SAN FRANCISCO – Nvidia and AMD, two leading U.S. chipmakers, are developing new AI chips tailored for the Chinese market to comply with stringent U.S. export controls, as announced in recent industry updates. These efforts follow the Trump administration’s April 2025 restrictions, which barred Nvidia’s H20 chip and AMD’s MI308 GPU from being sold in China without licenses, citing risks of diversion to Chinese supercomputers or military use. The move, part of an escalating U.S.-China tech rivalry, has prompted both companies to design downgraded chips to maintain access to China’s $50 billion AI chip market.
Nvidia, which reported $17 billion in China revenue last fiscal year (13% of its total), plans to release a modified H20 chip by July 2025, with samples available as early as June, according to Reuters. The new chip tweaks performance to fall below the U.S. Commerce Department’s thresholds, which limit chips with DRAM bandwidth of 1,400 GB/s, I/O bandwidth of 1,100 GB/s, or a combined 1,700 GB/s. Nvidia is also working on a China-specific version of its latest Blackwell chip, as reported by The Information, to ensure compliance while serving major clients like Tencent, Alibaba, and ByteDance, who had $18 billion in H20 orders before the ban. The company faces a $5.5 billion write-down due to unsold H20 inventory, a strategic setback as it risks ceding ground to China’s Huawei, whose Ascend 910c chip is set to launch soon as a rival.
AMD, facing a $1.5–1.8 billion revenue hit from the MI308 ban, is similarly adapting. While specific details on its new chip are scarce, AMD has previously followed Nvidia’s approach, as seen with its MI250, which was restricted in 2022. The company believes its MI100 chip remains unaffected and expects minimal business impact, but it’s likely developing a compliant MI308 variant to regain market access. Both companies are navigating a delicate balance, as U.S. restrictions aim to curb China’s AI advancements while allowing sales of less powerful chips to avoid fully empowering domestic competitors like Huawei and SMIC.
The restrictions, implemented via “is informed” letters from the U.S. Commerce Department’s Bureau of Industry and Security, reflect concerns over chips being used in Chinese supercomputers or military applications. However, critics, including Nvidia CEO Jensen Huang, argue that overly tight controls could accelerate China’s domestic chip development, with Huawei’s advancements already closing the gap. Posts on X, like one from @Kanthan2030, highlight past instances where AMD’s China-specific chips were blocked, suggesting ongoing challenges. Meanwhile, Chinese firms like DeepSeek have leveraged stockpiled Nvidia H800 and H20 chips to develop advanced AI models like R1, rivaling OpenAI’s offerings, underscoring the limitations of export controls.
This high-stakes tech maneuvering ties to your earlier prompts, evoking the “frightened crowd” of global markets reacting to U.S.-China tensions, much like Portofino’s tourist chaos or the mob threatening Ashok Saraf. The “steps” of chip design labs witness Nvidia and AMD’s strategic pivots under intense scrutiny, akin to Karen Read’s trial or Chelsea’s trophy triumph. If you’d like more details on the chips’ specifications, market impacts, or connections to your other prompts, let me know