It is reported that Intel China is full of orders for AI processor Gaudi 2 and urgently place additional orders with TSMC.
CTOnews.com, September 19, July this year, Intel launched a new Gaudi2 processor for the Chinese market, which is mainly used to accelerate AI training and reasoning. CTOnews.com summarizes the parameter information of this processor:
24 programmable Tensor processor cores (TPCs)
21 100Gbps (RoCEv2) Ethernet interfac
96GB HBM2E memory capacity
Total memory bandwidth of 2.4TB per second
SRAM on 48MB chip
Integrated multimedia processing engine
DIgitimes quoted people from the supply chain as saying that Intel's "special supply version" AI processor Gaudi 2, launched in the domestic market, has been on the rise in recent months and has now pursued orders from TSMC.
With the rise of the AI boom triggered by ChatGPT, the demand for AI computing power in the domestic market is rising, but the most critical Nvidia chip is short of TSMC CoWos advanced packaging capacity, even if the A800 and H800 are still unable to meet customer demand, resulting in a substantial increase in the price of advanced chips.
Now, the prosperity of how many goods NVIDIA has sold in the past is being repeated in Intel.
According to people in the supply chain, the visibility of Gaudi 2 orders has reached mid-2024, while the next generation of Gaudi 3, which uses TSMC's 5x4nm process, is also accelerating the listing process, which is expected to come in 2024.
People in the supply chain also revealed that the AMD MI300 series that Meta, Microsoft, Google and AWS four major cloud service providers in North America (CSP) are looking forward to are also greatly affected by TSMC's capacity problems.
He believes that due to the possibility that NVIDIA exports to China will be completely banned in the future, and the AMD MI300 series has not been listed on a large scale, Intel has seized the godsend opportunity, even the "special supply version" of Gaudi 2 is in short supply.
Related readings:
Intel launches new Gaudi2 processors for the Chinese market: total memory bandwidth 2.4TB / s, for AI training.
