When diversification becomes an illusion: The hidden risk in Emerging Markets

Emerging markets have been here before - but today's concentration is being driven by a different set of dynamics. A small group of companies has risen rapidly to the top of the market, bringing renewed attention to what is driving these market dynamics and how it compares with previous periods.

Rapid shifts in earnings, valuations and share prices are creating a more nuanced picture than headline numbers might suggest. Looking beneath conventional valuation measures reveals some important distinctions between the companies currently leading the market.

For investors, the key question is not simply how concentrated emerging markets have become, but what sits behind that concentration. We revisit the emerging markets landscape through a value lens and explore what the changing nature of market leadership could mean for investors.

The developed market story

We have noted elsewhere(5) that high concentration of market cap in few names has been a dominant theme in recent years. Pre-COVID we saw high concentration in just three or four stocks in MSCI EM and the beginnings of increased concentration in MSCI World(6) through the growth of the “Magnificent 7”(7). The growth of the Magnificent 7 continued post COVID, with some small fluctuations in the exact names and weights, through to today.

Our paper5 from early 2025 showed much of this, with the top 10 names in MSCI World making up about 25% of the entire benchmark at that time (out of approximately 1,300 stocks). At the same time, the top 10 names in RQI Developed Markets Value Strategy made up little more than 10% of the portfolio and reflected much better Value characteristics (like forward earnings yield) as well. The MSCI World concentration remains to this day.

Five EM Stocks

However, our focus here is the recent and dramatic return to excessive concentration in emerging markets. The top 5 stocks today (Taiwan Semiconductor Manufacturing Company, SK Hynix Inc (8), Samsung Electronics Co Ltd, Alibaba Group Holding Ltd and Tencent Holdings Ltd) make up about 30% of the benchmark, up from 18% at the end of 2023.(9) TSMC alone now accounts for almost 15% of the benchmark (up from 6% at the end of 2023). If we move to the top 10 names, they account for about 35% of the benchmark (note that the MSCI EM benchmark has about 1,200 names). See Chart 1 for the individual stock weights in MSCI EM over time.

To emphasise the point, Chart 2 shows a stacked plot of the same weights. As noted, these 5 stocks now make up about 30% of the benchmark, with TSMC 15% or so on its own. Alibaba and Tencent have shrunk as weights in the benchmark but have still shown strong price growth. Alibaba price is up about 70% from the start of 2025, and Tencent price is up only about 25% from the start of 2025 (having sold off 30% from its high in September 2025).(10)

Note that the dislocation in Chart 2 is due to Alibaba entering the MSCI EM benchmark in mid-2021 when it listed in Hong Kong (it was already listed and trading as an American Depositary Receipt (ADR) in the US).

Chart 1: Weights of top 5 stocks in MSCI EM Jan 2015 to Apr 2026

Just three stocks today

The main story today concerns just three stocks: TSMC (TPE: 2330), SK Hynix (KRX:000660) and Samsung Electronics (KRX: 005930). This is not so much a story of EM benchmark concentration per se as it is a tale of three stocks and the concentration of the EM benchmark in earnings linked to data centre growth and AI. We might indeed say that this is an AI investment concentration issue rather than simply holding only a few large names.

The dramatic growth of SK Hynix and Samsung Electronics started in about September 2025 as the demand for memory chips in data centres started to grow sharply. These memory chips (notably NAND, DRAM and HBM)(11) are only mass produced by several companies: SK Hynix and Samsung Electronics in Korea, and Micron in the US produce about 95% of all DRAM and HBM chips.(12)

Returns on these stocks have been extraordinary indeed: since the start of September 2025, in their respective local currency terms, SK Hynix is up 610%, Samsung Electronics is up 315% and Micron is up 571%.(13)

Chart 2: Stacked area plot of top 5 stocks in MSCI EM

It would be entirely reasonable to think that these price moves have meant large valuation multiple increases, making them look very expensive. In fact, the opposite has happened – these price rises have lagged the earnings growth forecasts, so forward PE multiples have shrunk, making these stocks look cheap despite their recent returns. Chart 3 shows this for SK Hynix and Samsung Electronics.

