The OECD Forecast Contains a Warning About AI Investment That Has Not Been Priced
The OECD's June Economic Outlook contains a specific risk scenario that equity market valuations have not yet reflected: in the worse-case scenario, where Strait of Hormuz disruptions persist and energy prices remain elevated, global inflation rises by 0.4 percentage points in 2026 and 1.3 percentage points in 2027, and — critically — "investment including in energy-intensive AI would weaken significantly, with increasing risks of financial market repricing." That is the OECD's chief economist Stefano Scarpetta identifying AI infrastructure investment as a specific casualty of sustained energy cost inflation, not as a protected category immune to macroeconomic conditions. Data centers are among the most energy-intensive infrastructure investments in the modern economy, consuming power at densities that make them directly sensitive to electricity cost inflation. The hyperscaler AI capital expenditure plans that the equity market has been pricing as demand certainty are not immune to a scenario where power purchase agreement costs rise faster than projected and the ROI on AI infrastructure deployment deteriorates accordingly.
The Mercatus Center's June 2026 economic situation analysis adds a domestic manufacturing dimension that compounds the international energy risk. U.S. manufacturing employment stood at 12.59 million workers in January 2026, slightly below the prior year — meaning the tariff-driven reshoring narrative that justified significant industrial policy investment has not yet produced measurable employment outcomes. Deloitte's 2025 manufacturing survey found that more than three-quarters of manufacturers cited trade uncertainty as their top concern, and the Mercatus analysis notes that policy uncertainty — not knowing what tariff actions come next — is functioning as a "wrecking bar" on capital investment decisions more damaging than the known tariff rates themselves. The U.S. GDP growth of 1.6% in Q1 2026, while improved from the prior quarter's 0.5%, reflects an economy absorbing large and simultaneous policy shocks — government employment reduction, tariff volatility, and energy cost inflation — with resilience that is genuine but fragile.
Taiwan's 9.6% Growth Forecast Is the Economy That Explains the Rest of the Map
Taiwan's government projects 9.6% GDP growth in 2026 — the highest in 16 years — driven almost entirely by AI chip exports expected to grow 39.8% for the year. South Korea's 139% year-on-year growth in high-end AI chip exports in Q1 2026 tells the same story with a different accent. These are the two economies most directly exposed to AI infrastructure demand, and their growth trajectories are diverging sharply from every other major industrial economy. The UN's World Economic Situation and Prospects mid-2026 update identifies global growth slowing to 2.7%, with growth remaining "weak in Europe" and "high debt and climate-related shocks posing significant risks in Africa." The gap between AI-adjacent economies and the rest is not a cycle-level divergence. It is a structural bifurcation that will persist as long as AI infrastructure investment remains the dominant driver of global capital expenditure growth, and that has specific investment implications for sectors exposed to both sides of that divide.
The industries most acutely caught between these two economic realities are the ones that provide physical infrastructure for AI but are priced as traditional industrials: power generation equipment makers, specialty gas suppliers for semiconductor fabrication, industrial cooling system manufacturers, and precision engineering firms supplying data center mechanical infrastructure. These companies are experiencing AI-driven demand growth in their order books while being valued on traditional industrial multiples that do not reflect the stickiness or growth trajectory of AI infrastructure as a demand category. The OECD's energy price warning introduces genuine downside risk to that demand, but the structural argument — that physical infrastructure for AI has a different demand profile than the broader industrial economy — remains intact for the duration of the current AI infrastructure buildout cycle, which extends well beyond the current energy price shock regardless of how the Strait of Hormuz situation resolves.
The practical investment implication running through this week's macro data is that sector selection within industrials has never been more consequential. The companies supplying physical infrastructure for AI — precision cooling, specialty gases, power management equipment, and data center mechanical systems — are experiencing a demand environment that is decoupled from the broader industrial cycle, but are valued as if they are part of it. Identifying which companies in this category have contracted AI-infrastructure revenue with hyperscaler customers, versus which are exposed to discretionary capex that could slow if the OECD's energy-price warning materializes, is the differentiation question that determines whether an industrial allocation in the current environment outperforms or underperforms the sector index through year-end.