What Groq does

Groq, Inc. is a privately held AI infrastructure company founded in 2016 that designs the LPU (Language Processing Unit), a deterministic, SRAM-based, statically scheduled inference accelerator, and operates GroqCloud, a hyperscale AI inference cloud spanning 13 data centers across North America, Europe, the Middle East, and Asia-Pacific. The company's technical core is custom silicon purpose-built for low-latency, high-throughput LLM serving — featuring an on-chip SRAM memory architecture with up to ~7x the memory bandwidth of leading GPUs and a synchronous low-diameter interconnect that scales to thousands of chips. On top of the LPU, GroqCloud offers an OpenAI-compatible REST API, a developer/enterprise tiered product with a Free, Developer (pay-per-token), and Enterprise (custom commitment) plan, plus add-on modules including Compound agentic AI, Tool Use with built-in and MCP-compatible remote tools, Speech-to-Text (Whisper), Text-to-Speech (Orpheus), OCR/Vision, Reasoning, Content Moderation, Structured Outputs, Prompt Caching, and LoRA Inference, served across Performance, Flex, and Batch processing tiers.

In December 2025, Groq entered a ~$20 billion non-exclusive LPU technology licensing and acqui-hire arrangement with Nvidia: founder/CEO Jonathan Ross and president Sunny Madra departed to Nvidia while Groq retained full ownership of its LPU IP and continued independently under interim CEO Adam Winter and CFO Matt Eng. The company subsequently rebranded its strategy as 'Groq 2.0', pivoting from standalone chip sales to an AI inference neocloud model, closing a $650 million growth-capital round in June 2026 (led by Disruptive and Infinitum) to scale to 200MW of compute capacity by end-2027. Groq monetizes through usage-based developer pricing, prepaid Service Credits, enterprise subscription contracts, and — historically — chip and IP licensing. Its customer base spans 3M+ developers and thousands of AI-native companies, alongside named enterprise customers such as Dropbox, Vercel, Chevron, Volkswagen, Canva, Robinhood, Riot Games, Workday, Ramp, McLaren F1, and the PGA of America; Forbes reported revenue near $100 million at the time of the December 2025 Nvidia transaction.

Name

Groq

Legal name

Groq, Inc.

Website

https://groq.com

Company type

Private

Founded year

2016

Operating status

Operating

Headcount range

101–250 employees

Short description

Groq designs the LPU (Language Processing Unit), a purpose-built inference accelerator, and operates GroqCloud, a hyperscale AI inference cloud across 13 global data centers serving 3M+ developers, thousands of AI-native companies, and enterprise customers including Dropbox, Chevron, Volkswagen, and Workday.

Ownership category

akta.pro rank

Product category

AI Inference Cloud Infrastructure

NAICS

Software Publishers (513210), Computer Systems Design and Related Services (54151), Computer Systems Design Services (541512)

SIC

Services-Prepackaged Software (7372), Services-Computer Integrated Systems Design (7373)

akta.pro primary industry

Model Governance, Risk & Compliance (GRC) Platforms (HDAAAKAA)

akta.pro secondary industries

AI Governance, Risk & Compliance (GRC) Platforms (HDAAAMAA), Enterprise AI Governance, Risk & Compliance Platforms (Model Risk, Audit, Policies) (HDAEANAE), Cloud Compliance, Audit & Continuous Controls Monitoring (CCM/GRC) (HDABAHAI)

Keywords

• AI inference cloud
• Custom inference silicon
• LLM serving platform
• Language processing unit
• Neocloud infrastructure

Headquarters

HQ city

San Jose

HQ country

United States

HQ region

North America

Offices5 records

Markets served

GTM type

B2B

Offering type

Software

Cost components

Technology or R&D, Infrastructure, Personnel, Supply Chain, Operations, Marketing or Sales

Revenue model

1. Developer tier — pay per token: Self-serve Developer plan charges customers per token consumed for LPU-powered inference, with higher token limits, chat support, and access to Flex and Batch processing tiers.
2. Enterprise tier — committed contracts: Enterprise plan sold via Order Form with custom commitments, scalable capacity, dedicated support, LoRA Inference, SSO/SCIM, 90-day audit logs, and prepaid Service Credits; supports multi-year, committed-spend contracts.
3. Prepaid Service Credits: Customers can prepay for Cloud Services by purchasing Service Credits (with optional auto-reload Balance Maintenance), which can be redeemed against usage; Service Credits expire 1 year after purchase.
4. Free tier: Free tier for developers to build and test on Groq APIs with community support; functions as a top-of-funnel acquisition channel.
5. Historical hardware / IP licensing (transitioning): Prior revenue included chip sales and a $20B non-exclusive licensing deal with Nvidia for Groq's LPU technology (December 2025); the company is now pivoting to AI inference cloud and infrastructure services rather than standalone chip sales.
6. Neocloud / AI inference cloud services: Strategic shift to operating 13 LPU-powered data centers across North America, Europe, the Middle East, and APAC, processing trillions of AI tokens weekly for developers and AI-native companies; aiming to scale to 200MW of capacity by end of 2027.

