Nvidia Continues to Bet on China
On April 16, Nvidia's stock price closed down nearly 7%, wiping out more than $148 billion in market value. The company said that restricting exports of the AI chip H20 customized for the Chinese market will result in an additional cost of $5.5 billion; there are reasons to worry that the Trump administration's latest regulations mean a "cut-off supply" of H20 to China.
On the same day, rival AMD's stock also fell by 5.8%. The company said that Washington's latest export restrictions on products such as the MI308 chip would cause losses of approximately $800 million. In addition, other AI-related chip stocks such as ARM, Broadcom, and Micron also fell between 2.5% and 4.6%.
Nvidia CEO Jensen Huang unexpectedly visited Beijing on April 17. According to official Chinese media reports, he held talks with Ren Hongbin, president of the China Council for the Promotion of International Trade. Huang expressed during the meeting that China is a very important market for Nvidia, hoping to continue cooperating with China.
This is Huang's second visit to Beijing in three months. He had similarly expressed a desire to continue cooperation with China on January 19 (the day before Trump took office).
According to a report by Yicai Global, Huang mentioned the latest export controls targeting H20 in his talks with Ren Hongbin, stating that the strengthening of such controls has already significantly impacted Nvidia's business. "Nvidia will continue to spare no effort in optimizing its product lineup to comply with regulatory requirements and remain steadfastly committed to serving the Chinese market."
As early as during President Biden's term, the US had already restricted Nvidia from exporting the most advanced AI chips to the Chinese market. The "China-specific" H20 chip was born at the end of 2023 under this backdrop, with slightly inferior performance compared to the current most advanced chips. At the beginning of this year, a Chinese startup, DeepSeek, released a new version of the AI large model, specially optimized for the H20 chip, claiming that it achieved performance comparable to the latest generation of OpenAI's GPT models from the US with extremely low training costs.
A Reuters report in late February stated that as many Chinese companies began deploying open-source DeepSeek models, the demand for Nvidia's H20 chip surged.
Source: DW
Original Article: https://www.toutiao.com/article/1829696919537664/
Disclaimer: This article solely represents the author's personal views.