TSMC Increases Capital Budget to $60–64 Billion, U.S. Expansion Reaches $26.5 Billion

• TSMC's second-quarter results and full-year guidance reinforce demand for AI chips, with the full-year capital budget raised to $60 billion to $64 billion.
• Market分歧 focuses on U.S. manufacturing costs, 2-nanometer ramp-up, and long-term capital returns; strong demand does not equate to unpressured profit margins.
• Underlying assets: TSM, NVDA, AMD, AVGO, ASML.

After announcing its second-quarter earnings on July 16, TSMC raised its full-year revenue and capital expenditure guidance and continues to advance its U.S. 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 2-nanometer technology, and U.S. customer demand described by management as multi-year structural demand. Yet the market’s focus has shifted from “whether there are orders” to “whether those orders can generate sufficiently strong returns.”

The reason is straightforward: the stronger the demand for AI, the greater the capital TSMC must invest. Building factories, purchasing equipment, and expanding advanced packaging all require upfront spending. Building a factory in the U.S. also entails higher costs, a longer ramp-up period, and a more complex operational environment.

TSMC Chairman Mark Liu provided an optimistic response during the earnings call. He emphasized that the company’s investments are driven by multi-year demand from leading U.S. customers, and it will maintain reasonable gross margins necessary for long-term expansion, without abruptly raising prices due to supply constraints.

The Q2 data itself was strong. TSMC officially reported $40.20 billion in USD revenue, a 33.7% year-over-year increase. In New Taiwan Dollars, revenue increased by 36.0% year-over-year. Net profit reached NT$706.56 billion, up 77.4% year-over-year.

Revenue structure also points to AI. Advanced processes, namely 7 nanometers and below, now account for 77% of wafer revenue. The 2-nanometer node contributed 3% in its first quarter. High-end GPUs, custom AI chips, and high-performance computing chips all rely on this capacity.

The financial report clearly indicates that AI demand has now entered TSMC’s revenue structure, rather than remaining just an order narrative. The company also raised its 2026 U.S. dollar revenue growth forecast to just over 40% and increased its full-year capital budget to $60 billion to $64 billion.

However, revenue growth rate is not the only factor. The higher the capital budget, the more investors will ask: When will this money become capacity? When will that capacity generate revenue? And can the revenue sustain past high profit margins?

TSMC’s current capacity expansion is not merely a typical inventory cycle-driven replenishment; it’s more like reordering the rankings for AI supply between 2026 and 2028.

AI chips require not only advanced manufacturing processes but also advanced packaging. Advanced packaging can be understood as efficiently integrating multiple chips together to enable faster coordination between computation, memory, and data transfer. For AI servers, packaging also determines final supply.

The company disclosed that 70% to 80% of its annual capital budget will be allocated to advanced processes, approximately 10% to specialty processes, and 10% to 20% to advanced packaging, testing, masks, and related areas.

Expansion in the United States is also proceeding under the same logic. According to company announcements and media summaries, after TSMC added $100 billion in investment in Arizona, the total committed investment in the United States has reached approximately $265 billion. The new investments will be used for several advanced logic wafer fabs and advanced packaging facilities.

Wei Zhengjia emphasized that the avoidance of sudden, large price increases should also be viewed within this framework. This does not mean TSMC lacks pricing power, but rather that the company places greater importance on long-term customer relationships and capacity commitments. For NVIDIA, AMD, and Broadcom, securing stable access to capacity is more critical than short-term pricing.

The expansion of U.S. investment is the most sensitive aspect of this revaluation. It reduces market concerns about concentrated production capacity in a single region and strengthens the narrative of a domestic U.S. AI supply chain.

But setting up a factory in the U.S. is not a free safety net. Labor, construction, supply chain infrastructure, and management costs are typically higher than in Taiwan, and the new factory must also go through a ramp-up phase for yield and utilization. The lower the yield, the higher the unit cost.

Therefore, expanding production in the United States carries two implications. It helps mitigate geopolitical risks and makes customers and the U.S. government more willing to view TSMC as part of AI infrastructure, but it may also reduce overall profit margins and dilute the past cost advantage of Taiwan’s ecosystem and ultra-high efficiency.

This will shift TSMC’s valuation anchor. Previously, the market was willing to pay for its technological scarcity and high profit margins. Now, investors must also assess whether its capacity, after U.S. localization, can maintain a sufficiently high return on invested capital.

TSMC can currently demonstrate that demand for AI is strong, expansion in the U.S. is accelerating, and the company is entering a phase of higher capital expenditure. However, these conditions do not directly guarantee better long-term returns.

Short-term pressure on gross margin. According to the company’s earnings call guidance, the steep ramp-up of 2-nanometer production in the second half of 2026 is expected to dilute gross margin by approximately 3 to 4 percentage points. The ramp-up of overseas facilities was initially projected to dilute gross margin by 2 to 3 percentage points, with potential expansion to 3 to 4 percentage points in later stages.

This indicates that management has factored in margin pressure into its guidance. Even with rapid revenue growth, the ramp-up of 2-nanometer production and startup costs for overseas facilities may temporarily pressure margins. If the market values the company as a high-margin leader in advanced foundry services, a downward revision in gross margin could amplify stock price volatility.

The medium-term variable is free cash flow. CFO Wendell Huang emphasized strong operating 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 significant cash, leaving dividend capacity and shareholder returns to be further tested.

In the long term, it depends on the yield and utilization rates of U.S. fabs. If U.S. advanced process and packaging capacity steadily ramps up and is consistently absorbed by AI customers, high capital expenditures will translate into stronger competitive barriers. If AI capital expenditures slow down after 2027, or if U.S. fab costs remain persistently higher than expected, TSMC’s growth narrative could be re-priced at a discount due to lower returns.

What’s most worth tracking for TSMC right now isn’t whether quarterly revenue continues to hit new highs, but the quality of its profits after heavy investment. The AI supercycle has provided the rationale for expansion; the market will now reassess whether the company can still generate the high-quality returns of the past, using metrics such as gross margin, free cash flow, and the ramp-up progress of its U.S. factory.