Current forecasts have this demand continuing for another year at least, and the customers of these companies are seeking contractual surety rather than negotiating on price.(14,15)

Chart 3: Forward PE for SK Hynix and Samsung Electronics

This leads to an important consideration with these two stocks. They are probably best considered cyclical, as they move with economic cycles, in the same way that resource stocks or perhaps tourism stocks might move. In periods of strong demand, their earnings inflate sharply but unwind equally sharply when the cycle turns.(16) Defensive stocks are the opposite of cyclical, in the sense that their earnings are not economic-cycle-sensitive. The dramatic growth in semiconductor demand reflects an early stage in the economic cycle for AI, so SK Hynix and Samsung can rightly be considered cyclical.

Cyclical stocks – like materials or resources – are known to be priced by looking through the cycle rather than on local inflated short-term earnings forecasts. The cyclical nature of these stocks may mean that a low forward PE may be a poor indicator of cheapness and so a cyclical firm that is apparently cheap using forward earnings forecasts may simply be pricing in full cycle earnings.

TSMC is a different story. Based in Taiwan, it makes the underlying chips for Graphic processing Units (GPUs) like those used by Apple, Nvidia, AMD and others – in fact it is their main supplier. As the saying goes “to make money in a gold rush, sell shovels”.(17) It is the world’s preeminent “foundry” for these chips and so is critical to the AI data centre growth. Due to this, it is much more defensive in its nature. It has doubled in value since September 2025, but this is just a continuation of a long-term trend - the value of the firm has quadrupled since the start of 2024 (in local terms).(18) It was already a large firm (6% of the MSCI EM) at the time, and even with overall market growth its overall weight has more than doubled in the benchmark. It became the largest stock in the MSCI EM benchmark near the start of 2021.

Unlike SK Hynix and Samsung, the earnings growth potential of TSMC has been firmly in investors’ minds for many years and recently has been more expensive than the market. In the past, as it grew earnings, it looked much better value. Chart 4 shows this. Note also that MSCI EM has become cheaper recently as well (driven largely by SK Hynix and Samsung Electronics).

These three firms look quite different to RQI EM Value. We do not concentrate in a few large stocks, nor do we concentrate in a particular market theme, like AI (which appears to be the story now). Instead, we avoid these localised stories in construction of our portfolios and focus on the economic size of a company over a long period of time. We do not chase price, or themes, so remain much better diversified.

Chart 4: Forward PE for TSMC and MSCI EM

It is important to note that our approach to Value does not centre on forward PE, although that is one component of our relative value family of alphas. Using forward PE in our charts and discussion is just a useful way to explain the issue at hand.

Longer term

The more recent increase in MSCI EM concentration mirrors the run up in late 2020, when Alibaba, Tencent and TSMC were dominant, with the top 5 stocks representing more than 20% of the benchmark, and the top 10 more than 30%.(19) Chart 5 shows these proportions. The most recent increase in concentration is more significant than that of 2020 – the top 5 stocks alone are now almost 30% and the top 10 are over 35%.

Aside from the very largest stocks, little has changed in terms of weighting. The drive to increased concentration is just with the top few names.

Chart 5: Concentration in MSCI EM since 2015

Finally, and not that this needs extra explanation, but a standard measure of concentration is known as the Herfindahl index, which measures concentration as the sum of the squared values of the data for each sample time. A larger score means higher concentration. Here we simply calculate this by squaring the percentage weights over time. Chart 6 shows this dramatic upward surge, far larger than in 2020.

Chart 6: Herfindahl concentration index for MSCI EM since 2015

RQI and EM Concentration

While we understand completely the need to benchmark against a reflection of underlying markets, and market-cap based indices offer the easiest and most efficient way to do this, many investors will struggle to be comfortable with such levels of concentration as true reflections of markets. Further, these levels of concentration can have detrimental effects on investments which track the benchmark – over-investment in overpriced or AI-themed stocks, inadvertently loading up on momentum as a style(20), difficulty to outperform when market leadership is so narrow. In our view, low tracking error quantitative investments have a much better chance of outperformance than bottom-up stock picking, as the aim is to target maximising active return for active risk taken, and so it is only small tilts about the cap-weighted benchmark that matter.

