TSMC Arizona's Profit Rose 662.8% — and About 90% of It Was Not From Making Chips

TSMC's Arizona operation earned NT$36.066 billion in the first half of 2026, up 662.8% from a year earlier. It is a striking number, and it is mostly not what it looks like. Roughly 90% of the first-quarter figure was investment income rather than profit from making chips.

What investment income means here

Investment income is money earned on financial holdings — interest, gains on securities, returns from stakes in other entities — not margin earned by manufacturing and selling wafers. A fab that has raised and parked very large sums of capital ahead of construction generates a lot of it. It is real money and it belongs in the accounts, but it says almost nothing about whether the factory is a good factory.

Strip it out and the picture is far more modest. Of Q1's NT$18.807 billion, NT$16.909 billion was investment income; in Q2, NT$14.603 billion of NT$17.259 billion, or about 85%. The operating contribution is the small remainder in both quarters, which is what you would expect from a site still ramping.

The second-quarter number is the more honest one

Arizona's Q2 profit fell 8.2% from Q1 to NT$17.259 billion, even as it stayed 307.8% above the same quarter last year. Investment income fell 13.6% over the same period — so the sequential decline is substantially the financial line normalising, not the fab faltering.

The headwind management points at is depreciation. A fab is an enormous fixed asset that must be written down over its useful life, and those charges land whether or not the machines are busy. TSMC's finance chief has guided that overseas expansion dilutes group gross margin by 2 to 3 percentage points in the early stages, widening to 3 to 4 points as the fabs mature — the cost of manufacturing outside Taiwan, stated plainly.

Why Arizona still matters

TSMC's four overseas units together earned NT$58.529 billion in the first half, up 215.4%, and Arizona is roughly 62% of that. The first Arizona fab has been in 4nm mass production since late 2024; a second, focused on 3nm, is scheduled for the second half of 2027.

So the honest read of a 662.8% headline is this: Arizona has flipped from losing money to making it, which is a genuine milestone for a site of this cost and political significance. But the profit is currently dominated by financial income, the operating economics are still diluted by depreciation, and the test of whether US manufacturing works for TSMC arrives when the second fab is running and the investment income is a smaller share of the total.