Nvidia hit with $5.5 bn charge as US clamps down on China chip exports

Nvidia to take $5.5 billion charge as US export curbs hit H20 AI chip sales to China; company warns of mounting challenges in key overseas markets

Nvidia announced on Tuesday that it expects to incur a quarterly charge of approximately $5.5 billion due to constraints on exporting its H20 graphics processing units to China and select other regions. Following the announcement, the company’s shares declined over 6 per cent in after-hours trading.

 

According to a regulatory filing, the US government informed Nvidia on April 9 that the company must obtain a licence to ship its H20 chips to China and several additional countries.

 

The latest export hurdle marks one of the clearest indications yet that Nvidia’s rapid growth trajectory may be hampered by increasing geopolitical restrictions. The US has maintained that these high-performance chips have the potential to power military-grade supercomputers. 

 

 

Nvidia is scheduled to release its financial first-quarter results on May 28.

 

H20 chip designed to comply with earlier curbs

 

The H20 was engineered specifically for the Chinese market to comply with earlier US export restrictions imposed in 2022 and updated in 2023. Despite the limitations, the chip is estimated to have brought in between $12 billion and $15 billion in revenue last year.

 

During Nvidia’s last quarterly earnings call in February, CEO Jensen Huang noted that revenue from China had halved compared to levels before the US imposed export controls. Huang also highlighted increasing competition in the Chinese market, with Nvidia identifying Huawei as a key rival in consecutive annual filings.

 

China still a key market for Nvidia

 

Despite the drop in revenue, China remains Nvidia’s fourth-largest market after the US, Singapore, and Taiwan. In the financial year ending in January, over half of Nvidia’s total revenue came from US-based clients, the company’s annual report stated.

 

H20 vs H100 and H200: Technological differences

 

Although based on the Hopper AI architecture introduced in 2022, the H20 chip features slower bandwidth and interconnection speeds compared to the H100 and H200 chips available in Western markets. Nvidia is now shifting its focus to its latest Blackwell-series AI chips.

 

Chinese AI firm DeepSeek used the H20 chip to develop its R1 model, which made waves earlier this year with its advanced capabilities. The use of Nvidia’s hardware underscores the significance of the H20 in the competitive AI research landscape.

 

Beyond current export limitations, Nvidia is bracing for another wave of controls set to begin next month under proposed "AI diffusion rules". These rules could further narrow the company’s ability to sell advanced AI hardware globally.

 

Nvidia has consistently argued that tighter restrictions may undermine the US’s position in the global tech race by stifling competition. In response to past curbs, the company has already moved some of its operations — like testing and distribution — out of China.

 

Export licence requirement extended indefinitely

 

The US government notified Nvidia on Monday that the licensing requirement for H20 chip exports will remain in place “for the indefinite future,” the company said in its filing.

 

Following the announcement, Nvidia shares have dropped 16 per cent so far this year, influenced in part by US President Donald Trump’s tariff proposals targeting top trading partners. Although some electronics, including chips and smartphones, received temporary tariff exemptions, officials signalled more sector-specific tariffs could be coming.

 

(With Reuters inputs)