For those investors who find this concentration concerning, Chart 7 shows that RQI EM Value has had a much lower level of concentration that the MSCI EM benchmark since around the middle of 2019. The weight to the top 10 names in both RQI EM Value and MSCI EM has recently increased sharply.(21)

Chart 7: Top 10 stocks in RQI EM Value and MSCI EM since 2015

As noted above, two of the three largest stocks in MSCI EM have actually grown prospective earnings faster than their price, so forward PE for them has shrunk – making them cheaper using that metric. We can see in Chart 8 Panel A that the same thing has happened in the top 10 names in both MSCI EM and RQI EM Value, and the spread has compressed as well. This is a local phenomenon driven by sharp moves in the average forward PE for just these two stocks, due to the potentially cyclical nature of the earnings for some of these stocks, and how they may not actually be “cheap”. RQI Value looks through this effect and is not as affected.

Chart 8 Panel B shows that this effect is reflected in the full benchmark and RQI EM Value as well, but the spread in value between the two is maintained. While the spreadbetween MSCI EM and RQI Value has reduced recently, as noted above it is due to simply to the two stocks we have discussed and is not systemic. The long-term average spread is around 3% and should be maintained - the investment case for Value is still very strong.

Finally, if we use trailing PE rather than forecast, the price rises have driven “expensiveness” in both, measured this way, and again the spread is maintained. Chart 8 Panel C shows this. This is useful demonstration of the need to examine Value using both forward- and backward-looking metrics. Trailing PE is less susceptible to cyclicality but still shows the spread.

Chart 8 Panel A: Forward PE for top 10 stocks in RQI EM Value and MSCI EM since 2015 (cap-weighted average)

Chart 8 Panel B: Forward PE for full RQI EM Value and MSCI EM since 2015 (cap-weighted average)

Chart 8 Panel C: Trailing PE for top 10 stocks in RQI EM Value and MSCI EM since 2015 (cap-weighted average)

In the case of the three stocks, we have non-zero positions in all of them, but our positions relative to the benchmark are quite different, for different reasons:

As at the end of April 2026, TSMC remains our heaviest underweight, which has grown with recent price moves. While we hold almost 3% of the portfolio in this stock, and have an alpha[1] which favours it due to earnings momentum, its benchmark weight is now over 14%. We remain very comfortable in this position, as our strategy targets long-term mean reversion of expensive stocks. This of course is a result of the benchmark, not of our process.

SK Hynix is also an underweight in RQI EM Value. Note that while this stock appears to have a low forward PE, and so appears to be “cheap”, by all other measures it is not. Its alpha is positive, however. While not as pronounced as TSMC, it has run up far in advance of the weight that we assign to it, now at about a 2.3% underweight. Until late 2024, the benchmark weight was very similar to ours.

Samsung Electronics is different. It has a very long-term and successful track record in business and has a large core weight in our portfolio. This core weight is due to its large long-term business presence, generating a significant economic footprint.23Its market cap has grown strongly but in lock step with our core holdings, so we are overweight even with the recent price moves. Our alpha forecast is also positive.

Year to date, these positions have detracted somewhat from performance, with the semiconductor/memory chip run up we have spoken about above. TSMC and SK Hynix have detracted, but Samsung has been our strongest contributor.

This highlights an important point – RQI does not simply avoid holding the largest (market cap) stocks to reduce concentration. Instead, we hold stocks based on their economic size, a much slower moving concept, and so can be overweight large stocks (like Samsung Electronics) or underweight them (like TSMC or SK Hynix).24

Conclusion

Recent dramatic changes in market concentration in Emerging Markets (EM) has led us to look at its extent and how it affects our investment approach. Our main conclusions are that EM market concentration has returned to extreme levels, higher than even pre-COVID. For our process, concentration has also increased but by less. The Value spread has been maintained.

Two of the largest stocks have received huge upgrades to prospective earnings: SK Hynix and Samsung Electronics. Industry change – the demand for memory chips in AI data centres – is the primary driver, and this is not forecast to abate in the near future. This has meant that the forward PE for MSCI EM and RQI EM Value has actually come down – due primarily to industry changes make both firms look cheap, but in a cyclical way. TSMC (the largest stocks in EM) still looks expensive. RQI Value largely ignores this to focus on economic size, and so can hold very different positions in large (market cap) names.

Finally, as noted in the text, this appears to be much more than a simple benchmark concentration issue due to a few large market cap names. Instead, this seem to be much more a thematic concentration in AI and data centres, creating a run up in prices which can look cheap through one lens but may indeed be a cyclical story, where the market simply prices against long term earnings rather than short term cyclical effects.

Learn more

RQI Investors is a global investment manager, delivering active quantitative equities expertise since 2008. For further insights, please visit our website.