TSMC has posted a record third-quarter profit and set out capital expenditure of roughly $98 billion through 2028, according to a TradingView summary of the company's results. One number covers the present; the other covers the next four years. Together they frame the commercial question every customer and rival now faces.

The record quarterly profit is a confirmed result, not a projection. The $98 billion figure is a cumulative spending plan, and plans can change with demand. A reader should hold the two apart, even when the headline does not.

How large is the capex commitment?

The $98 billion covers spending to 2028, which works out to an average near $24.5 billion a year if the money flows evenly. It almost certainly will not flow evenly. Foundry capex tends to bunch around new fabs, new nodes and new equipment installs.

For context, that four-year total approaches the gross domestic product of a small advanced economy. It is a statement of intent, written in concrete and deposition tools.

What does the record profit confirm?

The third-quarter record confirms that demand for leading-edge silicon still pays. TSMC does not earn record profits in a soft market. Whatever is driving the quarter — advanced smartphones, high-performance computing, the general wave of AI silicon — the foundry sits at the narrow end of the funnel.

A record is also a baseline. The next quarter will be measured against it, and so will the spending plan.

Why spend $98 billion now?

The rationale is capacity. Advanced-node capacity is scarce, queues are long, and customers sign long agreements to secure wafers. Capex converts contracts into output. If demand holds, the money compounds the lead.

There is risk on the other side of the ledger. If demand softens before 2028, the same spending becomes fixed cost that weighs on margins. Foundries have lived through that cycle before. The memory is not fond.

Who feels the pressure?

Rival foundries face the arithmetic problem. Matching a $98 billion commitment requires a balance sheet and an order book of similar scale. Few have both.

Customers face a bargaining problem. Record profits at the supplier rarely translate into gentle pricing at the negotiating table. Capacity may expand by 2028; prices are another matter.

What should watchers track next?

The confirmation of quarterly results, quarter by quarter, will show whether the record holds. The pace of actual spending against the plan will reveal whether the $98 billion survives contact with the demand cycle.

TSMC has effectively told the market that it expects the leading-edge boom to run long enough to absorb the largest private industrial bet in semiconductors. The next four years will test whether the customers agree.