NVIDIA will be allowed to continue selling H20 graphics processing units in China after the Mar-A-Lago visit by CEO Jensen Huang NPR. Huang attended a dinner last Friday at the resort, whose price was at one million dollars per head, although it is unclear whether he had spoken to President Trump.
The White House did not comment on the axis, but it was planning to carry out the unimaginable restrictions in China’s sales in China, within days. NPR Reports that HUANG has offered more investment in AI’s data centers while Microsoft is Retreat The conversion threatens the definitions by raising the cost of construction.
The United States is already restricting the advanced H100 chips exports in NVIDIA to China, due to fears that the country will develop artificial intelligence for use in war and collective control. It is not an unfounded concern, as the country has often suppressed the brutality of its citizens, especially the ethnic minority of Ouigor. But CEOs find God when it comes to earning money. China is a major market for NVIDIA, so the company persuaded the government to allow it to sell chips with somewhat neutral performance, including H20.
To put the importance of China in NVIDIA in the context, it was recently reported Information Chinese companies bought $ 18 billion in H20 chips in the first three months of 2025, expecting an imminent ban. In this sense, the dowry to a million dollars for $ 16 billion of sales is a good return on investment.
Like Apple and Tim Cook CEO, Huang was keen to walk in a thin line on the theme of China. Early this year is She made a quiet visit To the country, where he strengthened the country’s role in NVIDIA growth. Huang noted that more than 1.5 million developers in China use Cuda, its software tool set to train artificial intelligence on its slices. He added that Nafidia cooperated with more than 3,000 startups in the country. “Over the past two decades, we have contributed greatly to updating one of the most dynamic markets in the world, and we are proud of our role in this ecosystem.”
One can argue that export restrictions on China were a double -limited sword. They may slow the country’s developments in weapons or monitoring, but the technology industry there has become innovative in making its models more efficient, and doing more with less. Nothing has been proven more than Deepsek R1The lightweight artificial intelligence model was built in just two months and less than 6 million dollars. The company used compatible H800 chips To train the model.
The H20 chip is particularly suitable for inference tasks, such as the technique of thinking started by Openai’s O1 and Deepseek R1 models.
Huawei is another company that has found silver lining in export restrictions. The smart phone maker stumbled after President Trump, during his first administration, was restricting the company’s access to western processors and programs. Last year, her first smartphone unveiled with Local chip and operating systemAnd an achievement that many critics have not believed that he would be able to achieve so quickly. No, there is no longer a need to worry about gentle playing with the Trump administration.
Kissing the ring may have worked for the CEO of NVIDIA, but many other technology leaders made the pilgrimage to Mar-A-Laggo, and it is not clear that they will get out of it a lot. President Trump’s administration is still noisy for great technology because of the belief that they were controlled by the conservative views. Dead She still faces FTC FTADespite the large donations and frequent trips by Mark Zuckerberg in favor of Carrie Trump.
Either way, the last step is another example of the ease of dealing with the Trump administration. If China is a clear bad actor on the world stage trying to become the new global power, then why does President Trump allow them to continue buying artificial intelligence chips? At least China may buy chips while slapped the tariff on its imported goods to America can reduce the trade deficit.
https://gizmodo.com/app/uploads/2025/01/Nvidia-CEO-Jensen-Huang-RTX-50-Series-GPU-1.jpg
Source link