Groq
LPU inference chips and the GroqCloud inference service
About Groq
Groq designs the Language Processing Unit, an inference-focused chip, and runs GroqCloud for hosted model inference and GroqRack for on-premises deployment. In late 2025 Nvidia struck a licensing deal for its technology and hired its leadership, while the company continued operating GroqCloud.
Common use cases
- Real-time chat inference
- Agent workloads
- Voice applications
Products and services
GroqCloud
Hosted inference API.
GroqRack
On-premises LPU clusters.
Language Processing Unit
Inference accelerator architecture.
Recent product updates
- Dec 2025: Nvidia agreed a non-exclusive license of Groq's inference technology; senior leaders moved to Nvidia, Groq continues independently and says GroqCloud keeps operating.
Key features
- Inference-focused chip
- Hosted API
- Low-latency serving
- On-prem racks
- Developer console
- Open model support
Integrations
- OpenAI-compatible APIs
- LangChain
- LlamaIndex
Groq alternatives
FAQs
What does Groq sell?
Inference through the GroqCloud API and LPU-based GroqRack systems for on-premises use.
Is Groq still independent?
Groq continues to operate GroqCloud after Nvidia's December 2025 licensing agreement, although its leadership team changed.
How is cloud usage priced?
GroqCloud is generally priced per token, with enterprise arrangements available.
Work at Groq? Request updates to this profile. Update this profile
Compare with similar providers
All AI Hardware providers5CData center and AI factory infrastructure provider
AetinaEdge AI computing platforms and modules
AivresAI and general-purpose servers and rack-scale systems
AltairSimulation, HPC workload management and data analytics software
AMDCPUs, Instinct GPUs and adaptive chips for AI from data center to edge
ArmProcessor architecture and IP licensed for data center to edge AI
Arrow ElectronicsTechnology distributor and solutions provider for components and enterprise computing
ASRock RackServer and workstation hardware maker