TSMC Earnings: Refining Guidance and Expansion Plans Amid Strong AI Demand

TSMC refined its 2026 US dollar revenue growth guidance to above 30% from about 30% and moved capital expenditure guidance toward the prior upper bound of $56 billion, signaling continued AI-driven demand strength. First-quarter revenue rose 6% sequentially to TWD 1.1 trillion ($35.9 billion) and gross margin expanded 746 bps to 62.3%. Morningstar kept its TWD 2,700/$428 fair value estimate unchanged, citing strong long-term fundamentals despite higher raw material costs and geopolitical supply-chain risks.

Analysis

TSMC’s guidance reset is less about a near-term earnings beat and more about de-risking the AI supply chain. The key second-order effect is that sustained capex at the top end implies the company is willing to pre-build capacity ahead of visible orders, which usually only happens when customer commit levels and packaging/wafer starts are materially tighter than the market assumes. That should compress the perceived scarcity premium in advanced logic and reduce the probability of a meaningful AI compute bottleneck over the next 6-12 months.

The more interesting signal is competitive, not operational: if 3 nm economics are converging toward corporate average, then the “too-expensive for leading-edge AI” objection weakens, and the moat shifts from pure wafer manufacturing to integrated process-plus-packaging execution. That is negative for Intel’s narrative because EMIB alone is not enough if foundry customers care about yield, capacity certainty, and time-to-ramp; it also raises pressure on other advanced packaging players to prove they can scale without sacrificing thermal or signal integrity. In practice, this favors the ecosystem names that enable high-density interconnect and substrate supply, while leaving laggards vulnerable to share loss in the premium AI chain.

The main risk is that consensus is likely underestimating how quickly geopolitical or commodity shocks can hit margins without immediately changing demand. Over a 1-3 month horizon, higher input costs are a nuisance; over 6-18 months, the real swing factor is whether non-Taiwan capacity and 2 nm execution absorb enough strategic demand to justify the capex burden. The contrarian takeaway is that the market may be overpricing Intel’s packaging threat and underpricing TSMC’s ability to monetize AI demand through multi-node, multi-region capacity expansion, which is a better setup for sustained multiple support than for a blow-off rerating.

Key Decisions for Investors

• Long TSM / short INTC as a 3-6 month pair trade: thesis is that the market is over-discounting Intel packaging share gains while underappreciating TSMC’s capacity, yield, and ecosystem advantages; target 15-20% spread capture, stop if Intel wins a credible AI socket-design announcement.
• Add to TSMC on any 3-5% post-earnings dip over the next 1-2 weeks: the revised capex/revenue stance is a de-risking event, not an acceleration trap; risk/reward favors owning pullbacks with a 6-12 month horizon.
• Buy call spreads on TSM (6-9 month tenor) rather than outright stock if you want AI upside with capped downside: the driver is continued premium-node utilization and packaging expansion, with lower implied-volatility bleed than stock ownership.
• Monitor advanced packaging suppliers and substrate names for a 3-6 month long opportunity: TSMC’s capacity buildout should pull through demand for high-end packaging inputs, but only names with tight yield control and non-China exposure should be favored.
• Avoid shorting the AI supply chain on this print: the better expression is relative value against cyclical semiconductor equipment names that have already priced in aggressive capex, because the risk is a slower capex digestion rather than a capex rollback.