整理 | 华卫
Yesterday, AI chip startup Groq confirmed that it has completed a new round of financing amounting to $750 million, with a post-investment valuation of $6.9 billion (approximately 49 billion RMB). This round of financing was led by investment firm Disruptive, with participation from institutions such as BlackRock, Neuberger Berman, and Deutsche Telekom Capital Partners, as well as existing investors like Samsung, Cisco, D1, and Altimeter who also increased their investments.
This figure exceeds the rumored data leaked during the financing news in July. At that time, it was reported that Groq planned to raise about $600 million, with a valuation close to $6 billion.
In addition to developing chips, Groq also provides data center computing power services. The company had previously completed a $640 million financing round in August 2024, at which time its valuation was $2.8 billion—indicating that its valuation has more than doubled in about a year. According to PitchBook (a globally recognized private equity and venture capital data platform), Groq has raised over $3 billion to date.
The core reason Groq is favored by capital is its attempt to break NVIDIA’s (the giant in the AI chip field) monopoly in the tech industry. Notably, Groq’s chips are not mainstream graphics processing units (GPUs) that support AI systems, but rather specialized chips named “Language Processing Units (LPUs)”; their hardware products are defined as “Inference Engines”—computers optimized for the rapid and efficient operation of AI models.
Groq’s products primarily target developers and enterprise customers, offering two service forms: cloud services and local hardware clusters. The local hardware consists of a server cabinet containing multiple sets of “hardware-software integrated nodes.” Both cloud and local hardware support running various mainstream open-source AI models, such as those developed by Meta, DeepSeek, Qwen, Mistral (a French AI company), Google, and OpenAI. Groq claims that its products not only ensure but also enhance AI performance in certain scenarios, while being significantly cheaper than competing products.
Groq’s founder, Jonathan Ross, has a strong technical background in this field, which provides important support for the company’s development. Ross previously worked at Google, where he was involved in the development of Google’s Tensor Processing Unit (TPU)—a specialized processor designed for machine learning tasks that was officially released in 2016, the same year Groq ended its “stealth mode” and made its debut. Today, the TPU remains the core computing power support for Google Cloud AI services.
According to Groq, its AI chips and computing power services currently support over 2 million developers’ AI applications; a year ago, this number was only 356,000.
Reference link:
https://techcrunch.com/2025/09/17/nvidia-ai-chip-challenger-groq-raises-even-more-than-expected-hits-6-9b-valuation/
Disclaimer: This article is a translation by InfoQ and may not be reproduced without permission.
Click the bottomto read the original textVisit the InfoQ official website for more exciting content!
Today’s Recommended Articles7 hours of continuous reconstruction without disconnection! The once unrivaled Claude finally meets a rival: Greg Brockman personally interprets major breakthroughs in AI programming81-year-old boss makes $100 billion to become the world’s richest while the company lays off! Old employees joke: “We have been replaced by GPUs”Apple doesn’t hype AI, but succeeds instead?Just now, lithography machine giant ASML was revealed to become the largest shareholder of Mistral! Netizens angrily criticize: ASML’s money will go down the drain