Coronation Global Emerging Markets Fund

PERFORMANCE

The Fund returned 13.9% during the second quarter of 2026. This strong absolute return was, however, 10.1% behind the return of the benchmark MSCI Emerging Markets (Net) Total Return Index.. The Fund, and active management in general, is experiencing a very tough period as the highly concentrated benchmark continues to be driven by businesses involved in the Artificial Intelligence (AI) supply chain at the expense of most other sectors. We are mindful that the relative return of the Fund is below investor expectations but believe that this period of underperformance is showing unique characteristics that have not played out in over 20 years. The last time the Emerging Markets (EM) universe was driven by a single industry was in the pre-Global Financial Crisis years, when commodity stocks dominated returns, and Brazil and South Africa were responsible for much of the index performance due to the commodity-heavy nature of their markets at the time. Fast forward to today, and the same theme is playing out with AI. In this case, Korea and Taiwan, which house the bulk of the world’s memory and cutting-edge chip capacity, are the two markets driving the EM Index’s return. In contrast, companies from large EM economies with strong reported GDP growth, such as India and China, have now fallen out of favour with investors, only a few years after being highly in demand.

To illustrate this, the charts below show: i) the top 3 contributors to the EM Index over 3, 6 and 12 months and how much just these 3 stocks (TSMC, Samsung Electronics, and SK Hynix) have contributed to the market’s return (between 63% and 78%) over these periods; and ii) the dramatic change in country benchmark weights for some of the markets in question over the relatively short timeframe of 6 years.

VALUATION AND CONCENTRATION RISK

Just 6 years ago, investors couldn’t get enough of China (41% weight in the EM Index), and China-only funds had been, and were still being launched due to a widely-held view that EM managers simply didn’t have enough exposure to China, which in turn was going to become the 2nd largest capital market in the world over time – perhaps one day even surpassing the US. Today’s weighting in the EM Index in very much out-of-favour China, sits at a mere 19%. A very short 1 year ago, India could do no wrong and was the glory child in EM, with concomitant very high valuations. Its weighting in the EM Index was 20% (peaking at 25%) and now sits at less than half that level (11%). In a similar vein, today Taiwan and South Korea combined make up over half of the EM Index (51%), some 20% higher than their combined weighting just a year ago. Like China making up 41% of the EM Index in 2020 and India at 25% in 2024/2025, we don’t view the current situation as sustainable, for the same reason: valuation. We are finding very little margin of safety in most Taiwanese and South Korean stocks today, as was the case with Indian stocks in 2024/2025 and Chinese stocks in 2020.

The next issue is that of concentration risk. If, for example, a fund wants to be ‘overweight’ Taiwan and South Korea, it would mean having more than half of the fund in just these 2 countries given that they account for 51% of the EM Index. Both countries share the same key underlying driver (AI-related capex), and both are arguably above-average risk, in part due to a combination of their history and the fact that their respective neighbours are China (in the case of Taiwan) and North Korea (in the case of South Korea). Putting valuation aside, we wouldn’t be comfortable having half of clients’ capital invested in these 2 countries.

As big an issue, in our view, is the concentration risk that sits at the stock level. TSMC now comprises 15% of the EM Index. Samsung Electronics and SK Hynix combined comprise 17%. The drivers of Samsung Electronics and SK Hynix are essentially one and the same: the memory (DRAM and NAND) markets, particularly their pricing. And once again, the underlying driver of these memory markets today is AI-related capex spend. We would argue that from a risk perspective, besides considering Samsung Electronics and SK Hynix separately on their own merits, one should also consider them together – same end-market, same underlying drivers, same country base. So effectively just 2 stocks (TSMC and ‘Memory Inc’ (Samsung Electronics + SK Hynix) now make up 32% of the EM Index. If one wanted to be ‘overweight’ in these stocks (as many do in these heady days) one would need to have, say, 35% of a fund invested in just these 2 stocks (TSMC and ‘Memory Inc’). Again, this would be taking on a very high level of risk in our view (absolute, not relative) and we are far more concerned about absolute (loss of capital) risk than relative (short-term underperformance of the benchmark) risk.

Almost all our mandates preclude us from holding more than 10% in a single stock. By definition, we are at least 5% ‘underweight’ TSMC (allowed a maximum 10% position vs the EM Index at 15%). We are not convinced that one should actually have much more than 10% in a single stock: the future is uncertain, and adverse conditions can arise. Similarly, each of Samsung Electronics and SK Hynix are now approaching 10% in the EM Index (and 20% combined, which again is how one should be thinking about them from a risk perspective, in our view). Our maximum weighting in these 2 combined (and of course assuming they were very attractively valued, which they are not today, in our view) would be somewhere between 10%-15% of the Fund. Compared to the EM Index weighting of 17%, our combined position in Samsung Electronics (held only through the preference shares) and SK Hynix is around 5% of the portfolio today. This has, of course, impacted performance, particularly over the past 6 months, but our view remains that the 2 shares offer little margin of safety today. They may be on 4x next year’s earnings, but this is simplistic and misleading, in our view. Memory prices have risen to well above any sensible, normalised level, and on more normalised memory prices and hence earnings, that multiple is closer to 20x than it is to 4x. Looking at it differently, SK Hynix trades on 40x 2025 earnings, which only concluded 6 months ago. Memory prices will likely remain high over the next few years due to AI-related demand, but will ultimately come down as more supply inevitably comes online – the same pattern seen before in many other commodities, whether that be oil or copper or iron ore.

