TSMC to Expand Chip Production in the U.S. with a $26.5 Billion Investment, Emphasizing AI Valuations' Increasing Focus on Cash Flow

TL;DR

· TSMC's Q2 performance and full-year guidance strengthen AI chip demand, with full-year capex raised to $60 billion to $64 billion.

· Market focus is on US fab costs, 2nm ramp-up, and long-term capital returns, as strong demand does not necessarily mean pressure-free margins.

· Related Tickers: TSM, NVDA, AMD, AVGO, ASML.

Following the release of its Q2 financial report on July 16, TSMC raised its full-year revenue and capital expenditure guidance, and continued to advance its US expansion plan.

This should have been a standard positive for the AI supply chain. High revenue growth, increased share of advanced processes, revenue contribution from 2nm, and multi-year structural demand described by management for US clients. However, the market discussion has shifted from "Do they have orders" to "Can the orders translate into adequate returns".

The reason is straightforward. The stronger the AI demand, the heavier the capital investment required by TSMC. Building fabs, purchasing equipment, expanding advanced packaging, all require upfront spending. US fabs correspond to higher costs, longer ramps, and more complex execution environments.

TSMC's Chairman Mark Liu provided a somewhat optimistic answer during the conference call. He emphasized that the company's investments are driven by multi-year demand from leading US clients, and will also maintain a reasonable gross margin necessary for long-term expansion, without abruptly raising prices due to supply constraints.

Strong Earnings Report Does Not Eliminate Return Concerns

The Q2 data itself is robust. TSMC officially disclosed USD revenue of $40.2 billion, a YoY growth of 33.7%. In New Taiwan dollars, revenue increased by 36.0% YoY. Net profit was NT$706.56 billion, up 77.4% YoY.

The revenue structure also points to AI. Advanced processes, meaning 7nm and below nodes, already account for 77% of wafer revenue. 2nm contributed 3% in the first quarter. High-end GPUs, customized AI chips, and high-performance computing chips all rely on this capacity.

The message conveyed by the financial report is clear. AI demand has already entered TSMC's revenue structure, rather than merely remaining in the order narrative. The company has also raised its 2026 USD revenue growth expectation to slightly above 40%, and increased its full-year capital expenditure to $60 billion to $64 billion.

However, revenue growth rate is not the only variable. The higher the capital expenditure, the more investors will inquire: When will this money translate into capacity, when will the capacity turn into revenue, and can the revenue sustain the current high profit margin.

The AI Demand Shifts the Expansion Cycle

TSMC's current expansion is not just a capacity addition in a normal inventory cycle, but more like a rearrangement for the AI supply between 2026 and 2028.

AI chips not only require advanced processes but also advanced packaging. Advanced packaging can be understood as efficiently putting together multiple chips to enable faster synergy among computing, memory, and data transfer. For AI servers, packaging also determines the final supply.

The company disclosed that 70% to 80% of the annual capital expenditure will be used for advanced processes, about 10% for special processes, and 10% to 20% for advanced packaging, testing, mask, and other processes.

America's expansion is also propelled by the same logic. According to company announcements and media summaries, after TSMC's additional $100 billion investment in Arizona, the cumulative commitment in the United States is about $265 billion. The additional investment will be used for several advanced logic wafer fabs and advanced packaging facilities.

C.C. Wei emphasized that there would not be a sudden significant price hike, which should also be viewed within this framework. This does not mean that TSMC does not have pricing power, but rather the company places more emphasis on long-term customer relationships and capacity commitment. For NVIDIA, AMD, and Broadcom, stable capacity allocation is more important than short-term pricing.

American Expansion Raises the Cost Benchmark

The expanded investment in the United States is the most sensitive part of this reassessment. It reduces the market's concerns about the concentration of capacity in a single region and allows the American AI industry chain to have a stronger narrative of domestic manufacturing.

However, building fabs in the United States is not a cost-free safety net. Labor, construction, supply chain support, and management costs are usually higher than those in Taiwan. In the initial phase of starting a new fab, there will also be yield and utilization rate climb. The lower the yield, the higher the unit cost.

Therefore, American expansion carries two implications. It helps alleviate geopolitical risks, making customers and the U.S. government more willing to see TSMC as part of AI infrastructure, but it may also reduce the overall profit margin, diluting the past "Taiwanese ecosystem plus ultra-high efficiency" cost advantage.

This will change TSMC's valuation anchor. In the past, the market was willing to pay for its technological scarcity and high-profit margins. Now, it also needs to assess whether the post-Americanization capacity can maintain a high enough return on invested capital.

Gross Margin and Cash Flow Evaluate the Expansion

TSMC's current proof point is that AI demand is strong, US expansion is accelerating, and the company is entering a phase of higher intensity capital expenditure. However, these conditions do not directly imply better long-term returns.

The short-term pressure is on gross margin. The company's conference call indicated that the steep ramp-up of the 2nm node in the second half of 2026 is expected to dilute the gross margin by approximately 3 to 4 percentage points. The dilution of the gross margin due to overseas fabs is initially expected to be 2 to 3 percentage points, which may later expand to 3 to 4 percentage points.

This indicates that the margin pressure has already been factored into the management's guidance. Even with rapid revenue growth, the 2nm ramp-up and the startup costs of overseas fabs may temporarily squeeze the profit margin. If the market values the company based on the high-margin leading-edge foundry, a reduction in gross margin could amplify stock price fluctuations.

The mid-term variable is free cash flow. CFO Wendell Huang emphasized a strong operational cash flow, indicating that the company is not expanding due to cash constraints. However, the $60 billion to $64 billion capital budget will still consume a significant amount of cash, and the space for dividends and shareholder returns will need to continue to be tested.

The long term depends on the yield and utilization rate of the US fabs. If the US advanced processes and packaging capacity ramp up steadily and are continuously absorbed by AI customers, the high capital expenditure will transform into a stronger moat. If AI capital spending slows down after 2027, or if the costs of the US fabs remain higher than expected in the long term, TSMC's growth story will be repriced based on return rate discounts.

What's most worth tracking for TSMC now is not whether quarterly revenue continues to hit new highs, but the quality of profitability after high investments. The AI supercycle has already given it a reason to expand, and the market will reassess whether this company can still deliver the high-quality returns of the past based on gross margin, free cash flow, and the progress of the US fab ramp-up.

Welcome to join the official BlockBeats community:

Telegram Subscription Group: https://t.me/theblockbeats

Telegram Discussion Group: https://t.me/BlockBeats_App

Official Twitter Account: https://twitter.com/BlockBeatsAsia