# Groq

> Source: https://aiwiki.ai/wiki/groq
> Updated: 2026-07-28
> Fact-checked: 2026-07-28
> Categories: AI Companies, AI Hardware
> License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) - attribute to "AI Wiki (aiwiki.ai)"
> Cite as: AI Wiki. "Groq." aiwiki.ai, 28 Jul 2026. https://aiwiki.ai/wiki/groq
> From AI Wiki (https://aiwiki.ai), the free encyclopedia of artificial intelligence. Reuse freely with attribution.

**Groq** is a United States artificial intelligence infrastructure company that operates an [AI inference](https://aiwiki.ai/wiki/inference) cloud and offers on-premises inference systems. It was founded in 2016 by former Google engineers Jonathan Ross and Douglas Wightman.[1][2] Ross had worked on Google's [tensor processing unit](https://aiwiki.ai/wiki/tpu) project, and his name appears among the contributors to Google's first published TPU performance study.[3] Groq developed a compiler-scheduled accelerator that its 2020 architecture paper called the Tensor Streaming Processor and that the company now markets as a language processing unit, or LPU.[4] In December 2025, Groq granted [Nvidia](https://aiwiki.ai/wiki/nvidia) a non-exclusive license to its inference technology, and Ross, company president Sunny Madra, and other employees moved to Nvidia; Groq and GroqCloud remained independent.[5] As of June 2026, Groq was led by chief executive Adam Winter and focused on operating and expanding its inference cloud.[6] This article covers the company, its history, financing, acquisitions, cloud business, partnerships, and leadership. For processor architecture, specifications, system design, and benchmark evidence, see [Groq hardware](https://aiwiki.ai/wiki/groq_hardware). Groq is unrelated to [Grok](https://aiwiki.ai/wiki/grok), the chatbot developed by [xAI](https://aiwiki.ai/wiki/xai).[7]

## Article scope

The company and hardware articles should remain separate. This page treats Groq as a corporate and cloud-infrastructure organization. It summarizes the LPU only to explain the company's products and strategy. The Groq hardware article is the place for the Tensor Streaming Processor architecture, deterministic scheduling, on-chip memory, inter-chip communication, rack configurations, specifications, and performance measurements. Product claims that depend on a particular model, precision, batch size, service tier, or date should not be copied into this company article.

The public record has used more than one legal-entity style. Groq's early securities filings identify "Groq, Inc." as a Delaware corporation formed in 2016, while Groq's current website terms identify Groq LLC or its subsidiaries or affiliates as the website operator.[1][8] This article therefore uses the public-facing name "Groq" unless a specific legal entity is material.

## History

Groq's October 2016 Form D listed Wightman and Ross as executive officers and directors, with Wightman signing as chief executive.[1] A contemporaneous account of the 2021 Series C described Wightman and Ross as co-founders and reported that Wightman had left the company more than two years earlier.[2] Ross subsequently served as founder and chief executive through the Nvidia licensing agreement in December 2025.[5]

The company initially concentrated on a software-defined processor and compiler stack. Its first peer-reviewed architecture description appeared at the 2020 ACM/IEEE International Symposium on Computer Architecture. The paper described a functionally sliced, single-core design with compiler-controlled instruction timing and data movement.[4] Those engineering details belong in the Groq hardware article; their corporate significance is that Groq developed both the accelerator and the software needed to deploy it.

Groq announced two acquisitions before its 2025 licensing agreement:

| Date | Company | Announced purpose |
|---|---|---|
| March 2022 | Maxeler Technologies | Add Maxeler's dataflow-computing team and high-performance-computing systems experience.[9] |
| March 2024 | Definitive Intelligence | Build out GroqCloud and create separate GroqCloud and Groq Systems business units; Definitive Intelligence co-founder Sunny Madra took charge of GroqCloud.[10] |

The Definitive Intelligence announcement also documented GroqCloud's soft launch on February 19, 2024. It described self-service access to the Groq API and a business model that sold metered inference capacity rather than requiring every user to buy a hardware system.[10]

## Products and business model

Groq's business centers on two delivery modes. GroqCloud provides hosted inference through developer APIs and public, private, or co-cloud instances. Its public plans include a free tier, usage-based developer access, and negotiated enterprise service. GroqRack provides an on-premises option for customers that need local or air-gapped infrastructure.[11] The supported-model catalog, prices, limits, and regional availability change over time, so they should be read from Groq's current service documentation rather than treated as permanent company facts.

The underlying service uses Groq's LPU systems and primarily hosts models published by other organizations. Groq sells the computing service used to run those models. The company's current API exposes endpoints for text and audio tasks and uses OpenAI-compatible endpoint paths for several operations.[12] Compatibility at the API level does not mean that Groq owns the hosted models or that every feature of another provider's API is available.