There is an argument that ‘this time is different’, and that memory is now structurally a better industry. This is due to HBM (High Bandwidth Memory: what is used in AI) being a more complicated and better business than traditional memory, and due to the possibility of 5-year contracts being signed between the memory companies and their customers. We continue to evaluate this, but at this stage, we would consider these changes (with the latter still being very much unclear) to the industry to be positive, but not industry-changing.

CONTRIBUTORS AND DETRACTORS TO RELATIVE PERFORMANCE

After being a detractor for some time, the Fund’s patience with Delivery Hero was finally rewarded. The share price more than doubled during the quarter, and it provided +2% in relative return. There were a few key drivers of this strong share price performance. The most significant driver was Uber’s acquisition of close to 37% of the company, and a subsequent offer to other shareholders to purchase the remaining stake at €33 a share (the share price on 31 March was €15.45), which Delivery Hero’s board has rejected. Uber may return with a higher offer, but there is also the possibility that key assets could be sold to individual bidders to unlock maximum value for shareholders. Operationally, the company has performed well: in the first quarter, orders reached almost 1bn across the platform, Gross Merchandise Value (GMV) grew 9% on a like-for-like basis, and revenue grew close to 20% (like-for-like). Importantly, all its regions saw decent growth, which suggests broader improvement across all its operations following several quarters during which one or more regions were struggling. Quick commerce now comprises almost 30% of GMV, and if this trend continues, the engagement of customers within Delivery Hero’s ecosystem is likely to increase. Finally, the CEO, who was broadly viewed as underwhelming, gave in to investor pressure and announced that he would step down in due course, with the board beginning the search for a successor.

A few other stocks that had a weak quarter contributed positively to relative performance by virtue of not being held by the Fund or being underweight. Alibaba, an underweight, added +0.4% to relative performance, as did Xiaomi (not owned, +0.4%), Petrobras (not owned, +0.3%) and Pinduoduo (not owned, +0.3%). Two owned stocks that positively contributed to relative performance were ASML (+0.5%) and MakeMyTrip (+0.3%). It is rare for several stocks that are not owned by the Fund to end up being positive contributors, but the concentrated market environment is clearly impacting how returns are generated for active managers.

As one would expect in such a period of negative relative performance, there were several large detractors. The largest of these was Samsung Electronics, which doubled over the quarter and, with the Fund’s 6% underweight, cost 3.4% in relative performance. PRIO (Brazil oil producer) gave back some of its recent gains, costing the Fund 1.7%, while Nu Holdings (Brazil and the broader Latin American banking sector) and AIA (long-term insurance across Asia) each cost 1.5% in relative performance. The ecommerce operators MercadoLibre, Coupang, and JD.com also continued to detract (1.1%-1.4% each). This mix of detractors rather sums up the recent performance cycle of the Fund. Our long-term views (and fair values) on the likes of Nu, MercadoLibre, Coupang, and JD.com are unchanged, and as such each of them are even more attractive now than what they were before.

It’s worth exploring the makeup of the market today and where we go from here. As previously mentioned, the top three weights in the EM Index now represent a third of the overall benchmark, driven in particular by the several-fold increase in the share prices of Samsung Electronics and SK Hynix. The key driver of their short-term earnings has been the rise in memory prices, and the underlying driver of these has been the unprecedented capex cycle of the global cloud players, which hit $800bn this year and is expected to exceed $1 trillion next year.

Memory is an important input into data centres, and for AI, it is HBM that is needed. In order to meet demand for HBM, the memory manufacturers have to pull capacity away from DRAM, leading to acute shortages in DRAM and sharp increases in prices. Margins, and profits have therefore gone up significantly in a very short period of time. Memory was previously around one-fifth of overall semiconductor industry revenue; it is now projected to reach half of industry revenue by 2027, as the other semiconductor players (like TSMC, the single most important stock in the industry) have not raised prices anywhere near the same degree.

With pricing having gone through the roof, memory is estimated to comprise around 35%-40% of the total cost of a data centre, up from mid-single-digit levels in early 2025.

