Nvidia’s $20B Groq Deal Faces Lawsuit From Former Engineers

Nvidia’s roughly $20 billion transaction with AI chip startup Groq is facing a legal challenge from two former Groq engineers who allege that the unusual deal unfairly treated some shareholders while transferring the startup’s technology and much of its engineering workforce to Nvidia. The complaint, filed in Delaware, argues that Groq’s board failed to secure the best possible value for all shareholders and structured the transaction without giving certain investors the voting rights they were entitled to, according to the lawsuit.

The dispute puts a new spotlight on an increasingly common Silicon Valley transaction structure: companies can obtain valuable technology and talent through licensing agreements and hiring arrangements without formally acquiring the target company. In Groq’s case, Nvidia received a non-exclusive license to Groq’s inference technology while Groq founder and CEO Jonathan Ross, President Sunny Madra and other employees joined Nvidia. Groq itself remained independent and has since repositioned its business around AI inference cloud services.

Key takeaways

• Two former Groq engineers and shareholders, Benjamin Serebrin and Joshua Rubin, filed a lawsuit in Delaware challenging the transaction.
• The plaintiffs allege that Groq’s board failed to maximize value for common shareholders and improperly structured the transaction.
• The lawsuit characterizes approximately $20 billion of consideration as roughly $17 billion for the technology license and $3 billion in Nvidia stock awards for selected employees.
• Nvidia is not named as a defendant in the Delaware shareholder lawsuit.
• Groq has called the lawsuit meritless and said the transaction delivered exceptional value to investors and employees.
• Nvidia has already commercialized Groq technology through the NVIDIA Groq 3 LPX inference accelerator.
• The transaction is also facing broader regulatory scrutiny over whether licensing and acqui-hire structures can effectively function like acquisitions without traditional merger review.

What happened between Nvidia and Groq?

Nvidia and Groq announced their agreement on December 24, 2025.

The companies described it as a non-exclusive licensing agreement covering Groq’s inference technology. Groq said at the time that its founder Jonathan Ross, President Sunny Madra and other members of the team would join Nvidia to help advance and scale the licensed technology.

Importantly, Groq did not disappear as a corporate entity. The company said it would continue operating independently under new CEO Simon Edwards, while its GroqCloud business would continue operating.

That structure was unusual because the transaction delivered several things Nvidia wanted from Groq without being presented as a conventional acquisition of the startup.

Nvidia obtained access to Groq’s technology, while many of the people who developed that technology moved to Nvidia. At the same time, Groq remained a separate business.

The financial value reported for the transaction was about $20 billion. Groq did not initially disclose a dollar figure in its announcement, but subsequent reporting described the transaction as a roughly $20 billion arrangement.

Nvidia later disclosed more detail in its annual report. The company said it entered a non-exclusive license agreement with Groq for its language-processing-unit technology and hired certain Groq employees. Nvidia said it did not purchase Groq’s customer contracts, existing products or equity interests.

For accounting purposes, Nvidia recorded $14.4 billion of goodwill and a $2.5 billion developed-technology intangible asset. The company said total consideration consisted of $13 billion paid at closing and $4 billion, including imputed interest, payable within one year.

That accounting disclosure is particularly important because it provides a primary-source view of how Nvidia itself recorded the transaction, while the $20 billion figure commonly used in media reports reflects the broader reported deal value.

Why are former Groq engineers suing?

The lawsuit was filed by former Groq employees who retained equity in the company.

Benjamin Serebrin and Joshua Rubin allege that Groq’s board improperly transferred valuable assets and talent to Nvidia while leaving some shareholders with an underpriced stake in the remaining company.

The complaint was filed in Delaware’s Court of Chancery and was unsealed on October 5. The case is identified as Serebrin v. Ross, Delaware Chancery case No. 2026-1291.

The central argument is not simply that the $20 billion transaction was too small.

Instead, the plaintiffs challenge how the value was divided and how the transaction was approved.

They allege that the board gave Nvidia access to Groq’s most valuable assets — its technology and engineering workforce — while the remaining shareholders were left with a much less valuable company.

The lawsuit also alleges conflicts of interest involving members of Groq’s board and argues that the directors did not adequately pursue alternative transaction structures or attempt to maximize value for all shareholders.

Those claims remain allegations. No court has determined that Groq’s directors breached their fiduciary duties.

The $17 billion license and $3 billion employee pool

One of the most important parts of the lawsuit concerns how the reported $20 billion transaction was allocated.

According to the complaint and reporting on the case, approximately $17 billion was attributed to the technology licensing component, while about $3 billion was allocated through Nvidia stock awards for selected Groq employees who joined Nvidia.

The plaintiffs argue that this structure disadvantaged common shareholders.

