Groq Reportedly Raising $650M to Scale Inference Cloud After Nvidia’s $20B Tech Deal

AI chip startup Groq is reportedly seeking $650 million in new funding from existing investors, according to sources who spoke with Axios. The capital raise comes as the company pivots its focus toward scaling its inference neocloud business, which relies on its custom-designed chips and systems to power AI applications after they have been trained.

Context: The Nvidia ‘Not-a-Qui-Hire’

In December 2025, Groq struck an unusual agreement with Nvidia valued at approximately $20 billion. While not a full acquisition, the deal involved the departure of several senior Groq employees to Nvidia and the licensing of Groq’s hardware technology to the chip giant. For Groq’s investors, the arrangement functioned as a lucrative payout — one that would have marked Nvidia’s largest purchase had it been a complete buyout.

Now, those same investors are being asked to reinvest. The new funding round is intended to fuel Groq’s expansion of its inference cloud platform, which allows developers and enterprises to host applications that require rapid processing after an AI prompt is submitted.

Why Inference Matters Now

Inference — the computational work that happens after a model generates a response — has become a far larger market than model training in the current AI landscape. As companies deploy more AI-powered tools in real-time environments, the demand for low-latency, efficient inference hardware has surged. Groq’s architecture, known for its speed and deterministic performance, positions it to compete with established players like Nvidia in this growing segment.

Leadership and Investor Backing

Groq is currently led by interim CEO Adam Winter and CFO Matt Eng. According to Axios, the company’s existing backers Disruptive and Infinitium have agreed to backstop the round, committing to purchase any remaining shares if other investors decline their pro-rata allocations. This structure suggests confidence among lead investors despite the broader market’s cautious stance on AI hardware startups.

Conclusion

The reported $650 million raise signals that Groq’s investors see long-term value in its inference-focused strategy, even after the company’s unusual partial exit to Nvidia. As the AI industry shifts from training massive models to deploying them at scale, Groq’s specialized hardware and cloud services could become increasingly critical. The outcome of this funding round will likely influence the competitive dynamics of the inference chip market in the coming year.

FAQs

Q1: What is Groq’s inference neocloud business?
Groq’s neocloud is a cloud platform that allows developers to run AI inference workloads on Groq’s custom-designed chips, which are optimized for speed and low latency.

Q2: Why did Nvidia pay $20 billion for a ‘not-a-qui-hire’ deal with Groq?
The deal allowed Nvidia to license Groq’s hardware technology and hire key senior employees without a full acquisition, giving Nvidia access to Groq’s architecture while Groq’s investors received a substantial payout.

Q3: How does inference differ from AI model training?
Training involves teaching a model using large datasets, while inference is the process of using a trained model to generate predictions or responses in real time. Inference demand is growing faster as AI applications are deployed at scale.