The News
Donald Trump is expected to meet Friday with Nvidia CEO Jensen Huang, the first meeting between the US president and the head of the leading producer of advanced AI chips.
Multiple reports said the meeting was planned before the Chinese start-up DeepSeek’s latest AI release, which shook Silicon Valley this week with its new, low-cost chatbot, and erased $589 billion of Nvidia’s stock market value.
Trump and Huang are expected to discuss AI policy, the Financial Times reported. Nvidia has long protested new export controls limiting sales of its powerful AI chips to China. Many Republicans have argued that the DeepSeek developments show the need for even tougher limits on exporting AI technology to China.
SIGNALS
Western countries scrutinize DeepSeek
Huang and Trump’s meeting comes as the US is probing whether China’s DeepSeek evaded Washington’s restrictions on AI chip exports by buying Nvidia’s advanced semiconductors via intermediaries in Singapore, Bloomberg reported. The Chinese firm said it used Nvidia H800 chips, which were legal to export in 2023, while Nvidia has suggested that DeepSeek did not violate Washington’s restrictions. Still, DeepSeek’s R1 model, which its developers claim achieved results comparable to those of leading US AIs on a fraction of the computing resources, is facing increasing Western scrutiny. Italy blocked DeepSeek’s highly popular app over concerns about user data, and authorities in France and Ireland said they planned to question the firm over its privacy measures.
Investors debate Nvidia’s future after DeepSeek shock
Nvidia’s record drop in value over DeepSeek’s low-cost AI model was because investors feared that chips will become less important if fewer are required to train cutting-edge AI models. However, some in Silicon Valley, including Nvidia CEO Huang, argued that reducing the cost of training AI models would actually spur investments in AI chips, as more companies can afford to deploy AI systems. While the US chip giant remains best known for producing advanced chips for training AI systems, Nvidia has said it earns as much revenue from chips deployed to process user requests of finished models, and expects this demand to rise in the near future, the Financial Times reported.
Washington debates whether DeepSeek’s success spells failure of US’ AI chip restrictions
DeepSeek’s release caused a frenzied debate in Washington about whether the model proves that US efforts to limit China’s AI progress have failed. CNN’s Fareed Zakaria argued this new Chinese model cast doubts on whether it is possible to rein in China’s AI progress at all, or whether the curbs have simply forced Chinese developers to be more innovative. An AI expert at RAND cautioned against proclaiming the failure of US restrictions, noting that DeepSeek trained its model on chips China can no longer purchase, and that US export controls will not be fully felt until China needs to upgrade its data centers — and will find itself unable to do so.