GroqCloud
A managed inference cloud offering API access to leading open models with fast token generation, serving developers and enterprises that need low-latency, cost-efficient inference at scale.
TECHNOLOGY VENDOR · DATA CENTER
Groq Fast, low-cost AI inference on purpose-built LPU infrastructure.
Groq is an AI inference company founded in 2016 by former Google engineers. It pioneered the Language Processing Unit (LPU), a deterministic accelerator designed for running large language models with very low latency, and operates GroqCloud, an inference cloud used by more than five million developers. In December 2025 NVIDIA licensed Groq's inference technology on a non-exclusive basis, and Groq continues as an independent inference cloud business.
- Headquarters
- Mountain View, United States
- Founded
- 2016
- Category
- Data center
WHO THEY ARE
About Groq.
Groq is an AI inference company founded in 2016 by former Google engineers. It pioneered the Language Processing Unit (LPU), a deterministic accelerator designed for running large language models with very low latency, and operates GroqCloud, an inference cloud used by more than five million developers. In December 2025 NVIDIA licensed Groq's inference technology on a non-exclusive basis, and Groq continues as an independent inference cloud business.
Groq runs 13 data centres across North America, Europe, the Middle East and Asia-Pacific, including a facility in Dammam, Saudi Arabia, that has served production traffic since early 2025 under a USD 1.5 billion agreement with the Kingdom. Groq is an official inference provider to HUMAIN, giving regional enterprises and government bodies access to high-speed inference with in-country data processing.
Industries served
- Government and sovereign AI
- Financial services
- Telecommunications
- Technology and software
- Healthcare
- Media
LATEST SOLUTIONS
The portfolio, today.
LPU and NVIDIA Groq 3 LPX
Purpose-built inference accelerators with on-chip SRAM and deterministic execution; Groq is among the first providers bringing the NVIDIA Groq 3 LPX platform to market alongside Vera Rubin systems.
GroqMetal dedicated infrastructure
Dedicated bare-metal inference capacity tuned for speed and reliability, giving organisations full control over their environment for sovereign, regulated or high-volume workloads such as national AI services.
GroqCore and GroqAssured
A tested inference stack that turns dedicated capacity into production-ready performance without in-house infrastructure expertise, plus enterprise-grade governance, auditability and control for organisations with strict compliance requirements.
Sovereign AI deployments
Regional inference capacity delivered with partners such as HUMAIN in Saudi Arabia, supporting Arabic language models and day-zero availability of new open models for local enterprises and government.
AI FOCUS
ALJAMMAZ AI LAYERInference infrastructure built for the agentic AI era
Groq's entire business is AI inference. Its LPU architecture avoids the caches and branch prediction of general-purpose processors, delivering predictable, high-throughput token generation for large language models. The company positions inference, not training, as the bottleneck for AI adoption, and is scaling capacity from 57 MW toward more than 200 MW to serve real-time and agentic workloads.
GroqCloud hosts leading open-weight models and combines LPUs with NVIDIA GPUs in disaggregated configurations following Groq's admission to the NVIDIA Cloud Partner programme in 2026. In Saudi Arabia, Groq's Dammam facility powers HUMAIN's inference services, including HUMAIN One, and delivered OpenAI's gpt-oss open models locally on day one.
- LPU-based inference with deterministic, low-latency execution
- GroqCloud API access to leading open models
- Dedicated GroqMetal capacity for sovereign and regulated workloads
- Official inference provider to HUMAIN in Saudi Arabia
AI factories
Where Groq fits in the four-layer AI stack we bring to partners.
Explore the layerWHAT'S NEW
Recent announcements.
Groq closes USD 350 million Series A
Led by Disruptive, the round funds expansion of Groq's inference cloud, which now combines Groq 3 LPUs with NVIDIA GPUs across 13 data centres, following Groq's admission to the NVIDIA Cloud Partner programme.
USD 650 million raised to scale inference cloud
Growth capital led by Disruptive and Infinitum accelerates expansion of GroqCloud toward 200 MW of capacity by 2027, serving Fortune 500 enterprises and more than five million developers.
Non-exclusive inference technology licence with NVIDIA
NVIDIA licensed Groq's inference technology on a non-exclusive basis; founder Jonathan Ross and other leaders joined NVIDIA, while Groq continues as an independent company operating GroqCloud without interruption for customers.
Groq powers HUMAIN One enterprise AI platform
Groq's inference infrastructure underpins HUMAIN One, HUMAIN's real-time AI operating system for enterprises, extending the partnership under which Groq serves as an official inference provider in Saudi Arabia.
WITH ALJAMMAZ
Become a partner How we bring Groq to market.
01
Solution design & pre-sales
Architecture, sizing and proof-of-concept support so partners propose the right configuration first time.
02Enablement & training
Certification paths, workshops and demo access that build partner capability around the portfolio.
03Delivery, support & after-sales
Local stock, professional services, escalation and RMA handling across Saudi Arabia and the region.
READY WHEN YOU ARE
Build with Groq?
Speak with our team about availability, solution design and partner enablement.
