L'essor de l'IA risque d'aggraver les inégalités entre pays riches et pays pauvres, comme le révèle une étude majeure du groupe de réflexion Ayre, qui fait état d'un écart de 360 milliards de dollars

Une étude portant sur 56 pays du Commonwealth estime à 360 milliards de dollars l’écart de gains économiques potentiels liés à l’IA entre les 15 pays ayant atteint le seuil de préparation des jeunes et les 41 autres, sur 10 ans : 389,2 milliards contre 29,2 milliards de dollars. Singapour arrive en tête du classement, tandis que le Malawi et le Lesotho ferment la marche; seuls Maurice, l’Afrique du Sud et le Nigeria ont atteint le seuil parmi les 21 membres africains. L’étude avertit que l’accès à Internet ne suffit pas : compétences, connectivité abordable et débouchés professionnels seront déterminants, et souligne que le chômage des jeunes actifs en Afrique du Sud atteint 46,3 %.

Analysis

The investable signal is not a near-term AI revenue forecast; it is a warning that connectivity is not the binding constraint in many markets. Where power reliability, affordability, foundational digital skills and routes into employment lag, rising AI usage may translate into imported cloud/software spend rather than local productivity gains or durable domestic earnings. That favors established global AI platforms at the margin, while limiting the case for treating telecom coverage or headline youth enthusiasm as proxies for monetizable adoption. It may also shift eventual value toward power, skills and business-software ecosystems—not simply more network capacity.

The study’s 10-year modeled economic gains are too indirect to underwrite country or company earnings, and the stated data gaps weaken cross-country ranking precision. Do not extrapolate the weakest readiness scores into a broad emerging-markets short: low starting capacity can create substantial catch-up optionality if power access, education and employment pathways improve. Conversely, readiness is not sufficient evidence of job creation; weak labor absorption could heighten political pressure for regulation, localization or redistribution.

Over days, expect limited price impact absent policy or funding commitments. Over 1–3 months, watch Commonwealth budgets, power and digital-skills programs, and measurable adoption or employment data. Over 6–18 months, the key test is whether investment improves affordable usage and business productivity, not just infrastructure coverage. The thesis weakens if comparable data show sustained gains in productive AI use and youth employment; it strengthens if access expands without those outcomes.

Key Decisions for Investors

• No standalone trade on this report: treat it as a thematic risk signal, not a validated earnings catalyst. Avoid using the modeled GDP gap to justify country-level valuation changes.
• Watch global cloud and AI-platform providers as potential beneficiaries of imported AI demand, but require company-level evidence of emerging-market revenue conversion before adding exposure; verify regional revenue, customer mix and local infrastructure costs.
• For any EM digital-infrastructure position, monitor power reliability, affordable data usage, digital-skills outcomes and youth employment—not coverage alone. A deterioration in these indicators alongside new capex would weaken the monetization case.
• Set a 1–3 month policy watch for funded power, training and employment initiatives in lagging markets. Evidence of funded execution and comparable adoption data would support reassessing local digital and workforce-enablement exposure; announcements without delivery should not.