Capacity reservations and inventory buildup reshape component makers' 3Q26 outlook amid persistent shortages

The AI infrastructure boom is driving semiconductor supply chains into structural imbalance. Persistent shortages in advanced packaging, HBM3e memory, and CoWoS capacity are not only delaying GPU shipments but also forcing OEMs to reserve foundry slots 6–9 months ahead—effectively pricing out smaller players. Escalating U.S. export controls and seismic risks in Taiwan, China amplify supply chain fragility, with compliance costs now broadly passed on across the industry. NVIDIA and AMD are fast-tracking Samsung and SK Hynix as alternative sources, while Intel leverages its IFS strategy to lock in long-term cloud provider commitments. Over the next 18 months, component scarcity will worsen due to front-loaded capex and bloated inventories creating artificial demand. A slowdown in AI server orders could trigger a sharp midstream inventory correction, echoing the 2018 memory price crash.