(C) Tip Ranks
 

SANTA CLARA / BEIJING — NVIDIA is reportedly aiming to commence shipments of its H200 artificial intelligence chips to China by mid-February 2025, according to sources cited by Reuters. The plan marks a significant milestone in the shifting landscape of global semiconductor trade following recent policy adjustments by the U.S. administration.

The California-based chip giant intends to meet initial demand by utilizing existing inventory. The first phase of shipments is expected to involve between 5,000 and 10,000 chip modules, which translates to approximately 40,000 to 80,000 individual H200 units. NVIDIA has also reportedly informed Chinese clients of plans to expand production capacity, with a target to begin accepting new orders for the H200 starting in the second quarter of next year.

This strategic move follows an announcement by U.S. President Donald Trump on December 8, which permitted the export of H200 chips to China under the condition of a 25% fee. This decision represents a stark departure from the Biden administration’s stringent export controls that prioritized national security by barring China from accessing advanced AI hardware.

Despite being a predecessor to the current "Blackwell" generation, the H200 remains a highly coveted asset in the AI sector due to its performance within the "Hopper" architecture. While global supply for the H200 has been constrained as NVIDIA pivots toward Blackwell and the future Rubin lineup, the demand within China remains high. Major Chinese tech conglomerates, including Alibaba Group and ByteDance, have expressed keen interest in the H200, which offers nearly six times the performance of the lower-spec H20 chips previously available in the region.

However, the final execution of this export plan rests in the hands of the Chinese government. Beijing is currently weighing the benefits of acquiring advanced foreign technology against the potential risks to its burgeoning domestic semiconductor industry. There are concerns within China that allowing the influx of high-performance NVIDIA chips could slow the development of local alternatives.

To mitigate this, Chinese authorities are reportedly considering a "mixed purchase" mandate, which would require domestic firms to procure a specific ratio of Chinese-made AI chips alongside any H200 imports. Sources emphasize that until official approval is granted by Beijing, the delivery timeline and total volume remain subject to change.

"The entire operation is contingent upon the Chinese government's green light," a source noted. "Nothing is set in stone until the formal permits are issued." As the mid-February deadline approaches, the global tech industry remains watchful to see if this trade opening will successfully bridge the gap between U.S. commercial interests and China’s technological ambitions.