TSMC's latest update suggests the AI capex cycle still has room
The 35% first-quarter revenue jump suggests the AI capex cycle may still have room to run.
That impression was reinforced this week. At the shareholder meeting in Hsinchu, held against the backdrop of Computex, management had already raised its annual revenue forecast and said it was increasing capital spending this year. CEO C.C. Wei also linked demand to the broader adoption of AI models across consumer, enterprise, and sovereign AI applications. Investors with a bullish read see the signs of an infrastructure layer in the middle of an expansion cycle: revenue accelerating, spending rising, and customers still expressing a positive outlook for the AI industry.
The counterargument is that the stock is no longer a hidden story. With shares already repriced higher, investors are paying for TSMC to maintain its AI leadership. That makes Wei's pricing remark more notable. When asked whether TSMC could raise prices, he said, "I'd like to do that ... we still need to make money," but quickly added that the company does not want to raise prices abruptly like memory makers and is focused on sustainable, long-term operations. If pricing power is limited, the market has less room for disappointment.
So the core question remains: with hefty spending by American cloud-computing giants still pointing to strong compute demand, is TSMC simply expensive, or is its valuation still supported by an AI build-out that has not yet peaked?
Why TSMC still sits at a real bottleneck
The next test is not whether AI demand exists. It is whether TSMC remains the layer through which that demand becomes usable compute. Management has been direct about the constraint: demand is high and it can only produce so much, while analysts said 3-nanometre technology and advanced packaging continue to outstrip current production capacity. When the bottleneck is fabrication and packaging, the foundry is more than a passive beneficiary of AI spending.
Packaging matters as much as the node
Modern AI chips are not won on transistor density alone. The constraint runs through the broader production pipeline, including performance, yield, and the packaging capacity needed to turn chips into deployable systems. Reuters tied TSMC's strong quarter to both 3-nanometre technology and advanced packaging, which matters because packaging can expand the value pool beyond node shrinks alone. Fab construction, tool installation, and packaging integration also take time, so supply is unlikely to equilibrate quickly.
Demand still looks broad
The demand side still looks structural rather than narrowly tied to one product cycle. TSMC said customers continue to express a positive outlook for the AI industry, while American cloud-computing giants are expected to spend more than $600 billion this year on data centers. Management also said it would take a very long time to fully satisfy U.S. customer demand through production in the United States. That does not prove TSMC's share of that demand, but it does suggest the build-out is broad and still exceeds near-term supply. Of course, the company is also monitoring rising component costs, which is a margin watchpoint rather than a demand thesis.
What would strengthen or weaken the bull case
The bullish case depends less on dramatic price hikes than on TSMC's ability to keep converting AI demand into shipped compute without meaningful deterioration in economics. Right now, the clearest confirmation is still management's follow-through: TSMC has already raised its annual revenue forecast and said it was stepping up capital spending this year. In foundries, pricing power often shows up as customers accepting expansion plans even while the company deals with rising component costs.
Signals that support the thesis
- TSMC keeps advancing spending while demand remains broad across consumer, enterprise and sovereign AI applications.
- Hefty spending by American cloud-computing giants continues, suggesting hyperscalers still view advanced compute as strategic infrastructure.
- 3-nanometre technology and advanced packaging continue to outstrip current production capacity.
What could break it
- Spending rises but guidance stops moving higher, raising doubts about diminishing returns on capex.
- Customer outlooks weaken as pressure for clearer AI returns starts to show up in orders.
- Geographic diversification improves faster than expected and eases the supply bottleneck.
Stay constructive only if structural scarcity still holds. If it does not, the market is more likely to reprice TSMC as a premium cyclical foundry than as an irreplaceable AI bottleneck.