By Grace Kim

TSMC has reported a sharp rise in profit. The same week, AI-related stocks kept falling. The two facts sit uneasily together, and they frame the central question now facing semiconductor investors: whether the market has stopped believing the earnings or simply stopped paying for them.

TSMC, the world's largest contract chipmaker, said profit surged in its latest quarterly report, according to Capital Brief. The company did not signal deterioration in demand for advanced logic wafers. Its result confirmed that the silicon underpinning AI systems continues to sell at strength. That confirmation, however, did not lift the equities most exposed to the AI buildout.

AI stocks extended their decline after the report. Investors have spent weeks trimming positions across the sector. The pullback has hit names tied to accelerators, memory, networking and the broader AI infrastructure complex. The divergence between TSMC's income statement and the sector's share prices is the story.

Why did strong earnings not lift AI stocks?

The disconnect runs through valuation rather than fundamentals. TSMC converts AI demand into revenue and profit today. Many AI-linked equities price in years of future growth, and that pricing leaves little room for doubt. When sentiment turns, the stocks fall even when orders hold.

There is also a sequencing problem. Investors spent two years bidding up anything adjacent to AI. That trade concentrated gains in a narrow group of large-capitalization names. When those names wobble, indices wobble with them, and the selling spreads to second- and third-tier suppliers whose earnings have not changed.

The profit surge at TSMC shows the physical supply chain still works. Chips ship. Wafers sell. Revenue arrives. What has changed is the multiple the market assigns to expected future AI revenue, not the revenue itself.

What does TSMC's result say about real demand?

A surging profit at the foundry level is a hard data point. TSMC sits at the choke point of AI silicon. Every major accelerator program routes through its advanced nodes. When TSMC earns more, it means customers are buying more finished wafers, not merely talking about future capacity.

That makes the current selloff unusual by historical standards. In prior cycles, foundry strength led equity strength across the chip food chain. Today the link has broken. Analysts read this as a repricing of expectations rather than a demand shock, but the market has not finished testing that view.

The risk, of course, runs the other way too. If TSMC's momentum cools in coming quarters, the stocks that already fell will have fallen for cause. The foundry's forward guidance will matter more than its reported profit.

Who is exposed if the selloff continues?

The pressure points divide cleanly along the supply chain.

• Chipmakers levered to AI accelerator roadmaps face multiple compression even with intact order books.
• Suppliers of memory, packaging and networking gear trade on second-derivative exposure and fall harder on sentiment shifts.
• Hyperscalers funding the AI buildout must defend capital spending plans against investors now skeptical of near-term returns.
• The foundry itself holds a stronger position, because its profit is contracted manufacturing output rather than projected end-market growth.

TSMC's result is the strongest evidence yet that the correction is about price, not product. Chips keep moving through fabs at a pace strong enough to push foundry profit sharply higher, while the paper value of the AI sector contracts around them.

What comes next?

The next test is guidance. If TSMC maintains its demand outlook, the divergence between foundry earnings and AI equity prices will force one of two resolutions: stocks recover toward fundamentals, or investors conclude that even surging foundry profit cannot sustain current spending plans across the AI complex.