Yesterday, it was reported by DigiTimes that Intel is on track to become one of TSMC’s largest clients by 2023. The American chipmaker plans to leverage the latter’s 3nm node for its 3rd Gen Arc “Celestial” graphics cards (plus the iGPU chiplets of the 14th/15th Gen Core processors). TSMC’s 3nm node is going to begin volume production in the last quarter of 2022, followed by mainstream adoption by the end of 2023. Till then, however, Intel will need an alternative to power its 2nd Gen Arc “Battlemage” GPUs. That’s where the 4nm (N4) node which is an optimization of the 5nm process comes in.
In the second half of 2023, TSMC’s 3nm node will reach production volume of 50-60K wafers per month. Consequently, the cutting-edge process will only start generating revenue for the Taiwanese foundry in the first quarter of 2023 (mass production begins in Q4 2022).
In addition to Intel, Apple will also adopt the 4nm (N4) process in 2022 for its next generation of Bionic and M2 processors. All of TSMC’s other customers including AMD, Qualcomm, MediaTek, and NVIDIA will rely on the 5nm node till the end of 2023/early 2024. To keep clients happy, TSMC plans to introduce enhanced variant of its 3nm node “N3E” as it works to perfect its 2nm GAA process.