CPO unlikely to shake high-end CCL demand in the short term! Morgan Stanley: Limited impact before 2028, supply tightness remains unchanged.

Morgan Stanley released its latest Greater China technology hardware industry report on September 17, stating that although co-packaged optics (CPO) may reduce demand for high-end copper clad laminate (CCL) in the long term, the technology is not yet sufficient in the short term to pose a substantial impact on the high-end CCL market. Considering that CPO still faces numerous challenges in areas such as cost, manufacturing yield, reliability, maintainability, and thermal management, the bank expects that the impact on high-end CCL demand from 2026 to 2028 will be limited, and that AI systems' demand for high-performance CCL and the tight supply situation will continue. The report noted that on September 17, shares of Taiwan high-end CCL manufacturer Taiwan Union Technology fell 7.7%, while Elite Material fell 6.7%, even as the Taiwan Weighted Index rose 0.8% during the same period. After the sharp pullback in both stocks, the market refocused on a key question: as CPO technology accelerates its adoption in AI infrastructure, could demand for high-end CCL be undermined? CPO sparks market concerns: Will shorter high-speed electrical signal transmission distances reduce high-end CCL usage? This concern mainly stems from changes in the CPO architecture itself. In traditional AI server and switch architectures, high-speed electrical signals need to travel relatively long distances through printed circuit boards (PCBs) between chips and optical modules. As data transmission rates continue to increase, PCBs need to use high-end CCL materials with ultra-low dielectric constant (Dk) and ultra-low dissipation factor (Df) to reduce signal loss. The core idea of CPO is to integrate the optical engine more closely with the ASIC chip near the package substrate, thereby significantly shortening the high-speed electrical signal path that originally needed to pass through the PCB. In theory, this could reduce the system's reliance on ultra-low Dk/Df high-speed CCL, and therefore the market worries that as CPO penetration increases, the per-unit value of high-end CCL in AI systems could decline. However, Morgan Stanley emphasized that this is not a new industry controversy. The bank had previously discussed the potential long-term impact of CPO on high-speed CCL demand and believes that CPO is indeed a long-term risk worth watching for the high-end CCL industry, but not an imminent demand shock. Large-scale CPO deployment still faces multiple obstacles; impact limited before 2028 Morgan Stanley believes the market may currently be underestimating the engineering challenges facing CPO's large-scale commercialization, especially in scale-up architectures within AI clusters. The report pointed out that broader adoption of CPO still requires solving a series of issues, including cost, manufacturing yield, reliability, maintainability, and thermal management. Therefore, even if CPO has the potential to reduce high-end CCL usage in the long term, before 2028 the risk of it broadly weakening CCL demand across the entire AI system remains limited. This means that within the roughly two-year window ahead, the rapid expansion of AI computing infrastructure will remain the main DRIVE for high-end CCL demand, and the potential material substitution effect brought by CPO is temporarily insufficient to change this trend. Morgan Stanley therefore maintained its positive view on high-end CCL fundamentals. The report explicitly stated that it currently sees no obvious impact from CPO on Taiwan Union Technology's and Elite Material's near-term CCL demand or earnings forecasts, and that its previous judgment that high-end CCL supply will remain tight has not changed. CPO is a long-term risk; in the short term it is still difficult to change the high-end CCL supply-demand landscape From a longer-term perspective, Morgan Stanley did not deny the structural impact CPO may bring to the high-end CCL industry. As CPO technology gradually matures and the optical engine moves closer to the ASIC, the distance over which high-speed electrical signals need to be transmitted through the PCB decreases, and the per-unit usage and specification upgrade demand for high-end CCL could in theory be affected. Therefore, the bank acknowledged that broader adoption of CPO is a credible long-term risk to the value content of high-end CCL. However, the report clearly distinguished between "long-term technological risk" and "demand trends over the next two years": during the forecast period from 2026 to 2028, the pace of CPO penetration will not yet be sufficient to significantly change the outlook for high-end CCL demand. In other words, the long-term technological substitution logic that the market is trading has not yet translated into actual short-term order or earnings pressure. It is worth noting that Morgan Stanley maintained a neutral view on the Greater China technology hardware industry as a whole. At the same time, among the companies covered in the report, as of September 16, the bank maintained an "Overweight" rating on Taiwan Union Technology and also maintained an "Overweight" rating on Elite Material.