The room marked ‘AI’

As with any investment, your capital is at risk.

 

Market history has a habit of arranging the past like a museum exhibition.

One room for peak oil in the 70s; one for the rise of Japan in the 80s; another for the birth of the internet in the 90s. After a brief stop for coffee, you move on to China’s emergence in the 2000s, before finding yourself in the alarmingly titled ‘Software Is Eating the World' gallery of the 2010s. Every decade has its key theme and now, perhaps, we are standing in the room marked ‘AI’.

Anyone who has wandered through the Smithsonian or London’s Natural History Museum will have experienced the orderly rooms, detailed labels, and a general sense that everything in history unfolded neatly, calmly.

But clearly, markets are not lived and experienced like museum exhibitions. Day to day, they are messy, contested and full of stories that only later become obvious. And more than three decades of investing in emerging markets (EM) has taught us that while one exhibit may draw the crowd (and rightly so), the rest of the building is very rarely empty.

It is an analogy that feels particularly apt today as demand for AI has pulled parts of the EM asset class closer to the centre of global investor attention, leading to greater index concentration as a result. At the time of writing, the top ten stocks in MSCI Emerging Markets now account for roughly 35 per cent of the index, closing in on the S&P 500’s level (38.5 percent). AI heavyweights TSMC, Samsung Electronics and SK Hynix (“the big three”) account for nearly a quarter on their own.

For us, rising index concentration in AI hardware names largely means two key things:

1. As owners of these index giants, we must have a differentiated view from the market on each of them and where their future earnings power may still be mispriced; and
2. We must ensure that the portfolio is not a one way bet on AI, with other exposures that can compound for different reasons.

 

Know why you own the giants

In our clients’ portfolio, exposure to TSMC, Samsung and SK Hynix is greater than the index, as we have an unashamedly positive view on each company. This has been very important for investment performance, both in the near term and the long term.

Our long history with each has undoubtedly helped shape our view. We first held Samsung in 1995, TSMC in 1999 and SK Hynix in 2000. They are companies we have known across multiple semiconductor cycles, management changes, technology transitions, capital allocation mistakes, industry downturns and periods of extravagant market enthusiasm.

Few investors can bring the same length of perspective to all three. But long history is never a reason for complacency: in a concentrated benchmark, the penalty for being wrong in the largest stocks is higher. The first consequence of index concentration, therefore, is that we must be especially clear on why we hold the big constituents (or why we don’t) and where our view differs from the market.

To keep testing our judgement, multiple members of the team have spent time on the ground in South Korea and Taiwan in the first half of this year, meeting each of the big three while also speaking to smaller players, suppliers and customers across the AI supply chain.

Among those we met were companies pushing the boundaries of liquid cooling for chips, businesses creating new ways to connect not only chips but datacentres at scale, and more niche specialists focused on component parts that are critical to making the whole system work. Each meeting is an opportunity to test whether our conviction in our own holdings remains justified as the cycle, competitive landscape and market expectations evolve. Based on the facts today, we believe that each continues to earn its place in our clients’ portfolio.

For TSMC, our view is that the market still underestimates the breadth and durability of its earnings power and competitive advantage. We believe its moat is deeper than process technology alone: it rests on customer trust, yield learning, ecosystem depth and the extent to which customers are embedded in its manufacturing processes. For many customers, moving away from TSMC would be a major operational risk, one which should not be underestimated.

Samsung and SK Hynix are different. Memory remains cyclical, and we do not wish to understate that risk, but we do think the market is still too anchored to an old commodity memory framework. High-end memory is becoming more strategic, more complex and more customised. The nature of contracts is changing in favour of the memory designers. Ultimately, that may support a better and longer earnings profile than previous cycles would suggest.

While the big three are the most visible expression of our AI hardware exposure, they are not the only one. Accton (switches), Fabrinet (optics) and Chroma ATE (testing equipment), for example, are also connected to the same broad infrastructure build out.

Taken together, AI hardware represents around 40-45 percent of the portfolio. That is a significant exposure, but it is also deliberate, reflecting our conviction in the companies and the structural importance of the physical AI supply chain. As these names have strengthened, we have sought to keep the total portfolio exposure in check, trimming and recycling proceeds into the rest of the portfolio that is exposed to very different growth drivers. The result is a portfolio where AI hardware remains a meaningful source of risk and return, but not the sole driver (more on this later).

