Groq

Overview

Groq designed the LPU, a deterministic low-latency inference processor, and runs GroqCloud, an OpenAI-compatible API for open-weight LLMs, speech and other models from its own data centers. It now positions itself as an inference neocloud whose LPX systems work alongside NVIDIA GPUs.

In December 2025 NVIDIA took a non-exclusive licence to Groq's inference technology (about $20 billion) and hired founder Jonathan Ross and other leaders; Groq stayed independent and kept operating GroqCloud.

Plans & prices

Data centers · 13

Networks (ASN)

Similar providers

Together AIOpen-model inference API plus self-serve and reserved NVIDIA GPU clustersfalGenerative media platform with pay-per-use model APIs and serverless GPUsCirrascaleSan Diego AI cloud with NVIDIA, AMD, Qualcomm and Tenstorrent systems on term pricingHugging FaceAI model hub selling GPU Spaces, dedicated Inference Endpoints and routed inferenceDeepInfraLow-cost model inference API plus dedicated A100-to-B300 GPUs billed by the minuteFireworks AIFast inference and fine-tuning platform with per-second on-demand GPU deploymentsBasetenInference platform for dedicated model deployments and model APIs, billed per minuteGMI CloudNVIDIA Cloud Partner renting H100/H200/B200 and GB200 from its own data centers