Two months on from our February catch-up with Trinetra in February, the backdrop has moved on more than anyone would have liked. Oil is up roughly 37% since late February on the back of the Iran crisis, the MSCI EM index has added another leg on the Korea and Taiwan AI rally after a sharp drop in March. Meanwhile non-AI related emerging markets and the Trinetra portfolio have remained in the doldrums. 

In February we wrote that Trinetra’s “true EM” portfolio was behaving differently than the typical emerging market portfolio of prior cycles, less volatile, better insulated, and compounding earnings at a pace that did not need a cyclical tailwind to work. We flagged the Indonesia MSCI-downgrade threat as noise rather than signal, and we argued the portfolio’s risk-adjusted IRR was as attractive as it had been in years. Our conversation in this week’s What We Are Working On video reinforces that view, with a twist.

Start with the oil shock. Our own bottom-up work puts the portfolio’s beta to MSCI World amid an Iran crisis driven oil shortage and given relative energy dependency at around 1.1x, roughly in line with world equity market, which is perhaps counter-intuitive for an EM-only portfolio given the historical experience.  One reason is the portfolio’s 30%+ Latin American tilt (Brazil is a net energy exporter), its zero weight to Korea and Taiwan (the two most energy-import-dependent markets in the region), and its healthcare and internet cushion in names like Apollo Hospitals and Tencent. In fact, the portfolio dropped around 7% on the energy hit against a fall of 11% for MSCI Emerging Markets and 5% for MSCI World. 

Second, the AI question. It is worth saying explicitly: this portfolio is not AI-light. Tencent, one of the cleanest AI winners in the investable universe, is growing gaming revenue at twice the global industry rate with AI embedded, and its advertising division is growing faster again on AI-driven ad tech. It trades around 14x earnings against Meta at 24x for comparable implementation quality. CATL — 39% share in both EV batteries and grid storage — is a direct beneficiary of the AI and data-centre power build. 

Third, the memory cycle. Trinetra’s Andreas Valanides has covered the sector for years and he observes that the current up-cycle is already nine quarters long, longer than any in recent history,  and that SK Hynix and Samsung are over-earning. Apply long-run ROA and ROE and consensus NPAT implies 40–90%+ downside. The structural case rests on AI capex being a durable demand shift; the cyclical case, on inelastic supply finally catching up. He thinks the balance of risks has tipped.

Finally, the set-up. Crowding-out from Latin America and domestic-consumption EM into the Korea/Taiwan trade has made many EM stocks cheaper but not obviously riskier. This seems to be reflected in the fact that much of EM seems forgotten rather than detested which is reflected in the lack of day-to-day volatility. Meanwhile, Trinetra’s portfolio IRR has risen into the low twenties at roughly 11–12x forward earnings. The post-dot-com rotation into small-cap bricks and mortar value is the clearest historical rhyme: when the rotation came, it came quickly but not before investors' patience had been thoroughly tested.