Intel could soon finalize its deal with TSMC for its 3nm parts
Multiple industry sources have, on many occasions, stated that Intel would seek TSMC's help to manufacture some of its 3nm parts. However, what components will use the cutting-edge node is still shrouded by mystery. We know that the upcoming Arc Xe-HPG DG2 GPU will be fabbed on TSMC's 6nm process. Hence, we can reasonably assume that its successor could use the 3nm node along with components for Meteor Lake, Ponte Vecchio, and others. Now, a new report from Digitimes tells us that Intel and TSMC's deal is in the final stages of completion.
It also tells us that pilot production for TSMC's node has already begun. We learned earlier that mass production was expected to kick off in H2, 2022, at least for Intel parts. However, Apple has managed to secure the lion's share of TSMC's capacity for its 2022-bound hardware, so we can expect to see the Apple A16 Bionic get produced en-mass by then. An improved version of the node, dubbed N3E, will be finalized by 2023, followed by the 2nm N2 node in 2025.
Initially, TSMC's N3 output will be capped at 40,000 wafers per month. That figure is expected to increase to 60,000 in H1, 2023. For now, Intel and TSMC appear to be the only two contenders for the highly sought-after fabrication process, but we can expect other names to pop up soon. Intel has also called dibs on some of TSMC's 2nm production capacity, presumably to supplement its in-house 20A process.
So far, TSMC has maintained market dominance over its main competitor Samsung Foundries. However, that could change once both companies' 3nm processes enter mass production. Samsung is said to be using the newer GAAFET (Gate-All-Around Field Effect Transistor) technology for its 3nm parts, while TSMC is sticking with the tried and tested FinFET(fin field-effect transistor). It could give Samsung the much-needed edge to sway some customers away from TSMC.
Digitimes (in Chinese)