Nvidia has officially closed a massive $20 billion licensing deal with the artificial intelligence (AI) startup Groq. The strategic move gives the "Team Green" giant access to deep technical knowledge and key personnel who could otherwise have become dangerous competitors down the road. Although on legal paper this is not a merger, in practice it bears the clear hallmarks of a quiet takeover designed to avoid close scrutiny from antitrust regulators.
Nvidia's "Hackquisition" Playbook and the Push for Key Talent
When news of the $20 billion deal was confirmed on Christmas Eve (December 24), many assumed it was the largest acquisition in Nvidia's history. The company was quick to correct that, saying it was only a "non-exclusive licensing agreement" for Groq's inference technology. For Nvidia, which just reported profits of $32 billion last quarter, the $20 billion outlay underscores how strategically important it is to defend its lead against growing pressure from emerging chip architectures outside its GPU ecosystem.
The deal fits squarely into the "hackquisition" pattern that has become popular in Silicon Valley: a transaction that is not legally an acquisition but works like one in practice. Deals of this kind typically combine large cash payments, intellectual property licensing, and selective hiring of senior leaders. Microsoft, Google, Amazon, and Meta have all used this approach to gain technology without drawing a regulatory "yellow card." In this case, Nvidia's real targets appear to be CEO Jonathan Ross, president Sunny Madra, and the core engineers specializing in ultra-efficient AI inference chips.
Inference Technology and the Threat From TPUs
Groq built its reputation on a distinctive computing architecture designed for low-latency inference at scale. The company's first chips drew mixed reactions, with some analysts calling them a breakthrough and others doubting they could scale, but Nvidia's licensing decision suggests it sees real potential in folding Groq's design into its future products.
The people matter even more. CEO Jonathan Ross helped develop Google's tensor processing unit (TPU), the chip family that lets Google run large-scale AI models without depending on Nvidia GPUs. That experience makes him one of the few people with direct expertise in designing hardware that counterbalances Nvidia. The success of the TPU has forced Nvidia to guard against a future in which AI workloads shift from general-purpose GPUs to specialized inference hardware, where energy efficiency and speed are decisive. Bringing Ross's team "under its wing" lets Nvidia neutralize an independent rival and add valuable engineering brainpower.
Sidestepping Antitrust Rules to Neutralize a Rival
The structure of the deal carries consequences as important as the technology itself. A direct acquisition would almost certainly have triggered intense antitrust investigation, given Nvidia's dominant position in the market. By choosing a licensing arrangement, Nvidia achieves its goal of control without changing formal ownership. Critics describe these as "functional acquisitions," which let large corporations consolidate power and drain a rival of its talent without a merger, leaving the targeted company "hollowed out," much like the deal between Meta and Scale AI.
Announcing the deal over a holiday may limit immediate public attention, but antitrust concerns are unlikely to disappear. Nvidia's agreement, however cleverly built for speed and discretion, could still face long-term scrutiny if regulators conclude it is quietly stifling healthy competition in the AI hardware market.