Their contention is that employees who moved to Nvidia received benefits connected to the technology and their employment at Nvidia, while shareholders who did not participate in the Nvidia transition were effectively cashed out without receiving the full potential value associated with Groq’s technology, talent and future relationship with Nvidia.

The lawsuit further argues that Groq’s technology could have generated additional value through Nvidia’s ability to combine it with its enormous GPU business and distribution network.

That potential future upside and strategic synergy is a major part of the plaintiffs’ argument.

In other words, the dispute is not only about what Groq was worth when the transaction was signed. It is also about whether the board properly captured the value Nvidia was willing to pay for access to Groq’s technology and talent.

Why the deal was structured this way

The transaction reflects a broader shift in how major technology companies acquire AI capabilities.

Traditional acquisitions generally involve buying a company’s shares or assets and obtaining control of the business.

An acqui-hire is different. A company can instead obtain key employees while arranging separate transactions for technology or intellectual property.

This can be attractive when a buyer wants the people and technology but does not necessarily want to purchase the entire company.

Nvidia’s Groq transaction was particularly notable because Groq had developed specialized hardware for AI inference — the process of running trained AI models to generate responses.

Nvidia has historically been dominant in AI training hardware, but inference has become an increasingly important battleground as AI applications generate enormous volumes of real-time queries.

Reuters reported when the deal was announced that Nvidia’s agreement with Groq involved a non-exclusive technology license and the hiring of Ross, Madra and other Groq personnel. Reuters also noted that Groq’s technology focused on inference, where Nvidia faced competition from startups and established chipmakers.

The structure allowed Nvidia to bring important Groq technology and talent into its ecosystem while leaving Groq itself operational.

Nvidia is already using Groq technology

The legal dispute comes after Nvidia has already turned Groq’s technology into a commercial product.

In March 2026, Nvidia introduced the NVIDIA Groq 3 LPX, an inference accelerator based on Groq’s architecture and designed to work alongside Nvidia’s Vera Rubin platform.

Nvidia describes Groq 3 LPX as a rack-scale accelerator designed for low-latency AI inference.

The architecture combines Nvidia GPUs with Groq-derived language processing units to handle latency-sensitive workloads.

Nvidia says the system uses 256 interconnected LPU accelerators and is designed for highly responsive inference, including workloads involving agentic AI systems.

By August, Nvidia said Groq 3 LPX had entered full production, with AI cloud provider Nebius identified as its first adopter. Nvidia positioned the technology as an important component of infrastructure for agentic AI, where systems can generate large numbers of tokens across many inference steps.

That commercial progress adds another dimension to the shareholder lawsuit.

The plaintiffs argue that the value of Groq’s technology and its potential synergies with Nvidia were not properly reflected in what shareholders ultimately received.

Groq survived — but it became a very different company

The transaction did not eliminate Groq.

Instead, the startup shifted its business model.

Groq subsequently raised $350 million in a Series A round announced in August 2026, valuing the company at approximately $3.5 billion. Nvidia was expected to participate in the financing. Groq said the new funding would support expansion of its global AI inference infrastructure.

Groq said it operates 13 data centers across North America, Europe, the Middle East and Asia-Pacific and plans to substantially expand its infrastructure.

The company’s post-deal strategy is therefore different from its original identity as an independent AI-chip challenger.

Instead of focusing primarily on building and selling its own chip architecture, Groq is now positioning itself as an AI inference cloud provider.

This transformation is central to the lawsuit’s argument.

The plaintiffs contend that shareholders effectively lost the upside from Groq’s original technology business, while the company that remained behind was subsequently valued at $3.5 billion.

Whether that comparison proves damages in court is another question.

The $3.5 billion valuation came from a later financing transaction under different circumstances and therefore cannot automatically be treated as evidence of what Groq’s earlier shareholders should have received.

The board-conflict argument

The plaintiffs also challenge the role of investment funds connected to Groq’s board.

The lawsuit alleges that several funds with board representation had interests that were not fully aligned with all shareholders.

According to reporting on the complaint, the plaintiffs point to BlackRock, Social Capital, Infinitum and Disruptive as funds whose relationships with the remaining Groq business created potential conflicts. The funds themselves are not defendants in the lawsuit.

The argument is significant because Delaware corporate law places substantial emphasis on directors’ fiduciary obligations when they oversee transactions involving conflicts of interest or a sale of corporate assets.

The plaintiffs argue that the board should have taken greater steps to establish that the transaction produced the best available outcome for shareholders.

However, the lawsuit itself acknowledges an important legal uncertainty: there is no directly controlling Delaware precedent clearly establishing that an arrangement of this precise kind must be treated like a traditional change-of-control transaction.

That makes the case more complicated than a conventional claim that a company was simply sold too cheaply.

Nvidia is not the main defendant in the shareholder case

A critical distinction is that Nvidia is not named as a defendant in the Delaware shareholder lawsuit.

