Commentaires
Two views, one decision
19 novembre 2026

Over the last two weeks, our emerging markets team met with 30 technology companies across South Korea and Taiwan, covering semiconductors, memory, advanced packaging, testing, power equipment and cooling. Despite operating in different parts of the supply chain, most companies shared a similar message: demand is exceptionally strong, often exceeding available supply, with record backlogs and visibility extending well beyond historical norms.
Strong demand, cautious investors
A few observations stood out. A leading Korean memory maker expects shortages to persist through the end of the decade, while a major Korean electronic components manufacturer indicated that customers are financing more than half of their capacity expansions.
Among our emerging markets portfolio holdings, Chunghwa Precision Test Tech Co. Ltd. (6510 TT), a test interface supplier, is prioritizing higher-margin projects due to capacity constraints and indicated that its new capacity currently under construction is already fully booked. Acter Group Corporation Ltd. (5536 TT), an industrial engineering contractor specializing in cleanrooms, identified a shortage of engineers rather than orders as its main growth constraint.
Despite these strong fundamentals, investors we met during the trip were generally cautious. While 2027 is widely expected to be another strong year, concerns centred on the sustainability of AI-related spending beyond 2027, potential oversupply and whether current valuations adequately reflect these risks. This highlights an important distinction: strong corporate fundamentals and earnings visibility do not necessarily translate into attractive investment returns.
Since ChatGPT was launched in late 2022, AI adoption and infrastructure investment have accelerated significantly, benefiting a broad range of « picks and shovels » suppliers. However, the semiconductor industry has historically been cyclical. AI introduces a powerful new source of demand, but competition, capacity expansion and changes in customer spending still create cyclical pressures.
Capacity brings both opportunity and risk
At the same time, substantial new capacity in leading-edge foundry, advanced packaging and high-end memory has yet to come online. Our holding Eugene Technology Co. Ltd. (084370 KS), a Korean semiconductor equipment manufacturer, highlighted that increasingly complex memory manufacturing processes require more equipment per wafer, providing an additional source of equipment demand beyond growth in wafer production capacity.
Another development is the growing use of long-term agreements (LTAs) and customer prepayments to support capacity expansion. Our holding Kinsus Interconnect Technology Corporation (3189 TT), an advanced semiconductor substrate manufacturer, reported customer-funded equipment investments, and ISC Co. Ltd. (095340 KS), a Korean test socket manufacturer, which we also hold, expects LTAs to cover approximately half of its 2027 volumes. These contracts include minimum purchase commitments and compensation provisions but can be cancelled by mutual agreement. Such arrangements may reduce financial risks and improve visibility, but they do not guarantee future revenue or eliminate cyclical exposure.
A historical parallel
The telecommunications infrastructure boom around 2000 provides a useful historical comparison. The internet transformed the global economy, but many early infrastructure investors suffered from overcapacity and poor returns. The Fed’s monetary tightening in 1999–2000 also preceded the bursting of the technology bubble.
There are important differences today, particularly the stronger balance sheets of hyperscalers. Nevertheless, whether the substantial investments in AI infrastructure will generate adequate returns remains uncertain. Some neocloud providers also rely heavily on external financing. With the Fed having resumed rate increases in September 2026, tighter financial conditions could constrain their expansion or lead to financial distress, with implications for demand across the supply chain.
More capable, lower-cost AI models introduce another uncertainty. Greater efficiency could reduce computing requirements for individual applications, but also encourage broader adoption and potentially increase aggregate demand. Meanwhile, infrastructure spending remains concentrated among a relatively small number of hyperscalers.
AI could transform the global economy while generating disappointing returns for some investors. Strong near-term fundamentals do not rule out a downturn further ahead.
Holding two views at once
F. Scott Fitzgerald once noted that the test of a first-rate intelligence is the ability to hold two opposing ideas in mind simultaneously while retaining the ability to function. This observation is also relevant to investing. Research gives us the conviction to act, but we must remain open to evidence that challenges our conclusions. Even a well-supported investment thesis can be affected by unexpected developments, requiring us to reassess our views and adjust quickly.
We monitor whether easing shortages are accompanied by weaker order conversion, rising inventories or pressure on pricing and utilization. A combination of these developments would prompt us to reassess our investment thesis.
Ultimately, strong earnings do not guarantee attractive investment returns if already reflected in valuations, just as cyclical businesses can become attractive when expectations are excessively pessimistic. We cannot predict when the cycle will turn, but we can continue to evaluate new evidence, challenge our assumptions and adjust our investment decisions accordingly.