In the past few years, TSMC’s N5 node has almost been dominated by Apple chips. However, as TSMC’s production capacity gradually increased, more and more companies began to adopt new process technologies, and TSMC was also forced to work hard to increase its production capacity.
According to a DigiTimes report, TSMC will increase its N5 production capacity by about 25% this year to meet demand for N5 chips from companies such as AMD, Nvidia and MediaTek.
For example, it is reported that Nvidia’s Hopper computing cards (which now also includes Ada Lovelace consumer GPUs) will use the 4N process, while MediaTek’s N5 will be used for its high-end Dimensity chips, and the N4 will be used for flagship Tianji chips. Gui 9000.
Furthermore, TSMC’s N5 capacity can reach 120,000 wafers per month (WSPM), but 120,000 N5 WSPM is just a goal TSMC plans to achieve in early 2022. In addition, in order to meet more N5 customers, TSMC will build out some additional equipment, which is expected to increase N5 production to 150,000 WSPM by the third quarter of 2022.
Judging from the previous information, Nvidia plans to launch its Hopper computing card and start selling it in the third quarter, so TSMC must already be using the N4 node to achieve mass production of the H100 at this stage, although the output is not very high.
At the same time, in addition to TSMC’s largest customer, Apple, MediaTek is also determined to remain the second largest customer of TSMC due to its massive orders for Dimensity 8000/8100/9000 processors. It is currently unknown the capacity allocation of AMD, NVIDIA, Qualcomm and other manufacturers. plan.
Judging from existing revelations, AMD is expected to launch its next-generation Zen 4-based Epyc, Ryzen CPUs, and RDNA 3-based Radeon RX 7000 series GPUs this fall. Nvidia will also be launching its Ada Lovelace-based GeForce RTX 40-series consumer graphics cards around the same time, so TSMC will need a lot of capacity to produce these chips.