Huang Renxun, founder and CEO of GPU maker NVIDIA, publicly considered using Intel’s foundry services for the first time. Former well-known foreign analyst Lu Xingzhi said that although NVIDIA has learned the hard way of switching to foundry, using Intel’s foundry services To suppress the price increase of TSMC’s advanced process, it is also a necessary way to diversify risks. It is possible to know the progress of Intel’s process development, so as to achieve the purpose of knowing ourselves and ourselves for TSMC and ourselves.
Lu Xingzhi pointed out on his personal Facebook fan page that NVIDIA announced for the first time that it would consider cooperating with Intel’s foundry. He has several views to share:
- Mr. Huang disperses the risk of wafer foundry. It is true to consider cooperating with Intel to suppress the price increase of TSMC’s advanced process, but the painful lessons of Altera, Qualcomm, and NVIDIA before switching to foundry, NVIDIA is not ignorant, of course, as the CEO of NVIDIA , it is necessary to diversify the company’s foundry risk.
- Intel has to compete not only with NVIDIA in the GPU market, but also in the CPU market. It is a bit strange to hand over the most advanced designs to Intel (NVIDIA also announced that it will buy ARM knowing that it will not pass) It is strange, but both have other purposes). Mr. Huang said that it doesn’t matter. Before, many product content blueprints were shared in advance. Intel has long known the secrets of NVIDIA, but I think the original chip design is very different from the product, especially the tape-out time is very early. It is common sense that the foundry does not compete with customers. Could it be that the core values that Zhang Dong insisted on in the past were all nonsense.
- I think that the main reason is that Mr. Huang takes Intel as an arrow (TSMC is a smart person and it should be very clear), NVIDIA should be the last two of all TSMC customers (the other is AMD) will use Intel’s foundry Intel Foundry Services A company that makes a lot of advanced products (there is a good chance of mass production of some mature products). The point is to see if other customers can do it for Intel. Customers most likely to adopt Intel are system vendors such as Google, Facebook/Meta, Amazon, Alibaba, Apple, etc. Otherwise, Qualcomm, MediaTek, European, American, and Japanese automotive MCU and analog chip manufacturers, which have no obvious competition with Intel, especially after Intel bought the tower, these customers would suggest to stop and listen before they adopt IFS in large numbers.
- The last point, just like Mr. Huang knew that the ARM transaction would not go through, but he still had to go through the process. As smart as Mr. Huang, he should have wanted to understand the progress of various product processes through cooperation with Intel, but then he would become a spy of TSMC, who knew himself and his opponent.