Jensen Huang Resigns as Nvidia CEO Amid Smuggling Probe Questions

Jensen Huang has resigned as president and chief executive officer of Nvidia, giving up $108 million in share compensation and leaving the company he co-founded in April 1993. Chief financial officer Colette Kress, the company's longtime finance chief, steps in as interim CEO. The departure follows revelations of two affairs: one with an executive at Singapore-based Megaspeed International, who may be a Chinese government operative, and another with an executive inside Nvidia's own senior ranks.

The commercial stakes reach beyond the boardroom. According to The Information, Megaspeed and its affiliates have acquired $7.5 billion in GPU systems since 2023. A sister company of Megaspeed bought $4.8 billion in Nvidia GPU gear between June 2024 and February 2026. It is not clear whether Huang knew of the relationship when those purchases occurred.

What do regulators actually know?

The US Commerce Department had been investigating Megaspeed since late 2024, according to a timeline assembled by the New York Times. Those investigations reportedly stalled as President Trump and Huang grew close last year. The probes have since moved to the US Department of Justice, and their current state is unclear. Chatter persists that Megaspeed channelled Nvidia-based GPU systems, or remote access to them, to Chinese companies including model maker DeepSeek.

Two legal questions remain unresolved. First, whether Huang could face indictment for complicity in smuggling Inspur-made systems into China. Second, whether renting capacity on controlled systems to customers such as Alibaba constitutes a violation at all. The latter is not clearly illegal today. Alibaba builds its own AI models and would plainly benefit from access to GPUs that export controls forbid selling to Chinese organizations. Where Megaspeed obtained the money for GPU purchases and datacenter construction in Indonesia and China is also unknown, though a trail back into China is a reasonable inference.

The new Congress sworn in during January 2027 is expected to pursue the matter, with a possible presidential pardon for Huang should an indictment follow.

Who runs Nvidia now?

Harvey Jones, a former CEO of chip design software maker Synopsys and managing partner at venture firm Square Wave Ventures, becomes chairman of the board — the first chairman in Nvidia's history. Jones has sat on the board since November 1993. The board has started a CEO search, and Kress is expected to remain CFO once a permanent chief executive arrives.

Huang leaves with his finances intact. The GenAI boom made him the 15th richest person in the world, with $145 billion in Nvidia stock before this week's revelations. His 3.6 percent stake has taken losses. He is likely to attempt a quiet period repairing relationships with his wife and children, though the world is unlikely to let him stay quiet.

Can Nvidia run without him?

Huang famously ran a tight ship. Sixty top executives reported to him directly, so he knew exactly what each of them was doing and why. That structure is in place, and those five dozen executives know their jobs. The AI system roadmaps for the "Rubin" and "Feynman" GPU generations extend the company's path to 2030. Nvidia's production is sold out for 2027 already, and history plus rising demand suggest 2028 will follow. If there was ever a moment when Nvidia could largely run itself, it is now.

Yet Nvidia has never faced pressure from so many directions at once. Hyperscalers, cloud builders and AI model builders are all developing custom CPUs and XPU accelerators to run their own models at roughly half the cost of Nvidia's hardware and software stack. Nvidia is also championing open source and open weight AI models — a stance only it can afford, given the premium its compute and networking hardware commands. The open Nemotron models compete directly with the closed models of Nvidia's largest GPU customers, which has likely irritated them.

What lies beyond the sold-out years?

The rosy AI spending forecasts for 2029 and 2030 and beyond may not hold. Supply chain constraints for many components, and the difficulty of adding electricity capacity to grids worldwide, both threaten those projections. In the United States, dissatisfaction is rising over the preferential treatment datacenter operators have enjoyed for a decade and a half. Residential electric bills have climbed as datacenters proliferated, and public concern about GenAI's economic and cultural effects is growing.

Huang's exit also strips the AI industry of its most coherent and eloquent advocate. Nvidia must now test whether its sold-out order books and locked roadmaps can carry it through a future that was already far less certain before its co-founder was compelled to leave.