GAM: How Asia is powering the age of Artificial Intelligence

Ygal Sebban, Investment Director, Emerging Markets Equities at GAM

While the debate around artificial intelligence is often centered on the United States, some of the most important technological advances driving the AI boom are increasingly coming from Asia. The region has become an integral part of the global AI ecosystem, thanks to its world-class expertise in semiconductors, integrated circuit design, and a rapidly expanding software landscape.

The growing influence of emerging market technology is also reflected in market representation. The technology sector now accounts for 29% of the MSCI EM index, up from 21% in mid-2023 and 14% in 2018 - even surpassing the tech weighting of the MSCI World index (26.5%).¹ The key upstream hardware leaders - TSMC, SK Hynix, and Samsung Electronics - together represent nearly 20% of the emerging markets index.

Manufacturing at the Core of the AI Narrative - TSMC

In its recent fourth-quarter earnings briefing, Taiwan Semiconductor Manufacturing Company (TSMC) raised its five-year AI growth forecast from 40% to 50% and increased its 2026 capital expenditure outlook to USD 52-56 billion, above market expectations of around USD 48 billion.²

CEO Che-Chia Wei emphasized that discussions have expanded beyond direct customers, highlighting tangible commercial benefits and financial returns from AI adoption. TSMC is aligning its long-term capacity strategy accordingly.

From General-Purpose Chips to ASICs - MediaTek

AI is accelerating the shift from general-purpose semiconductors to application-specific integrated circuits (ASICs). Google’s Gemini 3 models - powered exclusively by internal TPUs rather than Nvidia GPUs - illustrate this trend.

Asia is home to leading ASIC developers, including Taiwan’s MediaTek, where ASIC-related revenues are expected to surpass traditional smartphone revenues by 2027.³ With 2026 likely to be a transition year, valuations appear supported at current levels, ahead of expected growth starting in 2027.⁴

Architecture Over Transistors - ASE Technology

For decades, semiconductor progress was driven by a simple principle: more transistors meant better performance. As chips shrank to nanometer scales and became more densely packed, processing power increased.

Today, however, performance depends increasingly on advanced memory integration and heterogeneous packaging. This shift elevates the importance of companies specializing in advanced packaging and testing.

ASE Technology, a global leader in outsourced semiconductor assembly and testing (OSAT) and electronic manufacturing services (EMS), is well positioned to benefit from rising demand for high-performance integration, alongside limited advanced packaging capacity at TSMC.

Strengthening the Memory “Supercycle” Narrative - SK Hynix & Samsung

We have been optimistic about memory since mid-2023, but the outlook has improved significantly since late 2025. Market conviction in a “memory supercycle” has strengthened, particularly for leaders like SK Hynix and Samsung Electronics.⁵

Two key drivers support this trend:

• Large-scale memory demand
• Memory becoming a critical bottleneck

As AI workloads shift toward longer-context inference, memory - rather than compute power - is increasingly becoming the limiting factor in system performance.

Applied AI and Physical AI - Hesai

A year ago, investor focus in tech shifted from upstream segments to midstream hardware. Early 2026 results now suggest that this momentum is moving further downstream toward software-driven applied AI and the rapidly expanding field of physical AI.

At CES, several major tech companies showcased new robotics platforms. One example is Hesai, a Chinese tech company listed in both the United States and, more recently, Hong Kong. As a global leader in LiDAR (light detection and ranging) sensors, essential for autonomous driving, robotics, and a wide range of physical AI applications, Hesai represents a classic “pick-and-shovel” strategy⁶ - providing critical infrastructure to a rapidly expanding ecosystem.

Risks

There are growing concerns about returns and potential bubble characteristics in GenAI. A July 2025 MIT report estimated that around 95% of GenAI projects are failing⁷, while valuations across parts of the tech ecosystem remain elevated.

At the same time, memory shortages are putting pressure on industrial and consumer electronics manufacturers, particularly PC and smartphone producers, due to limited supply and rising component costs, negatively impacting both volumes and margins.

Stay Invested, but Be Selective

Despite the hype, we believe it is too early to abandon the GenAI theme, but selectivity is essential.

In our view, the winners will be companies with:

• Clear leadership or defensible niche positions within the AI value chain
• Disciplined capital allocation
• Reasonable valuations
• Attractive long-term earnings trajectories

We believe that established leaders within the AI value chain are well positioned to benefit from the accelerating adoption of this technology and its increasingly visible economic impact.

Sources

1. Source: Bloomberg, as of December 31, 2025.
2. Source: Bloomberg, TSMC, “Q4 2025 TSMC Earnings Call”, January 15, 2026.
3. Source: Reuters, October 31, 2025.
4. Consensus estimates according to Bloomberg for 1-year forward / 2027. The price/earnings ratio of 18.7x is well within the historical range observed from 2020 to 2025. (Source: Bloomberg, December 2025)
5. SK Hynix: Market Outlook for 2026 - “Focus on the HBM-driven memory supercycle”, January 5, 2026.
6. A “pick-and-shovel” strategy refers to investing in the tools and infrastructure of an industry rather than its end products, based on the idea that during the California gold rush, sellers of picks and shovels often profited more than miners.
7. Source: Forbes, “MIT finds that 95% of GenAI pilot projects fail because companies avoid friction”, August 26, 2025.