TSMC Adds $100B to Arizona, Raises 2026 CapEx to $85B — How Long Can the Foundry AI Boom Last?

One-sentence conclusion: TSMC is pushing 2026 capital expenditure to a record $85 billion and adding $100 billion to its Arizona investment — AI demand is real, but the market is repricing what "a growth stock should cost."

Record revenue, falling stock — what's going on?

TSMC's latest quarterly results were spectacular on paper: net profit up 77% year-over-year, and July revenue up 45%. Yet the stock dropped more than 4% when the numbers landed. That gap between record earnings and a falling share price is the market's honest verdict on TSMC's future.

The reason sits in two announcements made alongside the results: an additional $100 billion investment in its Arizona fab (bringing total U.S. commitments to roughly $265 billion), and a raise of 2026 capital expenditure guidance to $85 billion. CapEx is tomorrow's cost. Investors see AI demand that is genuinely hot — but also a company that must spend at unprecedented scale to catch it, squeezing near-term margins.

Arizona itself is already turning around. According to BofA Global Research estimates and TSMC filings, the Arizona site generated about $1.41 billion in revenue in 2Q26 — up 17% quarter-over-quarter and 145% year-over-year, roughly 4% of TSMC's total. First-half net profit surged 663% versus a year earlier, making it TSMC's largest overseas profit contributor.

Where does the $85 billion go?

TSMC's capital spending flows into three buckets:

1. Advanced-node expansion: 1.6nm (A16) is slated for mass production as early as late 2026 — the node that anchors AI chip orders — while 2nm capacity ramps in parallel.

2. Overseas fabs: Arizona, Kumamoto (Japan), and Dresden (Germany) are all under construction simultaneously, steadily lifting the share of overseas capacity.

3. Advanced packaging: CoWoS and related AI-chip packaging capacity remains in short supply and keeps getting expanded.

The direct beneficiaries of higher CapEx are equipment makers — ASML and Applied Materials rose on the news, a textbook example of how the semiconductor equipment supply chain follows foundry spending.

From a two-horse race to a three-way fight

TSMC's near-monopoly is facing fresh pressure from challengers:

• Samsung: announced a 15% foundry price increase and is expanding its Pyeongtaek fab — a sign of tight advanced-node capacity and an attempt to reset market pricing.
• Intel: is winning orders, with Razor Lake reportedly using its N2X process and Socionext adding Intel 18A-P capacity — Intel Foundry is landing real business.
• Equipment bottleneck: ASML's EUV lithography capacity is itself scarce; whoever secures tools first gains capacity leverage.

For customers, a three-way race is good news — more supplier options, more bargaining power. For TSMC, intensifying competition means future pricing pressure, which helps explain the market's lukewarm reaction to record revenue.

The $100 billion American bet

The Arizona expansion is TSMC's largest-ever overseas investment, lifting cumulative U.S. commitments to about $265 billion. It's driven by the "AI chips made in America" trend: U.S. customers (Apple, NVIDIA, AMD) want critical silicon produced domestically to reduce geopolitical risk.

But overseas fabs come with a cost: higher operating expenses than Taiwan's home fabs, a yield-learning curve, and gross-margin dilution. TSMC's bet is that high utilization and AI order premiums offset the cost gap. The Arizona fab's 4nm line is running at full capacity with yields now on par with TSMC's flagship Taiwan fabs — evidence the model can work.

Three indicators to watch

Next time TSMC reports, don't just stare at net profit:

1. CapEx-to-revenue ratio: sustained CapEx above ~40% of revenue means the company is still in "spend to win the future" mode.

2. AI/cloud revenue share: whether AI-related revenue keeps growing determines how fast the CapEx pays off.

3. Overseas fab gross margin: when Arizona's margin converges with the Taiwan fabs, the overseas expansion thesis is validated.

FAQ

Q1: What do TSMC's $85 billion in 2026 CapEx go into?

Three areas: advanced-node expansion (1.6nm A16 and 2nm), overseas fab construction (Arizona, Kumamoto, Dresden), and advanced packaging (CoWoS) — with AI-related capacity as the top priority.

Q2: Is Samsung's 15% foundry price hike a threat or a help to TSMC's customers?

Short term it helps TSMC — a price hike gives customers more reason to stay, and in a supply-constrained advanced-node market, price increases are rational. Long term it signals intensifying competition, so TSMC must maintain its technology lead to keep pricing power.

Q3: What does 1.6nm mass production mean for phones and AI chips?

The A16 node offers better performance at equal power and higher transistor density than 2nm. It will power next-generation flagship phone chips and AI accelerators — the key node for flagship products around 2027.

Q4: Why did TSMC's stock fall despite record revenue?

The market prices future profit, not current revenue. $85 billion in CapEx raises depreciation and costs, compressing forward gross margins — and investors worry that if the AI spending cycle peaks, massive investment becomes idle capacity.

Q5: Has Arizona's profitability really caught up with TSMC's Taiwan fabs?

Not fully, but the gap is closing fast: 4nm yields now match the flagship Taiwan fab, first-half profit surged 663%, and Arizona is the largest overseas profit contributor. Overseas gross margins still trail the home fabs, which remains the key metric to watch.