Even so, we recognise this may not feel like isolated exposure to clients who already own AI elsewhere, particularly through US hyperscalers and semiconductor designers. It is reasonable to treat EM AI hardware as partially correlated exposure to the US AI investment cycle. In an AI disappointment, the North Asian hardware chain would not be immune in the short term.

That said, if we extend this line of thought there is an obvious argument that EM is a ‘cheaper’ option than its developed markets semiconductor counterparts. Valuation starting points are very different today. Samsung Electronics and SK Hynix, for example, still trade on far lower multiples of expected earnings than the US AI leaders, despite operating results that have continued to move ahead of expectations.

Still, over any meaningful investment period, we are not convinced that US AI leaders and EM hardware majors should be expected to move as one block. Firstly, one group is spending heavily to create and capture future demand, the other is adding capacity into bottlenecks where demand is already pressing against supply.

Moreover, in the face of increased concentration in the AI theme, MSCI EM’s latest rolling three-year correlation with the S&P 500 is below its 20-year average and well below its peak. The underlying Magnificent Seven data further argues against treating EM as a simple proxy for US mega-cap technology. EM is materially less correlated with Nvidia, Microsoft and Tesla than with Alphabet or Meta.

While none of this proves independence from US technology, it does highlight that EM AI hardware may be related exposure, but not repeated exposure. And if AI is indeed one of the most important technological shifts of the coming decade, that distinction does matter.

Clients may already own plenty of AI through the US market, but EM offers exposure to a different part of the value chain, often at far lower valuations, through companies whose role is less about promising future applications and more about supplying the physical infrastructure without which those applications cannot scale.

 

Not a one-way bet

If the first discipline is to know why you own the giants, the second is to remember that portfolios are not built on them alone. Market capitalisation tells us what has already become large, but it says less about where the next decade’s earnings growth will come from.

Clearly, no one knows that with any certainty. It may come from AI hardware, but it may also come from Indian infrastructure, Chinese coffee chains and hotpot restaurants, Brazilian fintech, Vietnamese consumption, lithium miners, or other forms of commodity scarcity.

The sensible response is to give the portfolio several ways to be right, rather than one increasingly large answer to one increasingly dominant question. Some of our largest overweights are not driven by AI capex, but by other long term structural trends. As a result, TSMC, Samsung and SK Hynix are not the only, or even necessarily the dominant, source of active risk. Together they account for only around one fifth of portfolio tracking error.

From a portfolio construction perspective, China also plays an interesting role. Recent correlation work done by our Investment Risk Team shows that several Chinese holdings are among the lowest correlation names against the big three.

The list includes businesses such as CATL, Ping An, China Merchants Bank, Tencent, and Meituan. A battery being fitted into an electric vehicle, a family buying insurance, friends messaging each other, a restaurant receiving a food delivery order – the companies catering to these daily interactions are exposed to a very different rhythm from the North Asian hardware in one of the world’s largest economies. It is one reason we have continued to source new ideas from China. More broadly, the point is not that any one of these areas will be the answer, but that the portfolio is built to compound from multiple sources rather than rely on a single dominant theme.

Years from now, the 2020s may well have its own room in the market history museum, and the label on the door may well read ‘AI’. That would be fair. AI matters enormously already; it matters to the EM benchmark, to the portfolio and ultimately to client outcomes. But emerging markets have never been a place where growth arrives obediently from the one direction in which everyone is already looking.

The danger in any well curated retrospective is that the main exhibit makes everything else look incidental. A diversified portfolio built for the next decade should leave room for the possibility that the future, inconsiderately, may have more than one story.

 

---

Risk factors 

The views expressed should not be considered as advice or a recommendation to buy, sell or hold a particular investment. They reflect opinion and should not be taken as statements of fact nor should any reliance be placed on them when making investment decisions.

This communication was produced and approved in May 2026 and has not been updated subsequently. It represents views held at the time of writing and may not reflect current thinking.

Potential for Profit and Loss 

All investment strategies have the potential for profit and loss, your or your clients’ capital may be at risk. Past performance is not a guide to future returns.

This communication contains information on investments which does not constitute independent research. Accordingly, it is not subject to the protections afforded to independent research, but is classified as advertising under Art 68 of the Financial Services Act (‘FinSA’) and Baillie Gifford and its staff may have dealt in the investments concerned.

All information is sourced from Baillie Gifford & Co and is current unless otherwise stated. 

The images used in this communication are for illustrative purposes only.

 

196102 10062752