TSMC is too busy pumping out 3nm chips for Apple's new M3 series processors, with the Taiwanese giant to see capacity utilization decline at the beginning of 2024, but in the second half of 2024, things will pick up as other 3nm customers will join the party. This includes AMD, Intel, NVIDIA, Qualcomm, and MediaTek.
According to Ctee, we are to expect TSMC's current 3nm capacity utilization rate to drop to 65% before things pick up in 2H 2024. US foreign investors are mostly optimistic about TSMC and its generative AI, semiconductor recovery cycle, 2nm technology leadership, and sped-up overseas factory expansion, so they are "optimistic about the long-term investment value" in the company.
Intel will be using TSMC's new N3B process for its next-gen, low-power architecture in Lunar Lake MX (LNL), while we'll also see Intel's upcoming Arrow Lake-H and Arrow Lake-HX processors made on 3nm, which will be "used to significantly fill TSMC's production capacity".
- Read more: TSMC's next-gen 3nm wafer cost are high: new CPUs, GPUs more expensive
- Read more: NVIDIA CEO flies to Taiwan, secures TSMC 3nm wafers for next-gen GPUs in 2024
- Read more: TSMC on 3nm : 3x silicon density over 7nm, 51% less power, 32% faster
The AI GPU side of things is where TSMC is pumping away non-stop, making NVIDIA's upcoming H200 and H20 AI GPUs, as well as AMD's just-announced Instinct MI300X AI accelerator. In 2024, we'll see the introduction of Huida's new B100, with all of these products being made on the N3 process node at TSMC.
The report continues, saying that TSMC is committed to launching new N3 processes, including the N3P and N3X, while N3E is being used to produce HPC and smartphone chips, which entered mass production this quarter. 3nm will be the long-term node in the future, with revenue proportions continueing to increase for TSMC.
TSMC has its more advanced 2nm technology progressing as scheduled and will enter mass production in 2025.