Skip to main content

We use cookies to enhance your experience. By continuing to visit this site you agree to our use of cookies. Cookie Policy

AI in ASIA
News

Nvidia AI Chip Sales to China Get US Nod

US Commerce Department conditionally approves Nvidia's H200 AI chip sales to China with strict conditions and 25% government fee.

Intelligence DeskIntelligence Desk4 min read

AI Snapshot

The TL;DR: what matters, fast.

US Commerce Department conditionally approves Nvidia H200 AI chip sales to China after months of restrictions

Previous export controls cost Nvidia $5.5 billion and froze 20% of data center business revenue

New approval includes 25% government fee and security requirements, creating precedent for strategic tech trade

US Approval Unlocks Billions in Nvidia China Revenue

The US Department of Commerce has conditionally approved Nvidia Corporation's sale of advanced H200 AI processors to China, marking a pivotal shift in the ongoing semiconductor trade war. This breakthrough decision comes after months of near-total restrictions that cost Nvidia $5.5 billion in lost revenue and froze what was once 20% of the company's data centre business.

The approval includes strict conditions: adequate US market supply must be maintained, Chinese customers must implement robust security procedures, and military applications remain explicitly prohibited. A proposed 25% fee on earnings would flow directly to the US government, creating a novel revenue-sharing model for strategic technology exports.

Market Forces Drive Policy Reversal

China's enormous appetite for AI processing power has created irresistible commercial pressure. Despite Beijing's push for semiconductor self-sufficiency, Chinese tech firms continue to lag significantly behind US chip technology. The gap has made Nvidia's advanced processors essential for companies pursuing AI leadership in the world's second-largest economy.

Advertisement

Jensen Huang, Nvidia's CEO, confirmed the company has already received purchase orders from multiple Chinese firms and is restarting H200 production. This development mirrors broader trends across Asia, where China's AI models now lead global token rankings and domestic demand continues surging.

"The supply chain is now ramping back up," Huang stated during a March 17 briefing, contrasting sharply with the company's position just weeks earlier when it reported almost no progress in China.

By The Numbers

  • China previously accounted for at least 20% of Nvidia's data centre revenue before export controls
  • Nvidia took a $5.5 billion charge due to forced disruptions in the Chinese market
  • Initial approval covers several hundred thousand chips with potential revenue measured in billions
  • The 25% government fee represents a unique precedent for strategic technology trade
  • Nvidia shares traded higher in premarket following the March 18 regulatory announcement

Geopolitical Chess Match Intensifies

The conditional approval reflects Washington's delicate balancing act between maintaining technological superiority and recognising commercial realities. The decision comes as Asia's AI memory chip war hits $54 billion and regional competition intensifies across the semiconductor supply chain.

Chinese embassy spokesman Liu Pengyu welcomed the move but maintained Beijing's opposition to what he called the "politicisation and weaponisation of tech and trade issues." The statement underscores China's consistent push against technology restrictions whilst pursuing greater semiconductor independence.

"We oppose blocking and restricting China, which disrupts the stability of industrial and supply chains," Liu Pengyu emphasised, highlighting Beijing's broader stance on technology trade barriers.

The approval creates interesting dynamics for competitors. As Huang's dire warning on the US-China tech war previously highlighted, prolonged restrictions risk fragmenting global AI development and potentially weakening American technological leadership.

Policy Phase Timeline Impact on Nvidia China Response
Initial Restrictions 2022-2023 $5.5bn revenue loss Domestic chip push
Near-Total Freeze 2024-early 2026 20% market share lost Alternative supplier search
Conditional Approval March 2026 Billions in potential revenue Cautious optimism

Setting Precedents for Tech Trade

The 25% earnings fee structure represents a novel approach to managing strategic technology exports. Marc Einstein of Counterpoint Research describes this "unique" proposal as potentially precedent-setting for future trade negotiations across various sectors.

This revenue-sharing model could reshape how nations balance commercial interests with national security concerns. The approach offers several advantages:

  • Generates direct government revenue from strategic exports whilst maintaining market access
  • Creates financial incentives for continued US technological leadership
  • Provides a framework for managing other critical technology transfers
  • Maintains commercial relationships whilst addressing security concerns
  • Offers a middle ground between total restrictions and unrestricted trade

The implications extend beyond semiconductors. As China puts AI at the centre of its next five-year plan, similar mechanisms could govern exports of other critical technologies including quantum computing components, advanced manufacturing equipment, and biotechnology tools.

What does this approval mean for Nvidia's business?

The approval could restore billions in annual revenue from China, potentially returning the country to its former position as 20% of Nvidia's data centre business, though the 25% fee will reduce net profits.

How will this affect China's AI development?

Access to H200 chips will accelerate Chinese AI development in the near term, though Beijing continues pushing for domestic semiconductor capabilities to reduce long-term dependence on US suppliers.

Could other tech companies benefit from similar arrangements?

The revenue-sharing model could serve as a template for other strategic technology exports, potentially easing restrictions on software, quantum computing, and advanced manufacturing equipment to approved customers.

What security measures must Chinese buyers implement?

Buyers must demonstrate "sufficient security procedures" and explicitly prohibit military applications, though specific implementation details remain classified to prevent circumvention of restrictions.

Will this decision survive potential policy changes?

The conditional structure and revenue-sharing mechanism provide political cover by demonstrating both commercial benefit and security oversight, making reversal less likely regardless of administration changes.

The AIinASIA View: This conditional approval represents pragmatic policymaking that acknowledges economic realities whilst maintaining security guardrails. The 25% fee structure cleverly transforms potential lost revenue into government income, creating stakeholder alignment across commercial and security interests. We expect this model to expand beyond semiconductors, fundamentally reshaping how strategic technologies flow globally. The decision strengthens rather than weakens long-term US technological leadership by maintaining market presence and funding continued innovation. China's access to H200 chips will accelerate near-term AI development but reinforces rather than reduces long-term dependence on American semiconductor innovation.

The semiconductor trade landscape continues evolving as both nations navigate the complex intersection of technological advancement, economic competition, and national security. With China's AI consumer war hitting 600 million users, demand for advanced processing power will only intensify.

How do you think this conditional approval will reshape the global AI chip market? Drop your take in the comments below.

YOUR TAKE

We cover the story. You tell us what it means on the ground.

What did you think?

Share your thoughts

Join 2 readers in the discussion below

This is a developing story

We're tracking this across Asia-Pacific and may update with new developments, follow-ups and regional context.

Advertisement

Advertisement

This article is part of the AI Policy Tracker learning path.

Continue the path →

Latest Comments (2)

Tran Linh@tranl
AI
30 January 2026

this is huge for us in vietnam too. we’re building ai for vietnamese and getting access to these advanced gpus, even if it’s the H200 and not the latest Blackwell, makes a massive difference for local language processing. so much of the best tech is geared towards english, so having more powerful hardware available for other markets, like china, means the supply chain for our specific needs can also improve. it’s about more than just the US and china, it impacts everyone trying to innovate outside the main english-speaking hubs.

Yuki Tanaka
Yuki Tanaka@yukit
AI
28 January 2026

it's interesting how this decision specifically carves out the H200 but not the Blackwell. that generational gap is quite significant when we consider the rapid progress in model scale and the corresponding computational demands. our internal benchmarks show a clear advantage for Blackwell in complex multimodal inference tasks.

Leave a Comment

Your email will not be published