Nvidia is set to launch a modified version of its H20 artificial intelligence chip for the Chinese mainland by July, according to three sources familiar with the matter. The adjusted design follows tightened U.S. export restrictions blocking sales of the original model, which had previously been approved for Chinese customers.
The downgraded H20 chip will feature reduced memory capacity and revised technical specifications to comply with U.S. trade rules. Despite performance limitations, sources indicate Chinese cloud providers and tech firms might reconfigure modules to optimize functionality—a strategy reflecting the region’s determination to sustain AI development.
With China accounting for 13% of Nvidia’s global revenue ($17 billion in FY2024), the market remains critical for the semiconductor giant. CEO Jensen Huang’s recent Beijing visit underscored this priority, occurring shortly after U.S. authorities imposed new licensing requirements for advanced chip exports.
Major Chinese tech firms including Tencent, Alibaba, and ByteDance reportedly accelerated H20 orders early this year amid surging demand for affordable AI solutions. Nvidia had secured $18 billion in pre-modified H20 orders by May, highlighting the region’s appetite for AI infrastructure despite geopolitical hurdles.
The U.S. has maintained semiconductor export limits since 2022, citing national security concerns. Nvidia’s adaptive approach highlights the growing complexity of balancing regulatory compliance with commercial interests in one of the world’s most consequential technology markets.
