Some investment trips are memorable because of the conclusions they generate. Others stand out because of the experiences that shape those conclusions. Our recent two-week research trip across Hong Kong, Taiwan and China was very much both.
Download the PDF here
Some investment trips are memorable because of the conclusions they generate. Others stand out because of the experiences that shape those conclusions. Our recent two-week research trip across Hong Kong, Taiwan and China was very much both.
We started off by attending an Asia focused investor conference in Hong Kong. One observation from the conference was particularly noteworthy: China’s surplus energy generation capacity may ultimately become a significant strategic advantage in the AI era, supporting domestic semiconductor development, AI deployment and industrial competitiveness. Combined with China’s position in rare-earth supply chains and increasingly competitive open-source AI models, the argument highlighted how several structural advantages
are converging.
On the ground in Taiwan – The epicentre of the AI build-out
With semiconductors accounting for approximately 56% (As of June 15, 2026) of the Taiwan Stock Exchange Index and technology hardware representing another significant share, virtually every conversation revolved around semiconductors, AI infrastructure or enabling technologies.
What stood out most was the remarkable consistency of the message across the value chain. Whether speaking with foundries, packaging specialists, power infrastructure providers, ASIC developers or memory manufacturers, management teams repeatedly emphasised that demand continues to exceed supply.
At TSMC and throughout the broader semiconductor ecosystem, AI was increasingly described not as another cyclical upgrade cycle but as a structural transformation. The industry now expects the global semiconductor market to surpass USD 1 trillion in 2026, years ahead of earlier forecasts. More importantly, the next phase of growth appears significantly broader than AI training alone.
The first stage of the AI cycle was dominated by training compute. Today, inference demand is accelerating rapidly as models move into commercial deployment. Meanwhile, AI integration into smartphones, PCs and edge devices has only just begun. Several companies argued that AI will ultimately penetrate far more end markets than smartphones ever did. As a result, many industry participants increasingly view AI as a decade-long demand driver rather than a short-lived capital expenditure cycle.
The message became even more compelling within memory and advanced packaging. High-bandwidth memory (HBM) demand continues to absorb industry capacity, creating structural shortages in traditional DRAM markets. Several industry participants expect undersupply conditions to persist through 2026 and potentially 2027, supporting both pricing and profitability. Advanced packaging investments are expanding rapidly, but current plans appear designed merely to keep pace with demand rather than create excess capacity.
Beyond semiconductors themselves, meetings with companies such as Delta Electronics highlighted how AI-related investment is spreading throughout adjacent sectors. Power infrastructure requirements are increasing with every generation of GPUs, creating substantial opportunities across cooling, power management and data centre infrastructure. Similarly, ASIC developers confirmed that hyperscalers continue to rely heavily on external design expertise despite expanding internal capabilities.
The implication for investors is clear: the beneficiaries of the AI cycle extend well beyond the most obvious semiconductor names.
On the ground in Shanghai – Witnessing China’s technological transformation
Conversations with industry experts, equipment suppliers and semiconductor manufacturers pointed towards rapidly increasing penetration of domestically developed AI chips. Factory utilisation remains high, supply shortages persist, and local ecosystems continue to expand despite external restrictions.
The prevailing market concern that increasingly efficient AI models might reduce computing demand was largely absent from discussions on the ground. Instead, the opposite argument dominated. Lower inference costs are enabling broader adoption, creating new applications, and ultimately increasing overall demand for compute resources. The dynamic resembles previous technology cycles, in which falling unit costs spurred significantly higher overall consumption consistent with what is known as the Jevons Paradox.
Beyond AI – China’s move up the value chain
The trip also provided valuable insights into China’s evolving consumer and industrial landscape.
Visits to Huawei’s flagship stores demonstrated how far domestic innovation has progressed. Huawei’s ecosystem now spans smartphones, laptops, operating systems and premium electric vehicles that increasingly compete on technological sophistication rather than price. The integration between hardware, software, AI functionality and autonomous driving capabilities was particularly impressive and highlighted the growing competitiveness of Chinese technology platforms.
Similar conclusions emerged from visits to Chinese consumer brands, including sportswear and apparel companies. The stores, branding and product positioning felt distinctly premium. In some cases, products incorporated advanced materials and specialised technologies that would traditionally have been associated with established international brands.
More broadly, these observations reinforced a key investment theme: China’s move up the value chain is occurring across multiple industries simultaneously. The transformation is no longer limited to semiconductors, batteries or electric vehicles. It is increasingly visible across consumer products, industrial technology and advanced manufacturing.
Investment conclusions
As the trip concluded, two investment conclusions stood out clearly. First, AI demand continues to appear structural rather than cyclical. Capacity constraints across semiconductors, memory and advanced packaging are likely to remain a defining feature of the industry for several years. The investment opportunity, therefore, extends beyond the headline beneficiaries towards the broader ecosystem of memory, advanced packaging, power management, cooling technologies and semiconductor equipment.
Second, China’s technological self-sufficiency is progressing faster than many investors appreciate. Across both industrial and consumer sectors, domestic companies are increasingly competing on innovation, engineering and intellectual property rather than cost alone. This is creating a growing universe of local champions capable of competing at increasingly sophisticated levels.
Ultimately, it came from connecting the macroeconomic narratives with real-world observations. Meeting companies across the value chain and testing our assumptions reinforced our conviction. For long-term investors, the opportunity set may prove significantly broader – and more durable – than current market expectations
suggest.
