Nvidia to Launch New AI Chipset for China Amid US Export Curbs

Nvidia To Launch New AI Chipset For China Amid US Export Curbs

Nvidia's new Blackwell-based AI chip is tailored to comply with US export rules and regain market share in China’s $50 billion data center industry.

Nvidia is preparing to launch a new artificial intelligence chipset for the Chinese market, designed to comply with US export controls while offering a more affordable alternative to its previously restricted H20 model. According to sources, the upcoming GPU—part of Nvidia’s latest Blackwell architecture—will be priced between $6,500 and $8,000, significantly lower than the $10,000 to $12,000 price tag of the H20.

The lower cost is attributed to the chip’s reduced specifications and simplified manufacturing process. Unlike the H20, the new processor will use conventional GDDR7 memory instead of High-Bandwidth Memory (HBM) and will not require advanced packaging technologies from Taiwan Semiconductor Manufacturing Co. (TSMC).

Despite diminished computing power, the new chip is expected to help Nvidia stay competitive in China’s AI and data center sectors, especially after US restrictions caused its market share in China to decline from 95% before 2022 to roughly 50% today. Nvidia’s main rival in the region, Huawei, has rapidly expanded its presence with the Ascend 910B chip.

CEO Jensen Huang recently warned that continued export curbs could further shift Chinese demand toward domestic alternatives like Huawei. In response to the April ban on the H20, Nvidia is also developing another Blackwell-based chip tailored for China, scheduled to enter production as early as September.

The regulatory limitations, including a cap on GPU memory bandwidth at 1.7 to 1.8 terabytes per second, have reshaped Nvidia’s product strategy. The new GPU, expected to reach around 1.7 TB/s using GDDR7, is believed to fall just within the export control limits.

Nvidia's earlier reliance on the Hopper architecture is no longer viable under current US regulations. The H20 ban forced the company to write off $5.5 billion in inventory and walk away from an estimated $15 billion in potential sales.

Still, Nvidia's competitive edge remains in its CUDA programming platform, widely used by developers to build AI models. While its new chip may not meet the highest-end AI training needs, it underscores Nvidia’s strategic pivot to retain a foothold in one of the world’s most critical AI markets.