Late last year, the United States tightened export controls on AI chips to China, prompting NVIDIA to halt shipments of its high-performance H100 and A100 processors.
In response, the company designed the H20 for export to China, a cut-down version of the H100, optimised for AI inference and cloud computing. This version was intended to meet US regulations, but faced licensing delays.
By early 2025, Chinese firms reported severe shortages, and NVIDIA warned of a potential $5.5 billion hit to its bottom line.
But now, the tides have shifted.
NVIDIA is set to resume shipments of the H20 chips to China, following assurances from the US government that license approvals are underway. CEO Jensen Huang, fresh from diplomatic visits to Washington and Beijing, confirmed that deliveries will begin soon.
The announcement marks a pivotal moment in the global chip wars, with NVIDIA positioning itself not just as a technology supplier but as a bridge between two rival superpowers.
Balancing Act in the Chip Wars
The resumption of H20 sales marks a critical moment in the ongoing AI technology tug-of-war. While the US has relaxed its grip slightly, it still maintains export control over advanced GPUs. NVIDIA’s announcement comes after months of trade tensions, smuggling reports, and product redesigns.
The company also announced the RTX PRO GPU, a China-specific chip engineered to comply with US regulations. The RTX PRO joins the H20 and other variants, designed to maintain NVIDIA’s presence in China while adhering to legal boundaries.
In a blog post published this week, NVIDIA wrote, “We’re building the world’s most advanced AI platforms and bringing them to every country. Our job is to enable developers everywhere.”
Huang’s recent trips included meetings with US lawmakers and Chinese officials, signalling NVIDIA’s effort to walk a diplomatic tightrope. Not just them, but AMD has also reportedly entered this game by announcing that it will soon resume shipping its MI308 AI chips to China.
AMD stated that the US Commerce Department intends to restart reviewing its license applications for sending these products to the world’s second-largest economy. While NVIDIA said it has been “assured” by the government that licenses to sell its AI chips will be granted.
Malaysia Closes the Loophole
While the US shifts towards granting limited chip export licences, Malaysia is moving in the opposite direction. The country’s investment, trade, and industry ministry (MITI) announced on July 14 that it will now require a 30-day advance notice and government-issued permits for the re-export of AI chips made with US technology.
“This initiative serves to close regulatory gaps while Malaysia undertakes further review on the inclusion of high-performance AI chips of US origin into the SIL of the STA 2010,” the ministry said in the release.
This decision follows investigations suggesting that Chinese companies have been acquiring NVIDIA chips through indirect routes. These include transhipment hubs such as Malaysia and Vietnam, as well as smuggling methods involving modified goods, seafood shipments, and prosthetics.
MITI said the new rules “align with Malaysia’s commitment to fair trade, and ensure that Malaysian ports are not exploited for unauthorised transactions.”
A $50 Billion High-Wire Act
NVIDIA is under pressure from multiple sides. On one hand, China remains a significant revenue generator, representing a sizeable share of the company’s AI chip market. On the other hand, US restrictions have dented its ability to sell high-performance models in the region.
According to previous reports, NVIDIA could have faced a $5.5 billion charge related to the halted sales of H20 chips to China. In addition to launching compliant alternatives like the H20 and RTX PRO, NVIDIA has also considered offering lower-cost versions of its flagship chips, specifically for the Chinese market.
These efforts aim to maintain market access while adhering to legal constraints. Just like NVIDIA, AMD had also previously indicated that it would incur $800 million in charges due to these controls on its MI308 AI chip.
Meanwhile, major Chinese firms such as DeepSeek have already experienced delays in product launches due to the H20 chip shortage earlier this year. The resumption of supply could accelerate AI development cycles in the region.
This could impact the country’s standing in the global open-source AI race, promoting the development of additional models, such as DeepSeek. In fact, a new contender is already emerging–Kimi K2 from Moonshot AI.
Kimi K2 is now drawing serious attention from AI insiders and outperforming some of the biggest names in the game. Its speed and expectations on benchmarks are sparking comparisons to DeepSeek’s breakout moment. It’s among the top 10 models this week, according to OpenRouter, an AI gateway for developers. It’s ahead of Grok 4 and GPT-4.1 models.
AI’s New Global Supply Routes
The broader impact of these developments goes beyond NVIDIA’s balance sheet. This moment reflects a shift in how countries, corporations, and regulators view the global flow of AI technology.
Export rules are evolving. Smuggling networks are being exposed. And firms like NVIDIA are becoming not just tech suppliers, but key players in international policy.
Anthropic, an AI research company, has recently warned of creative efforts to circumvent restrictions on GPU sales to China. These include disassembling and hiding processors inside unrelated products, a tactic known as “component laundering.”
The US Department of Commerce has increased enforcement actions, while companies operating in Southeast Asia now face greater compliance obligations.
In this landscape, NVIDIA is trying to maintain a role that keeps it in favour on both sides of the Pacific. With AI adoption growing rapidly, it remains unclear how long such a balancing act will be possible.
The post NVIDIA Reunites US and China With H20 Chips appeared first on Analytics India Magazine.
