Until a formal route appears, partner revenue could come from advisory and managed infrastructure work around customer deployments.
Moving workloads to Groq can diversify the cloud provider without necessarily reducing reliance on NVIDIA at the hardware layer. A second provider does not automatically reduce accelerator dependence. Partners helping customers compare AI infrastructure options should separate cloud-provider choice from hardware dependency when weighing longer-term sourcing decisions.
More news: NVIDIA is backing OpenAI’s 8GW Ohio project with computing infrastructure, investment, and credit support.