At Evertiq Expo Zurich on April 23, Claus Aasholm of Semiconductor Business Intelligence gave an unusually blunt read of the current market: aggregate growth is real, but it is hiding fundamental structural problems that will not resolve on their own. His framing -- a split semiconductor supercycle -- is worth taking seriously.
The materials point is the one that does not get enough airtime. Semiconductor revenue is growing fast, but raw material consumption is not tracking with it. When you double revenue in a compressed timeframe, you need corresponding growth in the physical inputs -- wafers, substrates, packaging materials, gases. That is not happening. Aasholm's read is that the industry is running a balloon phase: inflated margins at the top of the stack (GPUs, HBM) detached from the underlying supply economics. Memory at 80% gross margins is a signal, not a celebration.
The ICAPS situation is a different kind of problem. IoT, communications, automotive, power, and sensors are segments with available capacity and elevated inventory -- and some companies in those segments are still raising prices. Aasholm's explanation is credible: not demand strength, but a lack of visibility causing miscalculation, passed downstream as cost. Meanwhile, those same companies are deliberately shifting capacity toward AI-adjacent products for margin reasons, which is rational at the firm level and dangerous at the ecosystem level.
The automotive and industrial fallout is the piece with the longest tail. These are sectors with multi-year qualification cycles. When a component gets phased out, it does not get replaced quickly -- it gets replaced after years of engineering work and re-qualification. If the component was critical to a system already in production, the OEM is stuck. The industry is trading a reliable, low-margin component supply for higher-margin AI business, and the people paying the price are hardware engineers who need continuity of supply for embedded and automotive designs.