Extraterritorial Environmental Accountability of AI Data Centers via Transboundary Harm and Due-Diligence Norms

The bifurcation of artificial intelligence (AI) workloads into training and inference is propelling a global wave of hyperscale data-center construction, yet little is known about how carbon intensity and water stress differentials shape siting decisions, specifically regarding the trade-off between proximity to renewable-energy sources and latency-sensitive fiber connectivity or how the resulting externalities intersect with international environmental law. Drawing on a novel dataset of 247 AI-optimized projects announced between 2018 and 2025, this article couples event-study econometrics with life-cycle modeling to quantify offshore emissions surpluses. Every 100 g CO₂ kWh⁻¹ increase in host-grid carbon intensity raises the probability of attracting new capacity by 8.1 percentage points, while moderate water stress deters investment only when carbon advantages are absent. Median offshore emissions surpluses reach 348 kg CO₂ and 0.49 litres (L) of freshwater per delivered kilowatt-hour, yielding first-year quantifiable carbon leakage of 21.4 Mt CO₂ and 1.78 billion L for the sample volumes that satisfy the "appreciable harm" threshold anchoring the customary duty to prevent significant transboundary damage. Seventy-one percent of these impacts are traceable to companies now covered by the European Union's Corporate Sustainability Due Diligence Directive, yet fewer than one in five disclose project-level emissions, and only 7% report water use. Scenario analysis showed that 24/7 carbon-free procurement, hybrid cooling, and compute-adjusted border adjustments could halve both quantifiable carbon leakage and water-withdrawal differentials. The findings expose a latent governance gap but also chart a feasible path toward extraterritorial environmental accountability for cloud infrastructure.