Chapter 14: Rally People_2

In fact, there's no need to think hard about it, the hottest topic in recent years is the semiconductor ban. This includes advanced chips, software, and related materials.

For example, NVIDIA's AI chips, with the NVL72 supercomputing system built around the B200, have a whole machine price of up to 3 million US Dollars.

Let's compare it with the domestically produced Ascend 910C.

In terms of performance, the FP16 computing power of the B200 is about 31 times that of the Ascend 910C, and the memory bandwidth is more than 10 times greater.

The B200 has already seen mass shipments, whereas the production of the Ascend 910C is still limited, and the domestic market mainly uses the less powerful Ascend 910B.

Such a huge gap cannot be narrowed by simply increasing quantities. What others achieve in one year of computing would take you ten or twenty years, the efficiency disparity is too great.