Nvidia Corp. is targeting a return to the Chinese market before the Lunar New Year with its H200 chips, but multiple regulatory and competitive hurdles threaten its prospects of regaining lost ground in what was once its most important market.

The Santa Clara-based chipmaker plans to export between 40,000 to 80,000 H200 chips to China before February 17, primarily from existing inventory, according to Reuters citing people familiar with the matter. Nvidia has also arranged increased H200 production to supply the Chinese market in the second quarter of 2025. The company has not responded to requests for comment.

However, significant uncertainty remains over whether these exports will materialize. Chinese authorities have not approved any H200-related procurement plans, while U.S. regulators continue reviewing export licenses amid political opposition. The backdrop includes lingering security concerns after China's Cyberspace Administration raised issues about H20 chip vulnerabilities in July, which Nvidia has denied.

The potential revival comes after Nvidia's China business collapsed from a 95% market share to "zero" in artificial intelligence chips following export restrictions. CEO Jensen Huang's ability to rebuild that position now depends on navigating complex regulatory approval processes, intense domestic competition, and pricing pressures from mandated 25% revenue sharing with the U.S. government.

Regulatory Approval Process Remains Uncertain

The U.S. Commerce Department has sent H200 export license applications to the State Department, Energy Department, and Defense Department for review, a process expected to take no more than 30 days. President Donald Trump, who announced December 8 that he would allow H200 sales to China provided Nvidia pays 25% of revenues to the U.S. government, retains final decision-making authority.

Opposition persists within Congress and government agencies. Democratic Senator Elizabeth Warren and Representative Gregory Meeks have demanded the Commerce Department disclose more licensing information to assess whether approved chips could be used for military purposes. Some U.S. officials have reportedly requested Nvidia install tracking devices in exported chips, potentially triggering renewed security concerns.

Chinese authorities have signaled openness to advanced AI chip imports. Foreign Ministry spokesperson Guo Jiakun stated the government noticed reports about H200 exports and "has always advocated that China and the United States achieve mutual benefit and win-win results through cooperation." AMD CEO Lisa Su visited China last week, meeting with customers and officials in what appears to be preparations for re-entering the market.

Tech Giants Likely to Form Initial Customer Base

Alibaba, ByteDance , and Tencent are expected to form the primary customer base for H200 chips. These internet technology giants face intense competition in large language model training and AI infrastructure development, creating substantial demand for advanced chips.

China's smart computing centers and government technology projects have rapidly completed domestic substitution for AI chips, particularly in inference applications where domestic chips have demonstrated verified performance. Policy support and cost considerations favor domestic alternatives in these segments.

Reports indicate some Chinese companies have smuggled Nvidia chips through Japan and Singapore or directly deployed chips overseas to train their models, demonstrating strong market demand. Major technology companies previously submitted detailed procurement plans to regulators when H20 chips received export approval.

Revenue Impact Remains Limited Despite Sales Resumption

First-batch exports could generate between 1 billion to 4 billion in sales, based on market prices of US$20,000 to 40,000 per card and accounting for the 25% U.S. government levy. This compares to just $50 million in China data center revenue during Nvidia's fiscal third quarter, when no large procurement orders materialized.

Nvidia's fourth quarter guidance (November 2025 to January 2026) excludes any China data center revenue. In fiscal 2024, China generated $17.1 billion in revenue for Nvidia. Multiple institutions estimate China's AI chip market grew approximately 30% year-over-year in 2025 during Nvidia's absence.

By 2025, Nvidia has transitioned to Blackwell series chips as its mainstream product in the U.S. market, with H200 production capacity reduced and pricing subject to volatility, making Chinese market pricing particularly difficult to determine.

Domestic Competitors Narrow Performance Gap

H200 maintains advantages over domestic chips in large language model training scenarios, particularly in capacity and bandwidth specifications. More importantly, Chinese large language models were predominantly developed on Nvidia's GPU architecture and deeply embedded in the CUDA ecosystem, making H200 adoption cost-effective compared to migration expenses.

Domestic chip performance continues improving rapidly. On December 20, Moore Threads, known as "China's first GPU stock," unveiled its new "Chunxiao" GPU architecture with significantly enhanced computing density and energy efficiency.

Moore Threads claims its new "Huashan" AI chip, based on the architecture, comprehensively surpasses Nvidia's Hopper series in floating-point computing power, memory bandwidth, memory capacity, and high-speed interconnect bandwidth. Founder and CEO Zhang Jianzhong, formerly Nvidia's global vice president and general manager for Greater China, stated: "We can now confidently say that if you previously trained on Hopper, using our new product to train large language models will only perform better, not worse."

The company is aggressively building its MUSA ecosystem to challenge and potentially replace Nvidia's CUDA ecosystem. Huang has acknowledged that exporting downgraded AI chips to China is unwise, stating he is "no longer sure whether China will accept H200."

AMD Moves Aggressively to Capture Market Share

While H200 attracts attention, AMD and Intel are also targeting the Chinese market under the same export rules announced by Trump. As the world's second-largest GPU company, AMD derives 24% of total revenue from China, though its business there collapsed to minimal levels starting in the second quarter of 2025.

Su revealed in November that the company's MI308 AI chip has received U.S. government export licenses and AMD is prepared to pay taxes on China exports. Following Trump's policy announcement, AMD accelerated its China engagement.

On December 11, Su met with Chinese Ambassador to the U.S. Xie Feng to discuss AMD's operations in China and bilateral economic cooperation. On December 16, Su led AMD's executive team to China, first visiting partner Lenovo headquarters in Beijing.

On December 17, Minister of Industry and Information Technology Jin Leicheng met with Su to discuss cooperation in digital economy and artificial intelligence. Jin stated China will provide more cooperation opportunities for foreign companies including AMD and hopes AMD continues investing deeply in the Chinese market. On December 18, Commerce Minister Wang Wentao met with Su to discuss the company's China business development and strengthening cooperation.

Based on publicly available information, Su appears to have achieved faster progress and more positive regulatory feedback than Huang. AMD may secure first-mover advantage in returning to the Chinese market.