China's self driving was sentenced to court

China's autonomous driving and vehicle intelligence are increasingly aware of the strength of the US semiconductor ban. Among them, the US Department of Commerce issued semiconductor export control measures last month, basically locking in the future of China's autonomous driving. In the next 5-10 years, high-computing power chips become an insurmountable obstacle. This is called a palace punishment with no end.

 

This palace punishment is divided into two types.

 

One is to wield a knife from the palace. Before the introduction of the US semiconductor ban, China's CPU and GPGPU companies have launched chips ranging from 5nm to 14nm to NVIDIA and other giants, and claim to have reached or even surpassed competitors in computing power. After the United States restricted China's acquisition of high-computing power chips, NVIDIA A100 and AMD's equivalent products became the ceiling of such companies in China. This naturally encounters the problem of foundry.

 

Among them, Bicheng Technology and Pingtou Brother hit the wall first. On October 24, 2022, TSMC's series of products using the 7nm process for WallCheng Technology were stopped. The latter once claimed that the performance of its BR100 series products has surpassed that of the NVIDIA A100. Now, these applause claims have exceeded the ceiling of the US limit before the market verifies its true performance.

 

In order to avoid touching the red line set by the United States, Bicheng Technology and flat-headed brothers were forced to modify their own designs and reduce product performance. At the same time, TSMC will only ship the product if it is required to self-declare the output and output power of the chip to prove that the product has no problems under export control.

 

The other is to be tortured by someone else. After the US semiconductor ban came into effect, China's autonomous driving and automotive-grade chip fields were replaced by the slogan of localization. Under this climate, NVIDIA and other market gaps of more than 90% in the market segment have been left to China's local semiconductor manufacturers. However, under the reality that the self-sufficiency rate of computing power and control chips in the Chinese automotive industry is less than 1%, NVIDIA is irreplaceable for the time being.

 

Now, NVIDIA has launched the A800 series of products to fill the gap of the A100 within the safe range of computing power. As a result, China's semiconductor industry has placed bulk orders to buy alternatives to NVIDIA, while insisting that the US regulatory measures cannot stop the growth of China's autonomous driving, and the prospects for domestic substitution will be broader.

 

According to the common standard, the computing power requirement of L2 autonomous driving is roughly 10TOPS, and by L4 and L5 autonomous driving, the computing power needs to be increased to at least 1000TOPS. Without the support of high-computing power chips, L4 autonomous driving is the end.

 

In fact, when the manufacturing process relies heavily on foundries, China's autonomous driving and GPGPU companies are designing the future of automata driving on other people's territory, and are in the initial self-research stage.

 

During this period, I have successively sorted out the sales data of the global automobile market, and the European market has basically sorted out. Next are Asia, North America, South America, the Middle East, and Africa.