AI DOESN’T BUY COMPANIES ANYMORE. IT BUYS THE PEOPLE WHO MAKE THEM VALUABLE.
Nvidia–Groq may be more than a $20bn legal dispute. It may expose a fundamental change in what an acquisition means in the age of artificial intelligence.
There is an old way to acquire a technology company.
You buy the company.
You acquire its shares or assets, assume its liabilities, integrate its employees, obtain its intellectual property and submit the transaction to the corporate, tax and competition architecture developed over decades to govern mergers and acquisitions.
Artificial intelligence is beginning to produce something different.
The buyer may not need the company.
It may need the people.
The technology.
The intellectual property.
The knowledge embedded inside a relatively small number of engineers.
And perhaps the fastest route to those assets is no longer to acquire the corporation that contains them.
The Nvidia–Groq transaction now being challenged in Delaware provides an unusually clear case study.
According to the Financial Times, Nvidia’s 2025 arrangement with Groq involved approximately $20bn: a $17bn licensing payment whose proceeds were distributed among Groq’s backers and a separate $3bn Nvidia stock bonus pool for selected engineers who moved to Nvidia, including founder Jonathan Ross.
The lawsuit brought by two former Groq engineers alleges something more consequential: that Nvidia effectively obtained Groq’s core technology and hired nearly all of its engineers — potentially around 200 people — while Groq itself formally survived. Nvidia characterised the original technology agreement as non-exclusive. The allegations in the lawsuit remain allegations and have not been judicially established.
That distinction matters.
Because economically, the transaction raises a fascinating question:
What exactly is a company in the AI economy?
For an industrial business, much of enterprise value may reside in factories, distribution networks, physical assets, inventories, customer relationships and operating infrastructure.