Nvidia's AI chip cost performance has increased by 70% to compete with Huawei!
In 2023, artificial intelligence will usher in a year of artificial intelligence explosion, and ChatGPT, released by OpenAI, has triggered a large model war of artificial intelligence.
Of course, the biggest winner is still NVIDIA, because all artificial intelligence models are inseparable from GPU chips, so the competition of artificial intelligence is essentially the competition of GPU.
Nvidia's revenue, profits, and market capitalization are all growing wildly.
The United States is worried that the rise of China's AI technology will replace the United States, so it wants to block China's capabilities on GPUs to prevent the development of Chinese artificial intelligence.
As early as 2022, Nvidia's H100 and A100 series were banned from sale, so Nvidia launched a weakened version of H800 and A800, while H800 and A800 in the United States, etc., will also be banned from sale in 2023.
Nvidia has no choice but to continue to castrate its own chips, such as H20 and so on, but this castration is different from the last castration, the United States has great restrictions on H20, so the performance of H20 has been suppressed by 80%.
Of course, if there is no better processor than Nvidia in the Chinese market, then Chinese chipmakers may have to choose H20.
However, the emergence of Huawei has allowed Huawei's AI chips to compete with Nvidia in terms of performance, and Huawei has launched the CUDA ecosystem.
This also means that Chinese chip companies no longer need to buy GPU chips from NVIDIA.
And what about Nvidia? Of course, they can only lower the ** and seize the market through the ** war.
According to ** report, Nvidia has adjusted the pricing of H20, which is expected to be 70% cheaper than the original H100, the original RH100 was sold at about 300,000, and the current H20 is only 12,000 to 15,000 to 15,000.
The most important thing is that NVIDIA's ** is on par with Huawei's Ascend AI chips, and even cheaper than Huawei's artificial intelligence chips, obviously trying to suppress Huawei.
In the next step, if Nvidia's H20 sales decline, then Nvidia will reduce **, and Nvidia does not want to lose to the Chinese market.
This phenomenon illustrates a problem, if your products cannot be bought in China, then foreign manufacturers will definitely buy from your products at all costs.
But if China produces it, the price of foreign products will be reduced, and profits will be cut, which shows that if China wants to avoid being cut leeks by foreign manufacturers, it will have to rely on its own efforts.