Discussing Intel’s collaboration with TSMC is common in connection with the contract to manufacture programmable arrays, neural network accelerators, mobileye components, and cheap chipsets. Now sources say that orders for the introduction of 5nm GPUs from the Intel Xe family will also be added to the TSMC pipeline.
As Intel’s management has repeatedly stated, in recent years the company receives up to 20% of all revenues from the sale of products issued by third parties. At the same time, the firstborn of Intel’s 7nm process technology must be the GPU that will be part of the Ponte Vecchio computing accelerator. Intel representatives have often said that with the transition to 7nm process technology, the company will return to the previous lithographic standards rate.
Considering that Ponte Vecchio will make its debut at the end of 2021, Intel could launch 5 nm products by the end of 2023 or mid-2024. TSMC will begin mass production of 5nm products in the third quarter of this year. According to some reports, even NVIDIA will be interested in launching its GPUs with 5nm technology next year. Under such circumstances, Intel can maintain the uncomfortable status of an outsider.
The Commercial Times, citing industry sources, reports that most of the 5nm products on the TSMC assembly line will go into production next year. Most interestingly, Intel has announced its intention to persuade TSMC to manufacture its products in 5nm process technology up to and including 2022. In addition to programmable arrays (FPGAs), discrete graphics processors from the Intel Xe family can also be included in the corresponding product range. By the end of this year, TSMC estimates that the implementation of 5nm products will generate up to 10% of revenues. In addition to HiSilicon, Apple and Qualcomm, NVIDIA with its Hopper GPUs and AMD with Zen 4 processors and RDNA 3 generation graphics solutions will be TSMC’s customers in the 5nm process technology.