[E-Daily Reporter JAEMIN SONG ] Amid concerns that the semiconductor industry has passed its peak—the so-called “peak-out”—TSMC, the world’s largest semiconductor foundry, has announced record-breaking earnings that significantly exceeded market expectations along with aggressive investment expansion plans. This is lending weight to the analysis that the artificial intelligence (AI) semiconductor supercycle remains intact. Expectations for the second-half earnings of SamsungElectronics—which recently announced its highest-ever preliminary results—and SK hynix, set to report its earnings later this month, are also expected to rise significantly.

According to industry sources on the 16th, TSMC posted a net profit of 706.6 billion New Taiwan dollars in the second quarter of this year, a 77% increase from the same period last year. This is a record-high result that significantly exceeded market expectations (632.6 billion New Taiwan dollars). Revenue also rose 36% year-over-year to 1.27 trillion New Taiwan dollars, with a gross profit margin of 67.7% and an operating profit margin of 60.3%.

The market is paying even closer attention to TSMC’s outlook for the second half of the year than to these earnings results. Although concerns have been raised about the semiconductor market peaking—citing the recent slowdown in memory price increases and AI investment fatigue—TSMC has instead played its trump card by announcing expanded investment and an upward revision to its growth forecast, citing stronger-than-expected AI demand.

Wei Zhe-jia, TSMC’s Chairman and CEO, said during the earnings call that day, “AI-related demand remains very strong,” adding, “The AI megatrend continues to drive more computing demand.”

TSMC also revised its capital expenditure (CAPEX) forecast for this year upward by up to 14%, from the previous range of $52 billion to $56 billion to $60 billion to $64 billion. The company also raised its annual revenue growth forecast in U.S. dollars from “over 30%” to “over 40%.” This is interpreted as reflecting the company’s assessment that expanding production capacity is inevitable as demand for AI semiconductors is growing faster than expected.

The company is also significantly expanding its investments in the United States. TSMC has decided to invest an additional $100 billion in its Arizona production facility. CEO Wei explained, “This is to support the strong, long-term demand from major U.S. customers,” adding, “This investment will help further strengthen the U.S. semiconductor ecosystem and supply chain.”

TSMC’s confidence is also evident in its recent monthly earnings. The company previously announced that its June revenue reached 442.68 billion New Taiwan dollars, setting a new all-time monthly record. Analysts attribute this performance to high utilization rates of its 3- and 5-nanometer advanced processes—which produce most AI server chips—and growing demand for advanced packaging (CoWoS). The market expects the expansion of mass production for the 2-nanometer process and the increase in CoWoS production capacity to continue for the time being.

This is a positive sign for the domestic semiconductor industry as well. SamsungElectronics announced record-breaking preliminary second-quarter results—171 trillion won in revenue and 89.4 trillion won in operating profit—driven by the AI memory boom, while SK hynix is also expected to post record-high earnings based on rising demand for High-Bandwidth Memory (HBM) resulting from increased investment in AI servers. In particular, TSMC’s advanced processes and expanded CoWoS production are leading to increased AI chip production by global Big Tech companies, including NVIDIA, which is highly likely to directly benefit SK hynix and SamsungElectronics, the primary suppliers of HBM.

An industry insider stated, “Recently, there were some who spoke of a ‘peak-out’ based solely on memory prices, but TSMC is the company that feels the impact of customer orders first,” adding, “The fact that its earnings exceeded expectations and it has even increased its investment plans should be interpreted as a sign that AI demand is much stronger than anticipated.”