Pricing tiers

Go-to-market motion5 records

Core offering

Groq develops and operates the LPU (Language Processing Unit), a custom inference accelerator silicon pioneered in 2016, and delivers it as a vertically integrated AI inference cloud called GroqCloud. GroqCloud exposes an OpenAI-compatible API plus developer and enterprise consoles that serve open and proprietary LLMs, multimodal, speech, and agentic AI workloads across 13 global data centers with deterministic low latency.

Product overview

Groq is a single-vendor, vertically integrated AI inference platform: its purpose-built LPU (Language Processing Unit) silicon powers the GroqCloud developer and enterprise platform, which together form the core of its offering. The portfolio combines a hardware layer (LPU architecture, deployed across 13 data centers globally targeting 200MW by 2027) with a software/platform layer (GroqConsole, GroqCloud, GroqChat, Groq Playground) and a set of add-on modules — Tool Use (with built-in tools like Web Search, Code Execution, Wolfram Alpha, and Browser Search), Remote Tools and MCP, Compound (agentic AI), Speech to Text (Whisper), Text to Speech (Orpheus), OCR/Image Recognition, Reasoning, Content Moderation, Structured Outputs, Prompt Caching, and LoRA Inference. Service is offered in three billing tiers (Free, Developer, Enterprise) with Performance, Flex, and Batch processing tiers, and an OpenAI-compatible REST API plus Python/JS SDKs and a partner catalog of coding-agent integrations (Factory Droid, OpenCode, Kilo Code, Roo Code, Cline).

Differentiator

Problem solved

Functional benefit

Brands

• GroqCloud: Cloud-based AI inference platform / console and APIs (including Groq Playground, GroqChat and the OpenAI-compatible Groq API) that runs on Groq's LPU-based stack.
• LPU (Language Processing Unit)
• GroqCloud Console
• Compound

Products and services

• GroqCloud Cloud-based AI inference platform that runs LPU-powered inference across 13 data centers, serving developers and enterprises with low-latency, low-cost access to a catalog of open and proprietary LLMs, vision, speech, and agentic AI models via an OpenAI-compatible API.
• Groq LPU (Language Processing Unit) Custom-built inference accelerator chip with a deterministic, statically scheduled, SRAM-based architecture optimized for low-latency, high-throughput LLM and multimodal inference at scale.
• GroqCloud Console Online developer and administrator console for managing GroqCloud accounts, organizations, API keys, billing, rate limits, usage, logs, and audit logs; includes GroqChat chat interface and Groq Playground prompt experimentation interface.

Quantifiable outcome

• 7.41x faster chat speed and 89% cost reduction, enabling a 3x increase in token consumption (Fintool)
• +4 more outcomes

Named customers13 records

Ideal customer profiles2 records

Technology focussed Yes

API detail

Has API

Yes

API docs

API detail

Core technology

AI maturity

App detail

Integration9 records

AI capability14 records

Feature10 records

Partnerships

Six partnerships are on record, tiered strategic supplier — exclusive rack-scale system integrator for groq lpu-based products., flagship strategic partner — entered a ~$20b non-exclusive licensing and acqui-hire deal for groq's lpu technology., core manufacturing partner — fabricates groq's lpu silicon on samsung's 4nm process node., flagship partnership — official partner of the mclaren formula 1 team., ecosystem technology partner — primary agent framework for groq-powered apps. and strategic ecosystem alignment — groqcloud is openai-compatible for seamless migration..

