By Tom Whitfield

TSMC plans to spend $64 billion, and the headline number alone is enough to unsettle some shareholders. That reaction, analysts at Invezz argue, gets the story backwards. Heavy capital spending at the world's largest contract chipmaker has historically matched rising demand, not reckless expansion.

The figure lands at a moment when the whole industry is betting on artificial intelligence hardware. Every new data center GPU, networking chip, and smartphone processor ultimately needs wafers from a handful of advanced fabs. TSMC sits at the center of that chain, and its spending is the clearest signal of how its largest customers read future demand.

Why does $64 billion frighten investors?

Capital expenditure eats cash before it produces revenue. When a company commits tens of billions in a single planning cycle, shareholders see margins compressed and free cash flow shrink in the near term. The concern is simple: if AI demand softens, TSMC could be left with expensive fabs running below capacity.

That fear is not irrational. Chip demand moves in cycles, and past downturns have left producers with too much capacity and falling prices. The memory chip business has lived through several such waves, and investors remember them.

The anxiety, however, rests on a weak premise, as Invezz frames it: that TSMC spends first and hopes for demand later. In practice, the company's spending tracks orders already forming in its order book.

What does the spending actually buy?

The $64 billion covers the physical machinery of leadership: new fabs, advanced packaging lines, lithography and etch equipment, and cleanroom capacity. These assets take years to build, so TSMC must commit money today to serve customers in 2027 and beyond.

The mechanism matters here. Leading-edge chips require capacity booked far ahead, because a modern fab takes roughly two to three years from groundbreaking to volume production. If TSMC waited until demand was visible in quarterly reports, its customers would face shortages and would shift orders to rivals such as Samsung Foundry or Intel Foundry.

Spending, in other words, is defensive as well as offensive. It holds the company's position in a market where customers have few alternatives and where switching costs are enormous.

Why is the capex a bullish signal?

Invezz's core argument is that the spending plan should not deter investors because it reflects strength, not stress. Several forces support that reading:

• AI accelerator demand keeps pulling advanced capacity tight, and TSMC's leading nodes carry the bulk of that volume.
• The customer base — Apple, Nvidia, AMD, and Qualcomm among the largest — has demonstrated willingness to co-fund and pre-book capacity.
• Geopolitical pressure adds funded projects, notably fabs in the United States, Japan, and Germany backed by government incentives.

Each of these factors converts capex from a gamble into a contracted pipeline. Government support in particular shifts part of the bill away from TSMC's own balance sheet, which softens the margin question.

What should investors watch next?

The number that will settle the debate is not the capex itself but the revenue it unlocks. Investors should track utilization rates at advanced nodes, the pace of AI-related wafer orders, and gross margin guidance across coming quarters.

If those indicators hold, the $64 billion will read in hindsight as capacity bought cheaply ahead of a demand surge. If they weaken, the spending becomes a fixed cost the company must carry through a slower cycle — the scenario bears fear most.

Invezz's verdict is direct: the plan shouldn't deter investors, because the spending maps to demand that is already forming rather than to speculation about demand that might never arrive.

The takeaway in plain words: TSMC's $64 billion is a bet its customers have already placed, and the coming quarters of utilization and margin data will show whether that bet continues to pay.