US authorities are close to allowing the export of the latest Nvidia accelerators to Saudi Arabia

Over the previous two years, the US government has at least twice strengthened export restrictions on Nvidia computing accelerators, at one stage blocking the way for their free supply to Saudi Arabia due to concerns about the possibility of transit to China. Having achieved some concessions from the Saudi authorities, the Americans are now ready to soften their position.

Image source: NVIDIA

Semafor reports this, citing sources familiar with the progress of negotiations between the two countries. It is generally accepted that this week’s GAIN global summit in Riyadh, dedicated to discussing issues related to artificial intelligence, became a platform for such negotiations. Representatives from Groq, Google and Qualcomm also took part in the event.

Now the Saudi Arabian authorities, according to the summit participants, are working to satisfy the demands of the US authorities regarding the return of the right to supply Nvidia computing accelerators to this Middle Eastern country. Saudi Arabia is vying to receive H200 accelerators with Hopper architecture, which are still considered flagship, since deliveries of accelerators with the more advanced Blackwell architecture have not yet officially begun. Satisfying US national security demands in this area implies a cooling of Saudi Arabia’s relations with China, and the presence of Chinese companies at the event was less pronounced than before. However, Huawei and Alibaba provided their own stands at the summit.

According to all the rules of eastern diplomacy, Saudi Arabia is in no hurry to cut off all ties with China, since friendship with the United States may be short-lived. In the field of artificial intelligence, Saudi Arabia is still faced with a shortage of its own qualified specialists, so it is forced to depend on other countries, including China and the United States. But for the latter, it is attractive due to the presence of a large amount of inexpensive electricity, and for the implementation of large-scale projects in the field of AI, this factor becomes significant.