TSMC is poised to release its Q3 earnings report on October 16, and the outlook suggests the company is likely to excel, driven almost entirely by insatiable demand for Artificial Intelligence (AI) infrastructure. However, this financial and technological dominance is increasingly overshadowed by intense geopolitical pressure and questions about the long-term sustainability of the AI boom itself.
In this episode, we analyze the factors that may affect the short-term and long-term outlook of the giant’s leadership status, which may raise questions for TSMC management in the earnings conference.
Financial Strength and AI Dominance
Heading into the Q3 report, TSMC is showing clear momentum:
Robust Monthly Revenue: July 2025 revenue hit NT$323.17 billion, a massive 22.5% increase from June, followed by another 3.9% sequential increase in August 2025 (NT$335.77 billion).
Strong Guidance: The company’s 3Q25 net revenue guidance is projected to be between US$31.8 billion and US$33.0 billion.
Priceless Pricing Power: TSMC expects to maintain a strong gross margin of 55.5% to 57.5%. This margin confirms significant pricing power, directly enabled by their leadership in advanced manufacturing.
Advanced Node Leverage: Revenue from the 5-nanometer and 3-nanometer nodes combined made up about 60% of total revenue in Q2 2025. This cutting-edge capacity is limited, allowing TSMC to raise wafer Average Selling Prices (ASPs) for these workhorse nodes.
The Future is N2: TSMC is aggressively pushing its 2-nanometer (N2) node for mass production in the second half of 2025. This is expected to have faster adoption and a significantly larger customer pool compared to rivals.
Exchange Rate Advantage: The exchange rate is forecast at US$1 to NT$29 for Q3. However, the actual exchange rate has mostly stayed above NT$30 since August. That could result in a surprisingly good performance in revenues, as TSMC’s revenue is denominated in the local currency (this is a speculation based on the writer’s experience as a forex market reporter).
A Sustainable Recovery?
The broader semiconductor industry is also recovering strongly, with Omdia projecting double-digit year-over-year growth in 2025, led by the massive build-out of data centers for AI. Looking ahead to 2026, the global semiconductor industry is forecasted to grow by a “robust 12.5% annual growth rate”. This sustainability is backed by:
Memory Growth: Memory IC revenue is expected to remain in a high-growth cycle through 2026 due to AI deployment.
NAND Volume: NAND industry bit shipments (raw capacity) are projected to grow at an 18% Compound Annual Growth Rate (CAGR) between 2024 and 2029.
Advanced Packaging Boom: The 2.5D and 3D advanced packaging segment, essential for stacking high-performance AI chips (like TSMC’s CoWoS), is forecast to see a tremendous 27.9% CAGR between 2024 and 2029.
Other topics that TSMC chairman C.C. Wei may be asked about during the earnings conference:
The Geopolitical Silicon Shield Under Attack
TSMC’s technical leadership, however, has made it a central target in U.S.-Taiwan trade friction, directly challenging Taiwan’s “Silicon Shield” security concept.
The 50-50 Proposal: U.S. Commerce Secretary Howard Lutnick proposed that Taiwan should locally manufacture only 50% of the chips the U.S. currently imports from them, a radical suggestion aimed at slashing American reliance from the current 95%.
Taiwan’s Rejection: Taiwan’s top trade negotiator, Vice Premier Cheng Li-Chun, outright rejected the proposal, stating it was “not even discussed” in recent talks.
Strong Rhetoric: Taiwanese politicians reportedly called the U.S. proposal an “act of exploitation and plunder” and an attempt to “hollow out the foundations of Taiwan’s technology sector,” sending a very aggressive signal to Washington. However, Cheng did mention that Taiwan is willing to invest and create semiconductor industrial clusters in the US, just like the one in Hsinchu.
Market Fragmentation Risk: Analysts warn that escalating geopolitical tensions and mounting tariffs could lead to over-inventory builds across the global supply chain, potentially resulting in slower growth for the semiconductor industry in 2026.
The AI Bubble Debate
Adding to the complexity are significant questions about the financing of the AI boom, specifically around the massive investments being made by its largest players:
Circular Financing: Analysts are debating whether the massive demand is truly organic or if companies like NVIDIA are effectively “cosigning the mortgage” for their biggest customers, such as the reported US$100 billion non-binding agreement with OpenAI to deploy at least 10 gigawatts of NVIDIA systems.
Inflated Perception: NVIDIA’s pattern of similar deals with other customers like CoreWeave could give an “inflated perception of true organic market demand”.
Bubble Acknowledgment: Even key figures like OpenAI CEO Sam Altman and Meta CEO Mark Zuckerberg have openly acknowledged the possibility that the industry may be in an AI bubble, comparing it to past infrastructure build-outs that got ahead of near-term demand (e.g., the dot-com fiber optic glut).
Conclusion
TSMC’s Q3 earnings report is likely to confirm its technical and financial dominance, driven by an accelerating, fundamental shift toward advanced AI compute. However, this very strength has turned into a source of geopolitical vulnerability and leverage. The core long-term question remains:
Will the massive, apparently locked-in demand for AI infrastructure eventually force political concessions, or will intensifying domestic IC policies ultimately fragment the market enough to genuinely challenge Taiwan’s dominance by the end of the decade?
Reference:
https://investor.tsmc.com/english/quarterly-results/2025/q3
https://www.cnbc.com/2025/10/02/taiwan-us-chip-production-tsmc-trade-tariffs.html
What Lies Ahead for the Semicon Market - IDC 2026 Semiconductor Outlook Webinar
Global Semiconductor and Foundry Market Overview - by Simon Chen, Principal Analyst, Semiconductor Manufacturing
NAND Flash 2025–2026: From Cycles to Long-term Signals - by Claire Wen, Principal Analyst, NAND Intelligence