The supply dynamics suggest that shortages will persist into 2028, but we are already seeing early signs of both supply increases and demand destruction. In late June, Samsung Electronics and SK Hynix, alongside the Korean government, announced a combined ~KRW800tn (~$520bn) plan to build new domestic memory capacity. Spread over the coming decade, this represents a step-change relative to the industry’s historic capex on DRAM and NAND, and points to a very large increase in future supply. Chinese players CXMT and YMTC have also announced capacity increases. There is a reasonable chance that the industry eventually goes from chronically undersupplied to oversupplied, and if that happens, the impact on pricing would be severe (and negative). On the demand side, we are seeing evidence of memory being recycled by hyperscalers, and the prices of phones and tablets are rising to reflect more expensive memory, which will weigh on consumer end-demand. It will be impossible to time this from an investment perspective, but we do believe that, on balance, risks are skewed to the downside and have therefore limited the memory exposure in the Fund to reflect this.

More importantly than what we don’t own/are underweight in, we believe the Fund is trading at close to almost once-in-a-lifetime valuations. At the time of writing, the overall upside in the portfolio is 100%, with a five-year IRR of 24% p.a. We have been at these levels before, but rarely has it been so high, with the upside coming from some of the best businesses in EM. Take Tencent, for example: the Fund owns it both directly and indirectly through Prosus/Naspers, yet there is a high level of investor pessimism, driven by a concern that it has fallen behind in the AI race. We do not share this pessimism, based on the proprietary work we have done, and we continually engage with Prosus/Naspers management to encourage shareholder-friendly actions that would narrow the discount at which these 2 stocks trade to the value of the underlying Tencent stake, which, in any event, accounts for only 70% of our estimated NAV for Prosus/Naspers. Astute capital allocation by Prosus/Naspers could deliver investors returns of 30% p.a. from this point at relatively low risk, simply by selling non-core assets, continuing to trim their Tencent stake, and buying back their own shares. Note that since the buyback was initiated 4 years ago, Prosus/Naspers have bought back over 30% of the company: this is significant and would rank amongst the largest buybacks done globally.

Other stocks where the markets seem to be taking a view that they could be long-term losers in an AI world are the ecommerce companies like MercadoLibre (6.6% of Fund), Sea Limited (6.2%), and Coupang (4.1%). These are all very profitable businesses trading at low-teens multiples barely 2 years out, despite having a decade or more of well-above-normal growth ahead of them. In our view, these 3 are amongst the best businesses in EM, and each worth 2-3x the current share price.

PORTFOLIO ACTIVITY

The most significant buying during the quarter was in India, which had de-rated sharply after a prolonged period of pessimism tied to geopolitical and economic concerns. We took the Fund to an overall position of just under 12% in India. This is, in fact, the first time in the Fund’s 18-year history that we have more invested in India than the EM Index. We re-established a holding in Mahindra & Mahindra (0.9%), a turnaround candidate that had been hit by both a weak agricultural season and a broader economic slowdown, and built up Bajaj Finance (0.9%), where earnings growth of over 20% is among the highest of the Indian financials, while adding to the high-quality private-sector banks HDFC (3.6%) and ICICI (2.4%). In Korea, we re-initiated a position in Kia (0.7%) following further share price weakness.

In China, we started building two new consumer positions. Yum China (0.5%), the operator of KFC and Pizza Hut in the country, trades on around 13x forward earnings and offers a long runway of store growth with limited regulatory risk, while H World (0.3%), a leading hotel operator, is benefitting from the margin expansion of an increasingly asset-light, franchise-led model, supported by a strong loyalty programme and a net cash balance sheet that leaves room for shareholder returns. We also re-entered Asia Vital Components (0.6%), a Taiwanese supplier of thermal and cooling solutions to the AI server supply chain, on valuation grounds after it declined.

The Fund also holds a modest direct position in Tencent (0.5%), which we initiated during the quarter, complementing the much larger indirect exposure the Fund holds through Prosus and Naspers, and giving us a cleaner route to what we regard as one of the highest-quality businesses in EM. As discussed above, we do not share the market’s concern that Tencent has fallen behind in AI; in our view, its relatively measured capex reflects a deliberate strategy of leveraging its ecosystem at the right time rather than any lack of capability.

In Q2, we exited two positions entirely. We sold out of Baidu, where our conviction in its AI monetisation path weakened, with the proceeds helping to fund higher-conviction ideas; and we sold Melco Resorts, where governance lapses as well as high financial leverage, subdued top-line growth and ongoing regulatory uncertainty in Macau left better risk-adjusted returns available elsewhere.

The largest additions to existing holdings were in our highest-conviction digital consumer and fintech holdings, funded by trimming stocks that had performed strongly. We added materially to Nu Holdings (6.5%), now one of the Fund’s largest positions, and to MercadoLibre (6.4%), using their share price weakness to build both toward the top of the portfolio, and increased Grab (4.3%) once greater clarity on government commission caps removed a regulatory overhang. Within memory, we continued to reduce SK Hynix on valuation grounds, redeploying part of the proceeds into Samsung Electronics preference shares, which trade at a record-wide discount to the ordinary shares. The main sources of funding for the quarter’s purchases were this reduction in SK Hynix, profit-taking in Delivery Hero (0.9%) after it doubled, and a trim of the Fund’s large TSMC position (9.6%) as we continue to manage overall exposure to the AI supply chain.

Insights Disclaimer
Comprehensive Fund Factsheets