Northwind.
← Makers directory
Est. 2016 · USA

Groq

"LPU inference engines."

Founders
Jonathan Ross and Douglas Wightman
A die photo of Groq’s initial Language Processing Unit (LPU), showcasing its intricate design.
A die photo of Groq’s initial Language Processing Unit (LPU), showcasing its intricate design.
Chapter 01

The story

Groq was founded in 2016 by a group of former Google engineers, primarily Jonathan Ross, who was a lead designer of Google's Tensor Processing Unit (TPU). He was joined by entrepreneur Douglas Wightman, Groq's first CEO. Their motivation was to create a specialized AI accelerator, initially known as the Tensor Streaming Processor (TSP), to tackle the increasing demands of AI workloads.

The company received initial seed funding from Social Capital in 2017 with a $10 million investment, quickly followed by additional funding. This early support allowed them to develop their core technology, focusing on a unique architecture that could efficiently handle AI inference tasks. Their breakthrough lay in designing a deterministic processor that could achieve high performance and energy efficiency by meticulously controlling hardware components.

Groq's first product, the TSP, was a 14nm chip with a computational density of over 1 TeraOp/s per square mm. After the rise of large language models like ChatGPT, they rebranded their core architecture to the Language Processing Unit (LPU) to better reflect its primary application. This strategic pivot positioned Groq to address the burgeoning market for LLM acceleration.

Chapter 02

Challenges along the way

One significant challenge for Groq has been competing in the highly competitive AI chip market dominated by giants like Nvidia. While their technology offers impressive speed for AI inference, gaining market share and proving scalability against established players required substantial funding and strategic partnerships. The company rapidly raised significant capital, including becoming a "unicorn" after a Series C round in 2021 and a $640 million Series D in 2024, demonstrating continuous effort to fuel its growth and expand its infrastructure in the global AI landscape.

Chapter 03

Major milestones

  1. 2016
    Groq was founded by former Google engineers, led by Jonathan Ross and Douglas Wightman.
  2. 2016
    United States PatentFig. 32
    Deterministic tensor-streaming processor filings
    Assignee search
    Read filing ↗
  3. 2017
    Received initial seed funding of $10 million from Social Capital.
  4. 2021
    Raised $300 million in a Series C round, valuing Groq at over $1 billion and making it a unicorn.
  5. 2022
    Acquired Maxeler Technologies to enhance its dataflow systems expertise.
  6. 2023
    Selected Samsung to manufacture its next-generation 4nm LPUs, marking Samsung's first order at their Taylor, Texas fab.
  7. 2024
    Launched GroqCloud™ developer platform and acquired Definitive Intelligence.
  8. 2024
    Raised $640 million in a Series D round, valuing the company at $2.8 billion.
  9. 2025
    Secured a $1.5 billion commitment from Saudi Arabia for LPU-based AI inference infrastructure.
Chapter 04

Prototypes & people

A die photo of Groq’s initial Language Processing Unit (LPU), showcasing its intricate design.
productA die photo of Groq’s initial Language Processing Unit (LPU), showcasing its intricate design.
The official Groq logo, representing the company's commitment to AI acceleration.
logoThe official Groq logo, representing the company's commitment to AI acceleration.
On the shelf

Currently stocked

View all →
Sourced from Wikipedia. Images © their respective owners.