According to Reuters reporting today, the US Commerce Department will begin issuing licenses to NVDIA to export its H20 AI ACCENTER chips to China. NVIDIA filed necessary paperwork after the recent assurances of federal regulators.
Reuters cited sources familiar with the matter, which said the company expects it to cover the H20 delivery, which is a modified version of its advanced AI chips designed to comply with US export controls that prevent GPU’s most powerful models from reaching Chinese consumers.
Although NVIDIA has not publicly disclosed shipment quantity or exact timelines, analysts estimate that China has participated about 20-25 % of the company’s data center’s income before the latest sanctions. The move could help recover a portion of NVIDIA estimated $ 10-15 billion when Washington was in danger of tightening the rules in 2022.
Jensen Huang, CEO of Navalia in Competex in Taiwan.
News of Navalia shares increased by 4.2 % in the morning trade, and the rally semiconductor colleagues and AI infrastructure companies were extended.
Although some investors are rapidly, other market viewers noted that delays in shipment, obstacles to compliance, or the US -China policy change may increase. Regulatory clearance does not guarantee the immediate completion of orders, and any increase in commercial stress can disrupt exports once again.
Planned rehabilitation is an upside -down backdrop that aims to keep advanced AI chips away from China on national security concerns. The policy shift has already been criticized in Washington, with many lawmakers questioning the decision.
Last week, Sean Jim Bank (R in) and San Elizabeth Warren (D -MA) sent a letter to New Divine CEO Jensen Huang before a visit to China on July 11. He emphasized that he should refrain from negotiating with the Chinese military or any Chinese companies affiliated with intelligence, and warned that such engagement “could legalize companies that cooperate closely with the Chinese military or discuss the difference between exploitation in US export controls.”
The resumption of H20 shipment is expected to help Chinese hypertensive cloud providers and research institutes maintain access to modern GPU infrastructure for the training of large -scale AI models. Although the H20 NVIDIA flagship is less powerful than H100, it is important to maintain demand in the data center bloodout and AI workload that rely on CUDA acceleration.
Re -export reports have already sparked a rotation among Chinese firms to secure the supply of chips, which is the most capable model legally available in China, despite constant performance limits by the US administration.
The ongoing tension between trade AI growth and implementing export controls is designed to limit China’s modern semiconductor capabilities.
This article first appeared on our sister’s publication, HPCWIRE.
Relevant
