新兴市场经济展望与策略：人工智能与新兴市场增长分化

Johanna ChuaAC+852-2501-2357johanna.chua@citi.com CITI'S TAKE There are only a few potential EM winners (Taiwan, Korea, Singapore) in theUS-led AI capex cycle, but if the 1990s IT revolution were any guide,prospects for future productivity gains could accrue more to the “users”versus “producers”. Here too, EM looks relatively disadvantaged vs DM, asAI adoption prospects remain highly correlated with per capita GDP throughdigital-infra/skills, and occupational structure’s readiness for AI disruption.China’s lower-cost “frugal stack” could accelerate AI diffusion across EMbut its benefits are likely to be unevenly distributed to a few. It is too early toassess labor-market effects, but abundance of young job-seekers in Africa& Asia is a challenge. AI could narrow the skills-gap between DM and EM,but overall effects are unlikely to be large enough vs displacement andcommodificationof lower skilled tasks.India’s business processoutsourcing services export growth remains very resilient, while Philippines’is slowing sharply. Ernesto RevillaAC+1-212-816-2621ernesto.revilla@citi.com Gina SchoemanAC+27-11-944-0813gina.schoeman@citi.com Yuanliu HuAC+852-2501-2746yuanliu.hu@citi.com Luis E Costa, CFAAC+44-20-7986-9757luis.costa@citi.com The rapid growth of AI technologies could disadvantage most of EM vs DM led byUS.China is the only EM economy in the AI frontier producer stack, but even there,the effects are not large enough to offset domestic demand weakness. If the 1990sIT revolution were any guide, future productivity gains could accrue more to the“users” versus “producers”. Here too, EM looks relatively disadvantaged vs DM as AIadoption (current and prospects) remain highly correlated to per capita GDP throughdigital-infra/skills, and occupational structure’s readiness for AI disruption. Only a small group of EM economies are meaningful potential winners from theUS-led AI capex cycle.Taiwan, Korea and Singapore stand out through their role insupply-constrained (upstream) semiconductors, translating external demand intoproduction, terms-of-trade and investment gains. Lower value-added downstreamIT supply chain in Malaysia, Vietnam, Thailand or Mexico has smaller macroeffects. Data-center investment supports a few small hubs such as Malaysia andThailand but imported equipment and low operational employment constraindomestic spillovers; dividends may come from services exports and domestic AIadoption. AI’s labor-market effects may be emerging first among younger workers andoutsourcing hubs, with increasingly divergent outcomes.While we are still in theearly stages of AI adoption in EM, abundance of young job-seekers in Africa & Asiais a challenge. India’s BPO exports remain robust, potentially supported by deeperskilled-labor pools and expanding global capability centers. Philippine BPO growth,by contrast, has slowed sharply amid its heavier dependence on voice-basedcustomer services, which may be vulnerable to increasingly capable AI agents. What happens to EM if the “AI bubble” eventually bursts?We see threetransmission channels to EM: heightened risk aversion, spillovers from weaker realdemand (not only on AI capex but reverse wealth effects) and financial vulnerabilities(leverage, margin calls). In such a scenario, while US Treasuries are the typical safehaven, their hedging properties to equity market sell offs have arguably weakenedper BIS, creating a potential opening for some "high quality" EM local currencybonds, particularly if (and this is a big if) the shock is large enough to cause a largerepatriation of flows away from the US, resulting in a weakening of US dollar. See Appendix A-1 for Analyst Certification, Important Disclosures and Research Analyst Affiliations. Contents Artificial Intelligence and the EM Growth Divide4Perceptions about AI and EM4Only a few EM “winners” in the US-led AI capex cycle8AI Diffusion and Job Implications in EM10What happens to EM if “AI Bubble” Pops?14EM Strategy – There is only one game in town16Monetary Policy Watch17FX Views18Global Assumptions19EM: Key Macroeconomic Forecasts20Sovereign Bond Ratings Summary21 Uganda79Zambia79Latin America82Argentina83Brazil85CCA87Chile90Colombia92Mexico94Peru96Uruguay98Appendix A-1100 Artificial Intelligence and the EMGrowth Divide Johanna ChuaJohanna.chua@citi.com+852 2501-2357 Perceptions about AI and EM It is hard to ignore how important the AI story has been in driving global growth(via a boost to aggregate demand from US-led AI-related capex) that has beenable to help (partly) offset repeated policy-induced macro shocks, be it the tariffvolatility in 2025, or this year’s ongoing Middle East crisis.While much of thefascination with AI technologies is how this translates into expected earningsmomentum at the firm level,driving significant reflation and dispersion in globalequity market performance,we take a step back and look at a broader question forEM: how will a general-purpose technology like AI impactthe globa