UN warns about AI becoming another ‘Great Divergence’ between rich and poor countries like the Industrial Revolution

A UNDP report warns that without targeted investment and policy intervention AI gains will disproportionately accrue to wealthy nations, creating a potential 'Great Divergence' as poor regions lack skills, reliable power and connectivity. The analysis highlights economic and social risks — job displacement, exclusion from digital IDs and payments, heavy electricity/water demand with emissions implications, and cybersecurity/deepfake threats — and urges investment in digital infrastructure, regulation, education and social protections to democratize AI benefits.

Analysis

Market structure: AI reinforces concentration: hyperscalers (NVDA, MSFT, AMZN, GOOGL) and data‑centre REITs (DLR, EQIX) will capture outsized revenue and pricing power from GPU, cloud and colocation demand; expect 15–30% upside in consensus revenue for these names over 12–18 months if AI adoption continues. Suppliers of power, cooling and semiconductors will see step-up capex; this tightens supply for GPUs/TPUs and high-efficiency servers, pushing component lead times and premium pricing for 6–12 months.

Risk assessment: Key tail risks include rapid regulatory fragmentation (EU-style AI rules + export controls) that could remove 10–20% of addressable markets for certain models, large-scale cyberweaponization of AI leading to sudden capex pullback, or a power‑supply shock raising operating costs by >20% for data centers. Immediate risks (days–weeks) are volatility around policy announcements; medium term (3–12 months) are GPU supply cycles and utility constraints; long term (1–3 years) are social pushback and unequal adoption dampening TAM growth.

Trade implications: Prioritize core longs in cloud/AI infra and cybersecurity, overweight DLR/EQIX but hedge power risk with renewables exposure; use options to cap downside around earnings and policy windows (3–9 months). Underweight or avoid speculative “AI” small caps lacking >15% revenue growth and positive EBITDA—these are high default/rehypothecation risk if funding tightens.

Contrarian angles: Consensus undervalues the regulatory-cost channel and energy bottlenecks—buy the “enablers” of distributed AI (edge compute, micro‑data centers, grid upgrades) which could compound returns if on‑prem/edge adoption accelerates (18–36 months). Conversely, beware crowding in NVDA-equity; if GPU pricing normalizes, expect mean reversion of 15–25% in richly priced winners—use options to monetize that risk.

Key Decisions for Investors

• Establish a 3% portfolio long in NVDA (1.5%) and MSFT (1.5%) to capture AI infra growth; target +35% combined return over 12 months, take profits if combined gain >40% or cut to -20% loss.
• Allocate 2% to data‑centre REITs: DLR (1%) and EQIX (1%) for exposure to colocation demand; hedge 0.5% of portfolio by buying 6–12 month puts on DLR if implied vol <50%; hold 12–18 months.
• Add 1.5% risk allocation to cybersecurity: buy CRWD 6‑month 30‑delta calls sized to equal 1% notional and purchase PANW stock 0.5%; exit if sector order‑flow reverses or M&A premium compresses by 10%.
• Reduce/avoid small‑cap “AI” names lacking revenue growth >15% and positive EBITDA; if held, trim to <1% position size and liquidate if cash runway <12 months or funding spreads widen by 200bps.
• Implement portfolio protection: buy a 6‑month NVDA 25‑delta put spread sized to hedge 1.5% NVDA exposure (cost target <1.5% of portfolio) ahead of major policy/EU AI Act milestones expected within 3–9 months.