ASML or TSMC for AI? The Gatekeeper Looks Like the Safer Bet

ASML Has the Cleaner AI Play Right Now

For investors choosing between ASML and TSMC today, ASML looks like the cleaner AI play. The reason is simple: equipment tends to get ordered before chips generate lasting sales. ASML's latest numbers suggest the capex debate is settling in its favor, with a second 2026 guide raise to €43 billion to €45 billion in sales and 54% to 56% gross margin. That points to continued funding for advanced chip capacity, and ASML is positioned upstream enough to benefit early in the cycle.

Why ASML's position is easier to underwrite

TSMC's results were strong, but they still leave investors facing a harder question: whether AI demand is durable enough to justify more spending. The company reported NT$706.6 billion in net income and 36% revenue growth in the June quarter, confirming that demand remains healthy. At the same time, investors are still debating whether hyperscalers may be overbuilding computing capacity and whether that spending will earn the expected returns.

ASML has a simpler read-through. It is the only company making EUV lithography machines, and management says customers are accelerating capacity expansion plans. That gives ASML a clearer view of intent in the supply chain, even if TSMC is still responsible for turning that capacity into profitable chip volume.

ASML Orders Suggest the AI Buildout Is Still Continuing

One useful piece of fresh evidence is that ASML's supply chain is still moving. Equipment orders are generally an earlier signal of commitment than chip sales, because they reflect the decision to expand before that expansion has fully matured in revenue.

Why bookings matter more than finished-product sales

A useful analogy is a new flagship store: the spending signal usually shows up in fixtures, lighting, and equipment long before the store posts big sales. In semiconductors, ASML sits closer to that front-end commitment. If customers are still ordering tools for advanced capacity, it suggests they expect future demand to justify the investment.

ASML's latest numbers support that view. In the fourth quarter, the company booked €13.2 billion in orders versus €5.4 billion the prior quarter, a sharp pickup that also beat analyst expectations of €6.32 billion. That does not guarantee demand will stay elevated, but it does show chipmakers were willing to lock in capacity at that point.

Low-NA EUV shipments add another signal

The tool plan adds another layer. ASML expects 60 low-NA EUV tools in 2026, which is 25% more than in 2025. That fits the picture of customers continuing to prepare for future advanced-node demand.

That is the main appeal of ASML here: it does not have to prove that every new fab will be filled with profitable workloads. It benefits when foundries decide to expand. The main risk is still policy. New export restrictions can slow the flow from orders to shipments and revenue.

TSMC Remains the Cash Engine at the Center of AI Chip Demand

The real bull case for TSMC is that this is not a company struggling to fund the AI boom. It is one of the main places that spending is going.

TSMC's financial strength supports the buildout

TSMC's balance-sheet strength is hard to ignore. The company generated nearly NT$783 billion in cash from operations in the second quarter and still ended the quarter with TWD 3.1 trillion in cash. For a business that is building expensive advanced-node capacity, that level of liquidity matters.

Management also made the spending case explicit, raising its full-year 2026 capital budget to $60 billion to $64 billion. Roughly 70% to 80% is targeted at advanced process technologies, with the rest spread across specialty technologies and advanced packaging, testing, mask-making, and other areas. If AI demand remains structural rather than cyclical, that spending can widen TSMC's moat.

There is also no debate about operating momentum. In the June quarter, TSMC posted NT$706.6 billion in net income and 36% revenue growth.

The risk is execution, not demand alone

The complication is that TSMC still has to turn capex into profitable volume. Management has said it may not be able to fulfill demand from American customers for years even as new capacity comes online. That is bullish for leadership, but it also means the company has to execute a large buildout without wasting time or capital.

If AI demand softens after this spending wave, depreciation and other fixed costs can still pressure returns. In that sense, ASML gets paid when the decision to expand is made, while TSMC gets paid by showing that the expanded capacity earns an attractive return.

What Would Confirm or Challenge the ASML-over-TSMC Call?

ASML

Stay constructive on ASML as long as follow-through matches the signal already visible in the €13.2 billion bookings and the second 2026 guide raise: customers remain willing to commit capital to AI capacity now rather than delay those plans. The evidence we have points to customers accelerated their capacity expansion plans. If that changes, the thesis weakens quickly.

TSMC

For TSMC, the bullish case is simpler: keep the raised full-year 2026 capital budget paired with strong cash generation. That combination suggests the buildout is being funded internally rather than carried entirely by optimism.

The main invalidation point is also clear. If AI spending slows before it translates into more durable outcomes tied to robust global demand for AI computing, then TSMC starts to look more like an expensive pre-positioning bet than a cleaner earnings lever.

The two risks that matter most now

• ASML: tighter export controls could slow the translation of orders into revenue.
• TSMC: capex could outrun demand, so spending rises before cash flow fully confirms the story.

If orders cool first, ASML becomes the harder call. If spending rises without demand following through, TSMC becomes the riskier investment.