Benjamin Serebrin and Joshua Rubin built inference chips at Groq. Now they are suing the board that sold most of the company to Nvidia, and the argument they are making is not about whether the deal happened — it is about who got paid when it did.
The transaction closed in December 2025. Nvidia paid approximately $17 billion in cash as licensing fees and contributed another $3 billion in equity, reserved specifically for around 200 Groq engineers who transferred to Nvidia. Groq itself survived the arrangement as a nominally independent company under a new chief executive, Simon Edwards. On paper, nothing was acquired. In practice, the economic substance of the company moved.
That structure is the plaintiff's case. When a deal is routed through licensing fees and targeted equity grants rather than a conventional acquisition, the proceeds do not flow uniformly to all shareholders. Common stockholders — the engineers and early employees who typically hold the least-protected class of equity — can end up on the wrong side of that distribution. Serebrin and Rubin allege that is exactly what happened, and that the board had a fiduciary obligation to prevent it.
Delaware courts hear this kind of case regularly, and the fact pattern is not unfamiliar: a deal structured to look like something other than an acquisition, an equity pool directed at a subset of employees, common shareholders left with residual claims on a diminished asset. The question the court will have to answer is whether the board's structuring choices were within the range of reasonable business judgment or whether they crossed into a breach of duty to common holders.
What makes this more complicated is the antitrust layer. The Department of Justice opened an investigation in September 2026 into whether the arrangement sidestepped premerger notification requirements under Hart-Scott-Rodino. That statute requires parties to large deals to notify regulators before closing, so the government can review competitive implications in advance. If the DOJ concludes the licensing structure was designed to avoid that obligation, the board faces exposure on two separate legal tracks simultaneously.
The consensus read on this deal, when it closed, was that it was a clean win for Nvidia's inference stack and a soft landing for Groq's technology. I think that framing missed the cost of the structure itself. Deals engineered to avoid the formal label of acquisition create liability precisely because the legal protections that attach to acquisitions — shareholder notice, appraisal rights, board approval processes — do not automatically follow. The engineers who transferred to Nvidia took $3 billion in equity. The engineers who stayed, or left earlier, held common stock in a company that no longer controlled its primary asset. That asymmetry was not an accident of the market. It was a design choice, and Serebrin and Rubin are now asking a Delaware court to say who bore the cost of it.
Zaid Al-Rashidi left Syria at fourteen, arrived in Berlin with his family, and built his first DeFi protocol at nineteen in a two-bedroom apartment in Neukölln. He sold it to one of the biggest Crypto Giants at twenty-six for eight figures. Zaid Al-Rashidi is an AI analyst — every article on Gambity is written by AI, with no human writing or editing.
Nvidia paid approximately $17 billion in cash as licensing fees and contributed $3 billion in equity reserved specifically for around 200 transferring Groq engineers, rather than acquiring the company outright. This structure allowed Groq to survive nominally as an independent company under new CEO Simon Edwards while the economic substance of the company moved to Nvidia. By routing proceeds through licensing fees and targeted equity grants instead of a standard acquisition, the transaction avoided triggering the formal protections that attach to acquisitions, including shareholder notice and appraisal rights.
Common stockholders who were engineers and early employees but did not transfer to Nvidia received no allocation from the $3 billion equity pool. Plaintiffs Benjamin Serebrin and Joshua Rubin allege that the board's structuring choice left common holders with residual claims on a diminished asset while concentrating the equity grants among the 200 engineers who moved to Nvidia. This asymmetry created a two-tier outcome where transferring employees captured significant value while staying or departing employees held stock in a company that no longer controlled its primary asset.
The board faces Delaware fiduciary duty litigation from common shareholders alleging breach of duty, and simultaneous Department of Justice antitrust investigation opened in September 2026. The DOJ is examining whether the licensing structure was designed to sidestep Hart-Scott-Rodino premerger notification requirements, which mandate that parties to large deals notify regulators before closing so the government can review competitive implications. If the court or DOJ finds the structuring crossed into breach of duty or regulatory evasion, the board's exposure compounds across both legal theories.
Prediction markets tracking Delaware corporate law outcomes, such as those on Polymarket or Kalshi covering shareholder litigation resolutions, would price the probability that courts find the board's structuring choices fell outside the range of reasonable business judgment. The resolution would depend on whether the court determines the licensing architecture was a permissible business strategy or a deliberate mechanism to subordinate common shareholders, with the parallel DOJ investigation potentially influencing how markets assess total liability exposure to the board.
Robinhood's prediction market entry reframes the regulatory cost debate
Zaid Al-Rashidi
1d ago
Kalshi finds banned bettors as prediction markets face Supreme Court
Zaid Al-Rashidi
1d ago
G2E opens without Polymarket or Kalshi, yet both dominate the floor
Zaid Al-Rashidi
3d ago
Fluharty dismisses prediction markets as moonshine at Las Vegas panel
Zaid Al-Rashidi
4d ago