• Foxconnstrategic supplier — exclusive rack-scale system integrator for groq lpu-based products.Technology or Integration · 1 April 2026Selected as the exclusive rack-scale supplier for the Groq 3 LPX rack, which packs 256 language processing units per rack with 640 Tb/s scale-up bandwidth. Initial shipments of approximately 6,000 racks are scheduled for Q3 2026, followed by 10,000 racks in 2027. Foxconn plans to double its data center cabinet production capacity to 2,000 per week to support this and other demand.
• Nvidiaflagship strategic partner — entered a ~$20b non-exclusive licensing and acqui-hire deal for groq's lpu technology.Strategic or Co-development Partner · 1 December 2025Signed a non-exclusive licensing agreement for Groq's LPU chip technology in December 2025 (reportedly $17B–$20B), and hired away Groq's founder/CEO Jonathan Ross, president Sunny Madra, and other key engineering talent. Groq retained full ownership of its LPU IP. Nvidia productized the licensed technology as the Groq 3 LPX / LP30, manufactured by Samsung on 4nm, and integrated it into its Vera Rubin rack-scale platform. The deal is under U.S. Senate antitrust scrutiny (Senators Warren and Blumenthal).
• Samsung Foundrycore manufacturing partner — fabricates groq's lpu silicon on samsung's 4nm process node.OEM/ Whitelabel/ Licensing Partner · 1 September 2025Manufactures Groq's third-generation LPU chips at its Pyeongtaek campus using the 4nm (SF4) process. Samsung secured a deal to produce Groq's AI6-equivalent LPU for AI inference; Groq is one of the foundry customers (alongside Tesla) that has shifted to Samsung from TSMC for advanced-node AI chip production.
• McLaren Racing (McLaren F1 Team)flagship partnership — official partner of the mclaren formula 1 team.GTM or Marketing PartnerMulti-year global partnership providing the McLaren F1 Team with Groq-powered inference for decision-making, analysis, development, and real-time insights; co-marketed through a 'Partnership Spotlight' on groq.com and the McLaren newsroom announcement.
• LangChain (LangGraph)ecosystem technology partner — primary agent framework for groq-powered apps.Technology or IntegrationGroq is featured as a first-class inference backend in LangChain/LangGraph tutorials, with developers using Groq's OpenAI-compatible API and llama-3.3-70b-versatile as the reasoning backbone for agentic research assistants with tool calling, sub-agents, and persistent memory.
• OpenAI (API compatibility)strategic ecosystem alignment — groqcloud is openai-compatible for seamless migration.Technology or IntegrationGroq's API is OpenAI-compatible (https://api.groq.com/openai/v1) and supports a Responses API mirroring OpenAI's, plus day-zero support for OpenAI open models; reduces switching friction to two lines of code.

Scale indicators12 records

Recent moves8 records

Expansion highlights6 records

Direct peers

• Cerebras Systems: Direct peer building custom AI silicon and an inference cloud targeting low-latency LLM serving. Competes with Groq in the same inference-chip category and the same neocloud distribution model.
• SambaNova Systems: Direct peer offering a custom AI accelerator (RDU) and an enterprise AI cloud for inference. Competes with Groq in custom-silicon inference for enterprise and government customers.
• CoreWeave: Neocloud peer providing GPU-based AI compute and inference at scale. Competes with GroqCloud for the same AI-native enterprise workloads, though CoreWeave runs Nvidia GPUs rather than custom silicon.
• Lambda Labs: Neocloud peer offering GPU clusters and on-demand inference for AI developers. Competes head-to-head with GroqCloud on pay-per-token inference pricing and developer self-service.
• Together AI: Neocloud peer providing hosted open-source model inference and fine-tuning infrastructure. Competes with GroqCloud for the same AI-native developer and enterprise inference workloads.
• Fireworks AI: Inference-focused neocloud peer offering fast, low-cost open-model serving to developers. Directly comparable product proposition ("fast, low-cost inference") to GroqCloud's value proposition.

Broad incumbents

• Nvidia: Broad incumbent in AI compute and now a strategic partner of Groq via the ~$20B LPU licensing/acqui-hire. Nvidia productized Groq's LPU IP inside its Vera Rubin platform, making it both licensor/partner and direct inference competitor.
• OpenAI: Broad incumbent in generative AI and the API whose compatibility Groq emulates. Indirectly comparable as the dominant inference provider that Groq's OpenAI-compatible API is designed to siphon workloads from.
• Anthropic: Frontier AI lab serving its own Claude models via cloud inference at scale. Comparable as a major hosted AI inference provider for enterprise and developer customers.

Emerging players

• Hugging Face: AI platform offering open-model hosting, inference endpoints, and an enterprise catalog. Comparable as an emerging inference cloud alternative for developers; competes with GroqCloud on open-model availability and developer mindshare.

Market position

Strengths5 records

Weaknesses5 records

Competitive moat6 records

Key risks7 records

Key highlights7 records

Customer concentration

• Xx.com/groqinc
• YouTubeyoutube.com/c/GroqInc
• LinkedInlinkedin.com/company/groq
• Instagraminstagram.com/groqinc
• Facebookfacebook.com/groqinc

Compliance6 records

Revenue estimate

Valuation estimate

Number of profiles

Profiles8 records

Subsidiaries3 records

Funding overview

Funding rounds11 records

Investors69 records

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A2 records

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about Groq