Analysis of Jensen Huang's visit to China and H20 chip restrictions
I'm PortAI, I can summarize articles.
The U.S. government has indefinitely banned the export of its H20 chips to China, citing national security risks due to their potential supercomputing applications. As a result, NVIDIA is expected to record up to $5.5 billion in expenses for the first fiscal quarter ending April 27, covering H20 inventory, procurement commitments, and related reserves. This leads to a $5.5 billion loss for NVIDIA (inventory impairment + procurement commitment default reserves), accounting for 12.8% of its Q1 revenue forecast for fiscal year 2026. As the main product in China (accounting for 85% of revenue in China), the H20 was expected to contribute $12-15 billion in revenue in 2024. Post-ban, the annual shortfall could reach $18 billion. The contraction of the Chinese market (from 25% to 17% of global revenue) and the blockade of Singapore's re-export channels (77% of orders suspected to flow to China) have sharply increased supply chain disruption risks. Although the H20 is a "castrated" chip, its high memory bandwidth (4.0TB/s) and CUDA ecosystem still support China's AI inference needs. Huawei's Ascend 910B/C clusters perform close to the H100, and companies like ByteDance have increased their procurement of domestic chips to 60%. Under policy-driven substitution, NVIDIA's market share is rapidly eroding. Additionally, China's customs data shows a 10.2% drop in chip imports in Q1 2024 but an 18% increase in semiconductor equipment imports, signaling deeper restructuring of the domestic supply chain.
What signals is Jensen Huang sending with this China visit? Testing the waters for technology licensing and localized production, NVIDIA aims to maintain its presence in China by optimizing compliant product lines (e.g., the rumored H30 chip) and exploring technology licensing models (e.g., CUDA ecosystem partnerships). NVIDIA may follow Tesla's "Shanghai Gigafactory" model to promote localized production of Blackwell chips in China. The Blackwell Ultra chip (liquid cooling + CoWoS-L packaging) is expected to enter mass production in the second half of 2025, with FP8 computing power 47 times higher than the H20. If yield rates exceed 85%, it could offset losses in the Chinese market. For NVIDIA, despite constrained demand in China, new growth drivers include sovereign AI (nearly $1 billion in annual revenue), Tesla's autonomous driving (35,000 H100 clusters), and Spectrum-X Ethernet networks (100,000 GPU clusters). Global AI computing demand is projected to grow by 43% in 2025, with NVIDIA still holding an 80% market share.
Short-term bearish factors remain unresolved technically: the stock price has fallen 32% year-to-date, hitting a low of $86.62, with short interest still high. Policy black swans (e.g., additional restrictions on A800/L40S chips) could trigger a second dip. Investors should watch for key milestones in June, including Blackwell's mass production progress and U.S.-China tariff negotiations (90-day buffer period). If the H20 license receives "limited exemptions," it may trigger an oversold rebound.
Short-term policy risks remain high; a wait-and-see approach is advised. If Q1 earnings confirm the $5.5 billion impairment ceiling and free cash flow turns positive, consider buying the dip for a technical rebound.
The copyright of this article belongs to the original author/organization.
The views expressed herein are solely those of the author and do not reflect the stance of the platform. The content is intended for investment reference purposes only and shall not be considered as investment advice. Please contact us if you have any questions or suggestions regarding the content services provided by the platform.
