TSMC's Capex Hike: A Tactical Mispricing for Nvidia and Broadcom?

The immediate event is clear: TSMC just posted a blockbuster quarter. The company's fourth-quarter net profit is expected to jump 28% to a record high, beating analyst estimates. This surge follows a market-forecast-beating rise in Q4 revenue of 20.45%, cementing its eighth straight quarter of profit growth. The real catalyst, however, is what comes next. In its earnings call, TSMC's CEO delivered a powerful signal for the AI supply chain: AI-related revenue is set to grow at a compounded annual rate (CAGR) in the high-50% range through 2029.

This guidance confirms the AI boom is not a fleeting trend but a multi-year megatrend. Yet for TSMC's top customers, this is a double-edged sword. To meet this explosive demand, the foundry is planning a massive capital push. TSMC is stepping up its capital expenditures in 2026, forecasting $54 billion in capex for the year at the midpoint of its range, up from about $41 billion in 2025. This is a clear commitment to expand capacity, but the timing creates a near-term constraint risk.

The thesis is straightforward. Sustained AI demand is now baked into TSMC's long-term plan, which is bullish for the entire ecosystem. But the aggressive capex hike also signals that the foundry is playing catch-up. For customers like Nvidia and Broadcom, whose products rely on TSMC's advanced nodes, this means they must now compete for a finite pool of new capacity as it comes online. The record profit is a validation of demand; the capex surge is a warning of future bottlenecks.

The Immediate Impact: Supply Constraint and Customer Risk

The blockbuster profit is a sign of strength, but the real tactical story is a supply shock in the making. TSMC has already told its top customers that it cannot meet their full demand. According to a report, TSMC has told both Nvidia Corp and Broadcom Inc that it cannot provide as much production capacity as they want. This is a critical pivot. For years, the AI narrative was one of insatiable demand outstripping supply. Now, the constraint is shifting from demand to supply itself.

This creates a direct risk for Nvidia and Broadcom. When a foundry says "not enough," customers don't just wait. They start looking for alternatives. The immediate consequence is that these companies may face lost revenue opportunities as they cannot ship the chips they have designed and sold. More importantly, this forces a strategic re-evaluation. As the report notes, "When TSMC says 'not enough,' customers don't pause AI spending. They start looking elsewhere." This is the first crack in the exclusive relationship.

The power structure of the AI trade is beginning to change. For a long time, TSMC was the undisputed gatekeeper, with customers competing for its capacity. Now, with its own expansion plans still months or quarters away from delivering new output, the dynamic flips. The bottleneck favors any foundry with available capacity. This is why Intel's foundry push suddenly becomes relevant. Its pitch of "available capacity, geographic diversification, and political alignment" gains new weight when TSMC is full. The shift is from a demand-driven market to an allocation-driven one, where manufacturing access starts to matter as much as chip design.

Valuation and Forward Catalysts

TSMC's stock popped on the news, but the rally may be pricing in the good while discounting the hard part. The company's fourth-quarter earnings beat and bullish AI revenue forecast are clear positives. Yet its valuation now reflects the massive financial commitment behind that growth. The planned $54 billion in 2026 capital expenditures is a huge outlay, and the payback period for that new capacity will be measured in years, not quarters. This sets up a classic tension: high near-term capex pressure on cash flow versus the long-term promise of AI-driven demand. For now, the market seems to be betting on the promise, but the valuation leaves little room for error.

For Nvidia and Broadcom, the key risk is pure execution. The evidence is clear: TSMC has told both Nvidia and Broadcom it cannot provide as much production capacity as they want. That creates a direct threat to their ability to meet customer demand and ship products. The tactical question is whether they can secure enough of TSMC's future capacity to avoid significant order delays. Any visible slowdown in their own revenue growth or guidance could be the first signal that capacity constraints are biting.

The near-term catalysts to watch are specific and actionable. First, monitor TSMC's next earnings call for any updates on capacity allocation. The company has already signaled a supply constraint; further details on how it plans to manage that scarcity will be critical. Second, keep an eye on Intel's foundry progress. As the report notes, Intel's pitch of available capacity, geographic diversification, and political alignment gains real traction when TSMC is full. Any concrete news from Intel on ramping production for major customers would be a direct counterpoint to TSMC's bottleneck, potentially reshaping the supply chain dynamics for Nvidia and Broadcom. The setup is now about allocation, not just demand.