Nvidia Shifts Focus from China as AI Chip Demand Soars Elsewhere

Nvidia stops making H200 AI chips for China, reallocating TSMC capacity to its next-gen Vera Rubin platform as China data center revenue hits zero.

Nvidia has ceased manufacturing its H200 artificial intelligence chips for the Chinese market, reallocating those production resources at its primary supplier, Taiwan Semiconductor Manufacturing Company (TSMC), toward its upcoming Vera Rubin platform. This strategic pivot comes despite interim export approvals from Washington and underscores a market where sales to China have effectively ground to zero—a reality the company's current planning does not anticipate reversing.

China Sales Vanish Despite Export Permits

The decision, reported by the Financial Times, marks a significant operational shift. Although the Trump administration granted a license in January allowing exports of the more powerful H200 to China, no meaningful sales materialized. In its latest quarterly report, Nvidia explicitly stated it expects zero data center revenue from China for the current quarter.

Even the purposefully downgraded H20 export model for China played "no significant role" in the third and fourth quarters of fiscal year 2026. The company currently holds an inventory of approximately 250,000 H200 chips, a buffer sufficient to respond to potential market changes without immediately restarting dedicated production lines.

Vera Rubin Platform Takes Priority

The reallocation of H200 capacity coincides with the production ramp-up for Nvidia's next-generation Vera Rubin architecture. This new platform, encompassing six distinct chips, promises to slash the cost per token for AI inference by a factor of ten compared to the current Blackwell generation. Major cloud service providers, including Amazon Web Services, Google Cloud, Microsoft Azure, and Oracle Cloud, are among the first customers. A broad market release is scheduled for the second half of 2026.

Global appetite for Nvidia's latest AI accelerators remains voracious, with cloud providers and enterprises in the United States, Europe, and other open markets absorbing every available unit.

Fiscal 2026 Delivers Record-Breaking Performance

Nvidia closed its fiscal year 2026 (ended January 25, 2026) with staggering results. Fourth-quarter revenue reached $68.1 billion, a 73% year-over-year increase. For the full year, the chipmaker generated $215.9 billion in total revenue, representing growth of 65%.

The data center segment was again the primary engine, posting $62.3 billion in Q4 and $193.7 billion for the full year—a 68% annual gain. Bolstered by a gross margin of 75%, net income for the quarter stood at $43 billion, bringing the annual net profit to $120 billion.

Looking ahead, Nvidia has provided revenue guidance of roughly $78 billion for the first quarter of fiscal 2027, again forecasting no contribution from its Chinese data center business.

Upcoming GTC Conference in Focus

The next potential catalyst for the company is the GTC 2026 developer conference, scheduled for March 16-19 in San Jose. CEO Jensen Huang has teased plans to "present a chip that will surprise people," leading investors to anticipate details on the next-generation architecture and possible updates on the demand outlook.

The event will feature over 1,000 sessions covering AI data centers, robotics, digital twins, and scientific computing. Market observers will also be watching for any developments from the U.S.-China summit set for late March, which could potentially impact the stalled trade dynamic between the two nations.