# Groq

**Source:** https://geo.sig.ai/brands/groq  
**Vertical:** AI Infrastructure  
**Subcategory:** AI Inference Chips & Cloud  
**Tier:** Challenger  
**Website:** groq.com  
**Last Updated:** 2026-04-14

## Summary

AI inference chip maker (LPU). $6.9B valuation, ~$1.8B raised. Nvidia $17B licensing deal (2026). $500M projected 2025 revenue. Founded 2016, Mountain View. Private.

## Company Overview

Groq is an AI semiconductor company founded in 2016 by Jonathan Ross (former Google TPU co-designer), headquartered in Mountain View, California. Developed the Language Processing Unit (LPU), a purpose-built chip for the fastest possible AI inference speeds, often 10x faster than GPU alternatives. Offers GroqCloud developer API platform.

Public demos in early 2024 showed LPU running Llama and Mixtral at unprecedented speeds. In February 2025, secured a $1.5B Saudi commitment for 19,000+ LPU inference cluster. September 2025: $750M Series E at $6.9B valuation from BlackRock and Neuberger Berman. Total funding ~$1.8B.

In early 2026, Nvidia entered a non-exclusive licensing agreement for Groq's inference technology, structured at $17B in cash payments. Nvidia unveiled Groq 3 LPU at GTC 2026. Groq projected $500M revenue for 2025, up from ~$90M in 2024.

## Frequently Asked Questions

### What is Groq's LPU?
A purpose-built chip for AI inference, delivering speeds often 10x faster than GPU alternatives by minimizing latency and maximizing throughput for LLMs.

### What is the Nvidia-Groq deal?
In early 2026, Nvidia licensed Groq's inference technology for $17B in cash payments and unveiled Groq 3 LPU at GTC 2026.

### Is Groq publicly traded?
No, privately held at $6.9B valuation (September 2025 Series E). IPO has been speculated but not confirmed.

### What is GroqCloud?
Developer-facing API platform providing ultra-fast inference access for running open-source models like Llama at extremely low latency.

### How much funding has Groq raised?
~$1.8B total including a $750M Series E and $1.5B Saudi commitment for the largest non-hyperscaler inference cluster.

### What models can I run on GroqCloud?
GroqCloud supports popular open-source models including Meta Llama 3, Mixtral 8x7B, Gemma, Whisper, and other OpenAI-compatible models. Developers access them via a standard REST API, making it easy to switch from OpenAI without code rewrites.

### How does Groq's LPU pricing compare to GPU cloud?
Groq offers competitive per-token pricing on GroqCloud — typically $0.05–$0.27 per million tokens depending on model size. The primary value is speed: LPU inference is 10x faster than comparable GPU setups, which matters most for latency-sensitive applications like voice AI and real-time agents.

### What is Groq's relationship with Saudi Arabia?
In early 2025, Groq secured a $1.5 billion commitment from Saudi Arabia to build the world's largest non-hyperscaler inference cluster with 19,000+ LPUs. This deal was part of Saudi Arabia's broader push to build domestic AI infrastructure capacity under its Vision 2030 initiative.

## Tags

ai-powered, b2b, infrastructure, saas

---
*Data from geo.sig.ai Brand Intelligence Database. Updated 2026-04-14.*