It is expected that 2nm will enter mass production by H2 2025, with initial production capacity reported to be at 40,000 units. More importantly, the demand for 2nm is expected to be much more diverse this time, with most of the portion going towards the AI hype, fulfilling orders from the likes of NVIDIA and AMD. TSMC anticipates that the demand for 2nm is expected to be even higher than 3nm, which means the production figures are expected to reach an all-time high, since TSMC will be one of the only firms on ground to offer a competitive solution.

Interestingly, TSMC's 2nm node is expected to be one of the most expensive ones for clients to integrate into their products, which means that there is hesitance to adopt it. NVIDIA will likely rely on 3nm with its current Blackwell Ultra AI lineup, while AMD will employ 3nm for its Instinct MI350 AI accelerator lineup. We could see the debut of 2nm with Rubin and the Instinct MI400 series, but it will definitely come with higher costs. Moreover, a massive portion of demand is also set to come with ASIC developers, especially firms like xAI, Google and Meta.

TSMC has seen massive success with the AI hype, and it is expected to continue at this pace given that the demand will indeed see a rise. With competitors like Intel and Samsung slacking with their solutions, TSMC still maintains the lead, grabbing most of the semiconductor market share.

By: DocMemory
Copyright © 2023 CST, Inc. All Rights Reserved