Nvidia forms national security committee after Huang exit

A fractured circuit board suspended in mid-air, with one half glowing in warm gold light, the other dim and encased in translucent ice, wires fraying into smoke at the break. Background in deep indigo

Nvidia has launched a new board committee to oversee national security and export compliance, its first major move since CEO Jensen Huang stepped down amid allegations of a relationship with a Chinese national suspected of espionage. The committee will be chaired by director Suzanne Nora Johnson and includes Stephen C. Neal and John O. Dabiri. It will meet at least quarterly and receive direct reports from Nvidia’s existing national security and trade compliance teams.

A response to geopolitical fallout

The move follows intense scrutiny over Huang’s alleged ties to Huang Le (also known as Alice Huang), a Singapore-based executive at Megaspeed, a company reportedly responsible for importing $4.6–7.5 billion in Nvidia chips into Southeast Asia. US authorities suspect Megaspeed of facilitating access to restricted technology for Chinese AI firms. Huang had publicly opposed strict export controls, arguing they would accelerate China’s domestic chip development—a stance now under congressional review. He is scheduled to testify before US lawmakers later this month.

Washington has tightened export rules to prevent circumvention via third countries. The US Commerce Department announced an interim rule targeting AI chips and data centres outside the US, aimed at stopping Chinese firms from remotely accessing restricted infrastructure. This change was made in direct response to the unfolding Nvidia-related chip smuggling concerns.

Shifting power in AI hardware

Even before the scandal, Nvidia faced mounting competition. Its dominance in AI accelerators—holding 80% of sales, according to Silicon Analysis—has attracted challengers. AMD’s share has grown from 2% in 2022 to between 7% and 10% in 2026. Cloud giants Google, Amazon, and Microsoft now design in-house AI chips to reduce reliance on Nvidia and lower costs.

Startups are also gaining ground. Cerebras sells large single-die processors designed for training massive AI models. AMD recently acquired Taalas, a firm that optimises AI inference by burning model weights into silicon. Wafer, a startup using AI to port models to AMD hardware, claims it can run major open-weight models on AMD chips at least 80% as efficiently as on Nvidia’s, often within hours and with minimal human intervention.

The erosion of CUDA’s moat

Nvidia’s software advantage, particularly its CUDA platform, long served as a barrier to competition by simplifying AI development on its hardware. But AI models can now automatically rewrite code for alternative architectures. This reduces the friction of switching, threatening Nvidia’s ecosystem lock-in.

As Dan Hutchinson, a veteran chip analyst, notes, Nvidia risks entering an ‘Intel swan moment’—a reference to Intel’s 2019 leadership shift when CFO Bob Swan replaced engineering-focused CEO Brian Krzanich. That transition coincided with a decline in technical innovation, rising bureaucracy, and falling revenue. Huang’s reported personal conduct echoes Krzanich’s resignation, which also followed a relationship with an employee.

Patrick Moorhead of MoorInStrat adds that rising demand for inference and general-purpose CPUs—areas where Nvidia has limited presence—further weakens its strategic position. While Nvidia acquired Groq, a startup specialising in inference chips, in December 2025, it remains to be seen whether this will close the gap.

What it means

For UK, UAE, US, and Kazakh businesses relying on Nvidia, this moment signals increased regulatory risk and potential supply chain volatility. The new committee may tighten export practices, affecting availability in regions like Central Asia. Meanwhile, competitors’ progress suggests that multi-vendor AI infrastructure is becoming viable. Companies should assess diversification—both to mitigate geopolitical exposure and to leverage cost gains from alternative silicon.

Nvidia’s revenue rose from $26.91 billion in January 2022 to $215.94 billion in January 2026, fuelled by AI demand. Yet its dependence on China has waned: sales there now make up just 8% of revenue, down from 20% in 2024. Huawei’s advances in AI computing and the rise of Chinese open-weight models could further erode Nvidia’s global position if adopted abroad. While Chinese firms like Z.ai report training models entirely on Huawei chips, the scalability of these alternatives outside China remains unproven.

Sources

Leave a Reply

Your email address will not be published. Required fields are marked *