Crazy Robbing AI Chips: Computer Rooms Disassemble Used Goods, Sell 100000


Time:

2023-06-26

Key words:

Pigments

Since the beginning of this year, big models and AI 2.0 are not only a hot dream of creating wealth, but also increasingly become an expensive "game of power". Among the various thresholds constructed by technology, capital, talents, etc., there is also an important admission ticket-do you have a "way" to get enough NVIDIA high-end graphics cards.

01. Sell chips like Hermes

Since the beginning of this year, big models and AI 2.0 are not only a hot dream of creating wealth, but also increasingly become an expensive "game of power". Among the various thresholds constructed by technology, capital, talents, etc., there is also an important admission ticket-do you have a "way" to get enough NVIDIA high-end graphics cards.

Wang Fei is the sales manager of a server customization enterprise in East China. He described the shortage of Nvidia graphics cards to the "city": "The demand is too hot. Some people take the A100 80GB purchased before the chip embargo policy takes effect from the computer room and resell it." These old chips not only have no warranty, but also have a price of more than 100000 yuan compared with the unit price of more than 80000 yuan around August last year.

Even if the goods have been sold and used, the supply of video cards is still seriously insufficient. "You need to take cash before the other party is willing to sell it to you." However, the A100 40GB, which has relatively small display memory and low training efficiency, is not popular even if it is brand new. "More inquiries, fewer deals."

Specifically, this batch of sought-after graphics cards refers to Nvidia's "Tesla" series of products. When training AI large models, GPU (graphics processor, also known as display chip), which is more efficient than traditional CPU, especially high-end GPU, is hard to find.

In 2020 and 2022, Avida successively introduced A100 chip and H100 chip, of which A100 graphics card provides 80GB and 40GB video memory versions, while H100 provides 80GB video memory. The larger the memory, the larger the neural network that can run.

Starting in September 2022, the United States banned Nvidia and others from exporting high-end GPU chips to China. In response to the embargo policy, Nvidia has introduced special Chinese graphics cards A800 and H800 in 2022 and 2023 respectively.

However, nowadays, the A800 chip within the scope of the policy has also seen a prolonged supply cycle and rising prices in the market. Wang Fei revealed: "Before the big model became popular, the A800 would arrive in about two weeks. But now I conservatively estimate with the customer that it will take 8 weeks, even if it is optimistic, it will take 4-6 weeks."