That is, to see the overall processing performance of the chip divided by the die area, as a calculation method as a benchmark for assessing the export license. Under the new regulations, NVIDIA's special edition A800 and H800 no longer meet the performance density index requirements, and even consumer-oriented RTX4090 graphics cards are not exempted from the scope. In the face of this situation, it is reasonable to assume that AMD's magic version of the AI gas pedal before applying for an export license should have passed the internal compliance department of the company's review, but still hit a wall.
At present, the industry's interpretation of this matter is mostly focused on two levels, one is to emphasize the U.S. Department of Commerce BIS has a "caliber of flexibility", whether to give the export license is not entirely in accordance with the threshold of the paper regulations, one is to interpret from the application of scenarios, emphasizing the U.S. to China's AI chip arithmetic curtailment, the card of the big model of artificial intelligence upgrades.
Admittedly, both interpretation dimensions are very persuasive. But if you take into account the U.S. Department of Commerce for more than half a year on China's semiconductor technology route research anxiety, especially Huawei Mate60's listing, representing a major breakthrough in China's local chip process, the U.S. side of the series of reactions to this matter, AMD's "MI309" temporarily banned behind the scene there may be another dimension.
$NVIDIA (NVDA.US)$ $Advanced Micro Devices (AMD.US)$