TSMC Also Raised Its Forecast by 50%… Semiconductor Growth Is Exceeding Expectations / Has the Semiconductor Peak-Out Theory Been Proven Wrong Again? | SemiCon Research Lab CEO Noh Geun-chang

Summary

Noh Geun-chang, CEO of SemiCon Research Lab, argues that the AI infrastructure cycle and semiconductor demand are stronger than investors fear, citing TSMC's 50% upward revision to its 2030 market forecast and dismissing the AI speed-control debate as noise rather than a fundamental negative. He sees hyperscaler capex continuing despite rate hikes, automotive semiconductors doubling by 2030, and back-end packaging as attractive, while humanoid robotics adoption may be slower than hype. He also assesses US tariff-driven localization plans for SK hynix and Samsung Electronics, Intel's weak foundry position, and China's memory/lithography threat as longer-term or not immediate.

• AI infrastructure investment is expected to continue and accelerate despite additional rate hikes.
• TSMC's 50% higher 2030 market forecast is cited as evidence that semiconductor growth is exceeding expectations.
• AI speed-control concerns are framed as a hyperscaler cost issue, not a semiconductor negative.
• Automotive semiconductors are expected to double their share of semiconductor demand by 2030.
• Humanoid robotics adoption is seen as directionally right but slower than expectations.
• SK hynix is building US packaging capacity, and Samsung's US fabs may switch between foundry and memory.
• ASE shows strong back-end packaging results, while Samsung's October earnings may offset Anthropic IPO supply risk.
• China's YMTC/CXMT and lithography localization are watched but not immediate threats.