The claims are directed at Groq’s former leadership and board over their handling of the transaction.

That does not mean Nvidia is irrelevant.

The plaintiffs’ theory depends heavily on the proposition that Nvidia received extremely valuable technology and human capital through the transaction and that Groq’s board failed to negotiate or structure the deal in a way that maximized shareholder value.

But proving that theory requires establishing wrongdoing by the Groq directors, not simply demonstrating that Nvidia benefited from the arrangement.

Groq has rejected the allegations.

A company spokesperson said the licensing agreement delivered exceptional value for Groq, its investors and employees and described the lawsuit as meritless. Groq said it would vigorously defend itself.

Nvidia did not immediately provide a comment to reporters about the lawsuit.

The deal was already under regulatory scrutiny

The shareholder lawsuit is not the first controversy surrounding Nvidia’s arrangement with Groq.

The U.S. Department of Justice has reportedly been examining the transaction over antitrust concerns.

Bloomberg Law reported in September that the DOJ was investigating whether Nvidia structured the roughly $20 billion transaction as a licensing agreement and hiring arrangement in a way that avoided traditional merger review.

The issue matters because Nvidia has become one of the most powerful companies in the AI hardware market.

A conventional acquisition of a promising AI-chip competitor could receive substantially more regulatory attention than a technology license accompanied by the hiring of the target’s leadership and engineers.

That does not automatically make the Groq transaction unlawful.

The DOJ investigation itself does not establish wrongdoing, and there has been no finding that the transaction violated antitrust law.

But the combination of regulatory scrutiny and shareholder litigation demonstrates how closely regulators and investors are beginning to examine alternative deal structures in the AI industry.

Why acqui-hires are becoming a bigger issue

The Groq case comes amid a wider wave of transactions in which major technology companies have recruited AI researchers, executives and engineering teams through structures that stop short of conventional acquisitions.

The appeal is straightforward.

AI companies are often valued heavily on their talent, intellectual property and technical capabilities rather than on traditional assets such as factories or inventories.

Buying the entire company can trigger lengthy regulatory review.

Hiring its most important employees and licensing its technology can potentially deliver many of the same strategic benefits without formally taking control of the company.

That is why regulators and lawmakers have become increasingly interested in whether such structures effectively function as acquisitions.

The Nvidia-Groq transaction is particularly significant because it combined three elements: a major technology license, the movement of senior leadership and the transfer of a substantial engineering workforce.

The lawsuit now adds another question: even if such a structure is legally possible, what obligations does a startup’s board owe to shareholders when arranging it?

What the lawsuit could mean for AI deals

The outcome could matter beyond Groq.

If Delaware courts ultimately accept the plaintiffs’ theory, boards of venture-backed technology companies could face greater scrutiny when negotiating transactions that transfer a company’s core technology and employees without formally selling the company.

That could make future acqui-hire structures more expensive or more complicated.

Boards might need to conduct broader sale processes, seek competing offers, provide shareholders with greater information or obtain formal approval in situations that previously might have been treated as ordinary licensing and employment arrangements.

On the other hand, if Groq successfully defeats the claims, companies may have greater confidence in using similar structures.

That would be significant for Big Tech companies competing for AI talent.

The distinction between an acquisition and an acqui-hire is becoming increasingly important because the value of an AI startup can be concentrated in a relatively small number of engineers and a handful of proprietary technologies.

The bigger picture

The Groq lawsuit is ultimately about more than one $20 billion transaction.

It highlights the tension between the enormous strategic value of AI technology and the increasingly creative structures companies are using to obtain it.

For Nvidia, Groq’s technology has already become strategically useful. The company has incorporated the architecture into Groq 3 LPX, giving Nvidia another tool for the rapidly growing inference market.

For Groq, the transaction provided a large infusion of capital and a path into AI infrastructure after the loss of much of its original chip engineering organization. The company is now building an inference cloud business rather than simply trying to compete with Nvidia as a standalone chip designer.

For shareholders, however, the question is whether the value transferred to Nvidia was properly priced and whether the board acted in their best interests.

Those questions will now be tested in Delaware court.

Looking Ahead

The immediate focus will be on how Groq’s former directors respond to the complaint and whether the Delaware Court of Chancery allows the plaintiffs’ fiduciary-duty claims to proceed. The case could also clarify how Delaware corporate law applies to transactions that combine technology licensing, employee transfers and the effective hollowing out of a startup without a formal acquisition.

The broader industry impact may be even more important. As AI companies become valuable primarily because of their models, chips, intellectual property and specialized teams, traditional definitions of mergers and acquisitions are becoming harder to apply. The Groq case could help establish whether Silicon Valley’s new generation of “acqui-hire” transactions can remain outside conventional M&A rules or whether courts and regulators will increasingly treat them as acquisitions in everything but name.