Following the Nvidia agreement, Groq narrowed its stated strategy around cloud operations. In June 2026, it said it operated 13 data centers in North America, Europe, the Middle East, and the Asia-Pacific region, served more than five million registered developers, and processed trillions of tokens each week. These are company-reported operating figures, not independently audited metrics.[6] Groq also said the new financing would help equip its existing footprint with inference systems that include Nvidia's LPX platform using the licensed Groq technology.[6]

## Financing

Groq's disclosed equity offerings and major announced financings are summarized below. The table does not count the Saudi infrastructure commitment described later as venture financing, and it does not treat the reported Nvidia licensing price as a valuation of the continuing company.

| Date | Financing | Amount | Disclosed valuation or context |
|---|---|---|---|
| October-December 2016 | Early equity offering | About $10.3 million sold | Six investors were reported in the amended Form D; no valuation was disclosed.[1] |
| August 2018 | Equity offering | About $52.3 million sold | Sixteen investors were reported; the notice listed a total offering of about $60.7 million.[13] |
| April 2021 | Series C | $300 million | Co-led by Tiger Global Management and D1 Capital; Groq said the round brought total funding to $367 million.[14] |
| August 2024 | Series D | $640 million | $2.8 billion valuation; led by funds and accounts managed by BlackRock Private Equity Partners.[15] |
| September 2025 | Financing round | $750 million | $6.9 billion post-money valuation; led by Disruptive.[16] |
| June 2026 | Growth capital | $650 million | Led by Disruptive and Infinitum; Groq did not disclose a valuation.[6] |

The figures are not directly interchangeable. A Form D reports an exempt securities offering, an announced funding round reports company financing, a commercial infrastructure commitment may fund contracted deployment, and a technology-license payment compensates intellectual-property and related transaction rights.

## Nvidia licensing agreement and leadership transition

On December 24, 2025, Groq announced a non-exclusive license of its inference technology to Nvidia. The public statement said Ross, Madra, and other Groq employees would join Nvidia, while Groq would continue as an independent company, Simon Edwards would become chief executive, and GroqCloud would continue operating.[5] The announcement did not disclose a price.

Axios subsequently reported that the agreement was worth about $20 billion and that shareholder distributions were tied to that figure even though no Groq equity changed hands.[17] Nvidia's fiscal 2026 annual report separately said that the increase in cash used for investing activities was primarily attributable to purchases of equity investments and execution of the non-exclusive Groq license, but it did not assign a standalone price to the Groq transaction.[18] The approximately $20 billion figure should therefore be described as reported, not as a price jointly disclosed by the companies.

The structure attracted political scrutiny. In March 2026, United States senators Elizabeth Warren and Richard Blumenthal asked Nvidia for information about whether the license and hiring arrangement had been designed to avoid merger review.[19] Their letter was a request for information and an expression of antitrust concern, not a judicial or regulatory finding that the transaction violated the law.

Edwards's tenure as Groq chief executive was brief. Bloom Energy announced in March 2026 that he would join it as chief financial officer effective April 13.[20] By June 22, Groq identified Adam Winter as chief executive, Matt Eng as chief financial officer, and Alex Davis of Disruptive as chair. It also announced Alan Rice as chief operating officer and July start dates for Sinclair Schuller as chief technology officer and Rakesh Malhotra as chief product officer.[6]

At GTC in March 2026, Nvidia introduced the Nvidia Groq 3 LPX inference system as part of its Vera Rubin platform. Nvidia described the system as built around Groq-derived LPU technology.[21] That product is an Nvidia platform enabled by the license; it should not be presented as proof that Nvidia acquired Groq or GroqCloud.

## Infrastructure and partnerships

Groq has used partnerships to obtain data-center capacity and distribution. In February 2025, it announced a $1.5 billion commitment from the Kingdom of Saudi Arabia for expanded delivery of LPU-based inference infrastructure. The announcement said a GroqCloud region in Dammam was already serving traffic and described the commitment as funding for infrastructure delivery, not as an equity investment in Groq.[22]

In May 2025, Groq announced that it would be the exclusive inference provider for Bell Canada's planned Bell AI Fabric network. The announcement described a six-site target of 500 MW and an initial 7 MW Groq facility in Kamloops, British Columbia.[23] Those numbers were plans stated at announcement time; they do not establish that all six sites or the full capacity were completed.

In October 2025, IBM and Groq announced a go-to-market and technology partnership. IBM said clients could access GroqCloud through watsonx Orchestrate immediately, while support for IBM Granite models and integration work involving Red Hat's vLLM were described as plans.[24] This distinction matters because announced future integrations should not be written as completed features without later evidence.

## Name

The name Groq predates xAI's Grok chatbot. After xAI introduced Grok in 2023, Groq published a humorous post emphasizing that the organizations and products were unrelated.[7] The post was not an announcement that xAI had acquired, partnered with, or licensed technology from Groq.

