When Scarcity Works Backwards: Rate Hikes Can't Fix Capacity Constraints

Energy disruptions and logistics bottlenecks have redrawn the economics of modern manufacturing into a pyramid where winners sit not at the top of the chain but at the points of physical constraint. For the past three quarters, global inflationary pressure has remained a puzzle: demand growth, while strong, does not clearly justify the breadth of price increases across metals, shipping, semiconductors and energy. The mechanism driving inflation is no longer demand outpacing supply in the classical sense. It is that supply is broken into segments, and each segment is broken into pieces small enough that buyers cannot move freely between suppliers. The companies benefiting most are those whose customers have nowhere else to go. Consider the semiconductor industry, where Taiwan Semiconductor Manufacturing Company and Intel stand at opposite ends of the capacity question. TSMC's 3nm process reached full capacity utilization in 2026, with the company locked into a 25% to 30% revenue growth forecast through the end of the year. Its dominance is not a result of lower costs or better performance on a per-wafer basis; it is a result of being the sole source that can produce enough chips fast enough for major cloud infrastructure investments. Customers sign long-term, high-volume agreements not because they prefer TSMC's economics but because they must. Wafer output across TSMC's manufacturing footprint is at the edge of sustainable capacity. That fact translates directly into pricing power, independent of market conditions. Intel's 18A node, by contrast, has become a production success story that illustrates the opposite dynamic. The company ramped 30,000 wafers per month of 18A capacity across two facilities, in Phoenix and Hillsboro. Q2 2026 output was 25% above internal targets. Yet Intel's foundry business remained in the red, losing $2.1 billion before interest and taxes in the second quarter, and yields on 18A are only now reaching commercially viable levels in the third quarter. Intel's abundance of capacity, relative to proven yields, has flipped the traditional supply-demand relationship on its head. The company cannot charge premium prices because customers know a competing supply exists. At the same time, internal production demands on the same fabs are forcing Intel to shift 18A wafer capacity away from external customers back to its own Panther Lake processors. Intelligence of which the market is acutely aware. External foundry revenue was only $293 million in Q2, a fraction of the $5.8 billion foundry total, meaning TSMC's pricing power over Intel's has become asymmetric. The same pattern emerges in steel. US raw steel capacity utilization hit 88.9% for the week ending October 31, but the figure masks a radical divergence in regional price and availability. Nucor, the largest US steelmaker, raised its hot-rolled coil pricing from $1,240 per ton in mid-October to $1,280 per ton in early November, a $40 swing in three weeks. Three of those weeks saw $15 moves followed by a $10 move, suggesting the company was testing buyer resistance on each price point. The increases came not from volume growth but from the physical fact that certain mills and certain product grades had sold ahead several months. Nucor's backlogs now extend into mid-2027 for joist and deck products. The company's West Virginia sheet mill, the largest capital project in its modernization cycle, is entering commissioning with first coils expected in December and a 50% utilization ramp target by year-end 2027. That mill will add massive capacity but cannot supply orders for six to nine months. Until it does, customers needing specific grades and delivery windows have nowhere to go. Pricing reflects that immobility, not the state of underlying demand. Container shipping has entered a comparable trap. Maersk raised its 2026 profit guidance for the third time this year, despite loaded volumes falling 0.8% year-over-year. The company's average freight rate r