Nvidia deepens AI hardware lead With Groq deal and high-profile talent hire

Nvidia

Nvidia has moved to strengthen its grip on the AI hardware market by signing a non-exclusive licensing agreement with AI chip challenger Groq, while also bringing the company’s top leadership and key employees into its own ranks. The deal signals Nvidia’s intent to absorb alternative approaches to AI acceleration without formally acquiring the rival company.

As part of the agreement, Nvidia will hire Groq founder Jonathan Ross, Groq president Sunny Madra, along with several other employees. While CNBC reported that Nvidia is acquiring assets from Groq in a deal valued at around $20 billion, Nvidia has pushed back on the idea that this is an outright acquisition. The company told TechCrunch that Groq itself is not being acquired and declined to comment on the full scope or valuation of the arrangement.

Even so, if the reported figure is close to accurate, this would represent Nvidia’s largest deal to date. More importantly, it would give Nvidia access to a competing AI chip architecture that has been positioned as a potential alternative to traditional GPUs.

Groq has been developing a processor it calls an LPU, or language processing unit. Unlike GPUs, which are designed to handle a wide variety of workloads, LPUs are purpose-built for running large language models. Groq has claimed that its chips can execute LLMs at up to ten times the speed of conventional approaches while using roughly one-tenth the energy. Those claims have helped the company stand out at a time when efficiency and inference speed are becoming just as important as raw compute power.

Jonathan Ross is closely associated with that line of thinking. Before founding Groq, he worked at Google, where he played a key role in the development of the TPU, one of the earliest custom accelerators designed specifically for machine learning workloads. His move to Nvidia suggests the company is keen not only on Groq’s technology, but also on the expertise behind it.

The timing is significant. As competition in AI intensifies, demand for specialised computing hardware continues to surge. Nvidia’s GPUs have become the default choice for training and deploying AI models across the industry, but challengers like Groq have been pitching more focused alternatives, particularly for inference-heavy workloads. By licensing Groq’s technology rather than eliminating it through acquisition, Nvidia gains insight into that approach while keeping its options open.

Groq’s rapid growth helps explain Nvidia’s interest. In September, the company raised $750 million at a valuation of $6.9 billion. It has also reported a sharp increase in adoption, saying its hardware now powers AI applications for more than two million developers, up from roughly 356,000 the year before. That level of traction, combined with a differentiated chip design, has made Groq one of the more credible challengers in a market dominated by Nvidia.