## References

1. U.S. Securities and Exchange Commission. "Groq, Inc., Form D/A." December 9, 2016. https://www.sec.gov/Archives/edgar/data/1686725/000168672516000002/xslFormDX01/primary_doc.xml
2. Forbes. "AI Chip Startup Groq, Founded By Ex-Googlers, Raises $300 Million To Power Autonomous Vehicles And Data Centers." April 14, 2021. https://www.forbes.com/sites/amyfeldman/2021/04/14/ai-chip-startup-groq-founded-by-ex-googlers-raises-300-million-to-power-autonomous-vehicles-and-data-centers/
3. Jouppi, N. P., et al. "In-Datacenter Performance Analysis of a Tensor Processing Unit." Proceedings of the 44th Annual International Symposium on Computer Architecture (2017): 1-12. https://arxiv.org/abs/1704.04760
4. Abts, D., et al. "Think Fast: A Tensor Streaming Processor (TSP) for Accelerating Deep Learning Workloads." 2020 ACM/IEEE 47th Annual International Symposium on Computer Architecture: 145-158. https://doi.org/10.1109/ISCA45697.2020.00023
5. Groq. "Groq and Nvidia Enter Non-Exclusive Inference Technology Licensing Agreement to Accelerate AI Inference at Global Scale." December 24, 2025. https://groq.com/newsroom/groq-and-nvidia-enter-non-exclusive-inference-technology-licensing-agreement-to-accelerate-ai-inference-at-global-scale
6. Groq. "Groq Raises $650M to Scale Its AI Inference Cloud Business." June 22, 2026. https://groq.com/newsroom/groq-raises-usd650m-to-scale-its-ai-inference-cloud-business
7. Groq. "Welcome to Groq's Galaxy, Elon." November 9, 2023. https://groq.com/blog/welcome-to-groqs-galaxy-elon
8. Groq. "Terms of Use." Accessed July 28, 2026. https://groq.com/terms-of-use
9. Groq. "Groq Acquires Dataflow Systems Pioneer Maxeler Technologies." March 1, 2022. https://www.prnewswire.com/news-releases/groq-acquires-dataflow-systems-pioneer-maxeler-technologies-301492383.html
10. Groq. "Groq Acquires Definitive Intelligence to Launch GroqCloud." March 1, 2024. https://groq.com/newsroom/groq-acquires-definitive-intelligence-to-launch-groqcloud
11. Groq. "GroqCloud." Accessed July 28, 2026. https://groq.com/groqcloud
12. Groq. "Groq API Reference." Accessed July 28, 2026. https://console.groq.com/docs/api-reference
13. U.S. Securities and Exchange Commission. "Groq, Inc., Form D." September 5, 2018. https://www.sec.gov/Archives/edgar/data/1686725/000168672518000003/xslFormDX08/primary_doc.xml
14. Groq. "Groq Closes $300 Million Fundraise." April 14, 2021. https://groq.com/newsroom/groq-closes-300-million-fundraise
15. Groq. "Groq Raises $640M To Meet Soaring Demand for Fast AI Inference." August 5, 2024. https://groq.com/newsroom/groq-raises-640m-to-meet-soaring-demand-for-fast-ai-inference
16. Groq. "Groq Raises $750 Million as Inference Demand Surges." September 17, 2025. https://groq.com/newsroom/groq-raises-750-million-as-inference-demand-surges
17. Axios. "Nvidia deal a big win for Groq employees and investors." December 28, 2025. https://www.axios.com/2025/12/28/nvidia-groq-shareholders
18. Nvidia Corporation. "Annual Report on Form 10-K for the fiscal year ended January 25, 2026." February 25, 2026. https://www.sec.gov/Archives/edgar/data/1045810/000104581026000021/nvda-20260125.htm
19. Warren, E., and Blumenthal, R. "Letter to Nvidia regarding the Groq agreement." March 19, 2026. https://www.warren.senate.gov/wp-content/uploads/media/doc/letter_from_senators_warren_blumenthal_to_nvidia_on_groq_deal.pdf
20. Bloom Energy. "Bloom Energy Appoints Simon Edwards as Chief Financial Officer." March 26, 2026. https://www.bloomenergy.com/news/bloom-energy-appoints-simon-edwards-as-chief-financial-officer/
21. Nvidia. "Inside NVIDIA Groq 3 LPX: The Low-Latency Inference Accelerator for the NVIDIA Vera Rubin Platform." March 16, 2026. https://developer.nvidia.com/blog/inside-nvidia-groq-3-lpx-the-low-latency-inference-accelerator-for-the-nvidia-vera-rubin-platform
22. Groq. "Saudi Arabia Announces $1.5 Billion Expansion to Fuel AI-powered Economy with AI Tech Leader Groq." February 10, 2025. https://groq.com/newsroom/saudi-arabia-announces-1-5-billion-expansion-to-fuel-ai-powered-economy-with-ai-tech-leader-groq
23. Groq. "Groq Becomes Exclusive Inference Provider for Bell AI Network." May 28, 2025. https://groq.com/newsroom/groq-becomes-exclusive-inference-provider-for-bell-canadas-sovereign-ai-network
24. IBM. "IBM and Groq Partner to Accelerate Enterprise AI Deployment with Speed and Scale." October 20, 2025. https://newsroom.ibm.com/2025-10-20-ibm-and-groq-partner-to-accelerate-enterprise-ai-deployment-with-speed-and-scale