Nvidia just pulled off what might be the tech industry’s most expensive talent raid. The chip giant struck a “non-exclusive licensing deal” with Groq, an AI inference startup, that looks suspiciously like an acquisition in disguise. Groq’s founder Jonathan Ross and its president are joining Nvidia, along with key engineers. While neither company disclosed the price tag, CNBC reported it at $20 billion, nearly triple Groq’s $6.9 billion valuation from just a few months ago.
So what’s the big deal?
Here’s the thing: Nvidia makes the world’s best AI chips. Its GPUs power everything from ChatGPT to the most advanced AI labs. The company’s data center revenues hit $51 billion last quarter alone. It’s the undisputed heavyweight champion of AI hardware.
So why would the champion suddenly need to license technology from a startup?
The inference problem
To understand this, you need to know that AI has two distinct phases. First is training—where you feed massive amounts of data to create an AI model. This is like teaching a student everything they need to know. Nvidia’s GPUs excel here.
Then comes inference, the everyday work of actually using that trained model. Every time you ask ChatGPT a question or generate an image, that’s inference. And here’s the catch: inference is becoming the bigger business, but Nvidia’s chips aren’t perfectly designed for it.
Think of it this way: Nvidia’s GPUs are like Swiss Army knives, brilliant for gaming, data centers, AI training, and inference. But Groq’s Language Processing Units (LPUs) are like precision scalpels built for just one job: inference. They’re faster, cheaper to run, and consume less power.
Ross, who previously helped create Google’s tensor processing units (TPUs), designed Groq’s chips with embedded memory. This architectural choice means they can process AI queries at lightning speed without guzzling electricity like GPUs do.
Why this matters now
The AI industry is hitting an inflection point. Training massive models is expensive, but it’s a one-time cost. Inference, however, happens millions of times daily as users interact with AI services. As AI adoption explodes, inference costs are becoming the real bottleneck.
Groq’s chips address this perfectly. They’re cheaper to produce, faster to deploy, and more energy-efficient. In an era where data centers are struggling with power constraints, this matters enormously.
But Groq had a problem: scaling. Building chips is one thing; manufacturing millions of them and integrating them into the global AI infrastructure is another. That’s where Nvidia’s relationships with TSMC, its established supply chains, and its customer base become invaluable.
The acquihire trend
This deal follows a pattern we’ve seen repeatedly. Microsoft hired Inflection AI’s founders through a licensing deal. Meta spent $14 billion to bring in Scale AI’s CEO. Google did similar deals with Character.AI. These “acquihires” let tech giants snap up talent and technology without lengthy regulatory battles.
The structure is clever: Groq continues operating independently with a new CEO, its GroqCloud service keeps running, and Nvidia gets a “non-exclusive” license. But let’s be honest, when a company’s founder, president, and key engineers leave for the licensing partner, paying triple the startup’s valuation, it’s essentially an acquisition with extra steps.
The bottom line
Nvidia’s Groq deal is actually a sign of strength, not weakness. The company is acknowledging that as AI evolves from training to inference, it needs specialized tools. Rather than trying to build everything in-house, it’s acquiring the best inference technology available and integrating it into its ecosystem.
For Nvidia, spending $20 billion to cement its position in the inference market is pocket change compared to losing dominance in the fastest-growing segment of AI. The company generated $67 billion in revenue last quarter, it’ll earn back this investment quickly.
The real message? Even the AI chip king knows that in a rapidly changing landscape, admitting you need help is better than pretending you’re perfect at everything.
Sonia Boolchandani is a seasoned financial writer She has written for prominent firms like Vested Finance, and Finology, where she has crafted content that simplifies complex financial concepts for diverse audiences.
Disclosure: The writer and her his dependents do not hold the stocks discussed in this article.
The website managers, its employee(s), and contributors/writers/authors of articles have or may have an outstanding buy or sell position or holding in the securities, options on securities or other related investments of issuers and/or companies discussed therein. The content of the articles and the interpretation of data are solely the personal views of the contributors/ writers/authors. Investors must make their own investment decisions based on their specific objectives, resources and only after consulting such independent advisors as may be necessary.