Recently, a clear signal worth paying attention to—TSMC's capital expenditure in 2026 has hit a record high, reaching $52 billion to $56 billion, a 30% increase compared to this year. What does this number imply? Advanced chips and storage chips (DRAM, NAND Flash) are expanding production, and market contract prices are expected to rise by 33% to 60%. This wave of expansion is cascading upstream step by step, benefiting the most directly are equipment procurement, cleanroom engineering, packaging, and testing segments. These will see the fastest performance realization.

Regarding opportunities for domestic substitution, semiconductor materials are also gaining momentum—photoresists, electronic special gases, and core components have been digested by the short-term market, but their long-term growth potential remains promising, with barriers firmly in place. Leading cleanroom engineering companies are directly benefiting from large-scale construction demands of wafer fabs and storage factories; suppliers of core components like high-purity gas pipelines and vacuum chambers hold orders from top domestic and international equipment manufacturers; photoresist companies are moving from mid-to-low-end to high-end (KrF, ArF). In packaging materials, the dominant positions of lead frame and bonding wire companies determine their bargaining power in the industry chain. Advanced packaging-focused companies are deeply tied to the demand for storage and computing chips.

The second major certainty comes from the architectural upgrade of AI servers. Moving from single nodes to racks and clusters, network architecture must also upgrade, evolving from 400G to 800G and even 1.6T. The core market contradiction now is no longer about competing for orders but ensuring delivery. Capital is more favoring upstream core technologies and those "shortage" segments with tight capacity. The performance growth of optical modules has already been partially realized, with valuations reaching high levels for speculative play, but core optical chips (like lasers), specialized optical components (such as optical rotators), and high-end PCB materials (CCL, copper foil)—these higher barrier, supply-demand tight items—are becoming new certainties.

In optical modules, the global "dual giants" are the key performers. Companies that have successfully entered the 1.6T optical module supply chain are also involved in comprehensive AI hardware deployment. National teams are accelerating their layout of high-end optical modules like 800G/1.6T and cutting-edge technologies. Key components for high-density optical interconnects inside data centers directly benefit from AI data center construction. On the upstream core chip side, domestic leaders and breakthroughs in high-speed optical chips are critical for optical module upgrades. Currently, the global leader in Faraday rotators is in a state of supply shortage, and is also a key supplier of MEMS mirrors in optical circuit switching technology. Platform-based optical device companies provide one-stop solutions for optical module manufacturers and are deeply involved in next-generation technologies. Equipment suppliers expanding capacity through global optical module expansion are also riding this wave of benefits.

The third certainty comes from the critical point of robot industrialization. Humanoid robots are officially moving from R&D to large-scale commercial deployment, with giants like Tesla setting mass production targets (100,000 units by year-end), greatly boosting industry confidence. What is the biggest bottleneck in initial mass production? Cost reduction and delivery of core components. Therefore, suppliers of joint modules, dexterous hands, reducers, and servo motors have the highest performance elasticity and strongest order certainty. Harmonic drive leaders are already key suppliers for some giants; domestic RV reducer leaders are also capturing market share; robot component suppliers, servo motor-related parts companies, structural parts, or motor component enterprises are all directly benefiting.