Bank of America Raises TSMC Capex Forecast on Chip Demand By Investing.com

Bank of America raised its 2027 capex forecast for TSMC to $75 billion from $63 billion, above the $65 billion Street consensus, citing stronger demand for advanced-node capacity and high-performance computing. The bank expects TSMC to meet 2026 capex guidance of $56 billion and reaffirmed its Buy rating with a NT$2,560 price target. Arizona operations are improving, generating NT$39 billion in sales and NT$19 billion in net profit in Q1 2026, though margins excluding grants remain below corporate levels.

Analysis

The key signal is not higher spend, but a shift in spend mix toward front-end node tools. That implies TSMC is pulling forward capacity for the most constrained and strategic part of the supply chain, which should keep advanced lithography, deposition, etch, and metrology vendors tight well into 2027; the beneficiaries are the ultra-high-end toolmakers and select materials suppliers, while second-tier mature-node equipment exposure is likely to lag. This also reinforces TSMC’s role as the bottleneck allocator for AI and server silicon, which tends to widen the gap between leading-edge foundries and everyone else rather than meaningfully expanding industry-wide wafer supply.

The Arizona profitability datapoint matters mainly as a political and capital-allocation risk check, not as a margin thesis by itself. If U.S. operations are already earning respectable economics with limited subsidies, the strategic logic for additional domestic buildout strengthens, but so does the risk of policy pressure to front-load further localization across packaging, test, and eventually more advanced node capacity. That is a medium-term margin drag versus Taiwan, and it likely creates a hidden beneficiary set in U.S. semiconductor infrastructure, power, water, and industrial construction names before it becomes obvious in the chip tape.

The market’s likely mistake is treating the higher capex guide as a pure negative for free cash flow. For TSMC, elevated capex at this point is more a demand validation signal than a growth penalty, and the real valuation risk is not spending intensity but execution timing if equipment arrivals or facility ramp sequencing slip by 1-2 quarters. The contrarian setup is that the better trade may be the picks-and-shovels around the buildout, while TSMC itself stays range-bound until investors get evidence that 2027 capacity translates into sustained pricing power rather than just more depreciation.

Key Decisions for Investors

• Long TSMC on 6-12 month horizon on any pullback toward the lower end of its historical multiple range; risk/reward favors owning the demand validator, but size modestly because capex intensity can cap near-term multiple expansion.
• Add a basket long in advanced semiconductor equipment names with leading-edge exposure over the next 12-18 months; the setup is better than buying foundry shares because the spend mix signals orders before revenue and margins show up.
• Pair trade: long leading-edge tool/material suppliers vs short mature-node / commodity semiconductor equipment exposure over 3-9 months; the expected spread comes from front-end capacity allocation, not broad industry growth.
• Watch U.S.-linked semiconductor infrastructure and industrial contractors for a 6-24 month rerating; buy on weakness if domestic fab localization accelerates, with asymmetric upside if policy support increases.
• For TSM specifically, consider an options structure that benefits from grind-up performance but limits downside from execution delays, such as a 6-9 month call spread into the next capex and production update cycle.