The bottleneck after the chip
AI investment becomes useful only when a chain of physical systems starts working. The next question is who can deliver the connection.
AI & Compute
From model capability to usable capacity.
Research
485 TWh to 950 TWh: an outlook, not a delivery ledger
The IEA’s 2026 data-centre projection needs both its baseline and its uncertainty label.
The AI revenue ladder has missing rungs
A worked Microsoft example separates investment, lease recognition, cloud revenue and the missing evidence of asset utilization.
A chip sale is not yet working capacity
TSMC’s latest results show a growing supplier business. Following the same chips into a working system requires a different set of records.
The grid capacity hidden in plain sight
Better use of existing networks can open a faster route to power. The difficult part is turning technical potential into a dependable connection.
Signals
Power access may control the delivery date
A project-level hypothesis linking energization milestones to usable AI capacity.
Research briefs
From announced capacity to working infrastructure
An evidence protocol for following AI projects across chips, power and capital.
From the Knowledge library
Agentic AI in organisations
Interviews with organisations across regions and sectors examine how AI agents are deployed and governed, including practical benefits and obstacles.
View originalLinear and Nonlinear Econometric Models versus Machine-Learning Models: Evidence from Realized-Volatility Forecasting(Revised)
View originalAI and the global economy: implications for central banks
View originalAI disruption in private credit: exposure to software firms in BDCs
View originalThe OECD AI exposure measure
Maps AI capabilities to occupations to examine exposure across different kinds of work.
View originalKey Questions on Energy and AI
An updated examination of the relationship between artificial intelligence and energy systems.
View originalAI meets trade
Examines trade linkages and how the economic gains from AI may be distributed across countries.
View originalThe agentic AI landscape and its conceptual foundations
Sets out concepts for understanding AI agents and the emerging landscape of agentic systems.
View originalExploring possible AI trajectories through 2030
Explores alternative paths for AI development through 2030 and their policy implications.
View originalDigital Progress and Trends Report 2025: Strengthening AI Foundations — Overview
An entry point to the infrastructure and capabilities that support AI adoption.
View originalAI adoption in the education system
Examines AI adoption in education and the conditions affecting its use.
View originalAI and the global productivity divide
Examines differences across countries in the potential productivity effects of AI.
View originalThe effects of generative AI on productivity, innovation and entrepreneurship
A review of experimental evidence on generative AI, with attention to task differences, user experience and gaps in evidence on long-term business outcomes.
View originalEnergy and AI
A global and regional examination of electricity for AI, alongside AI applications in energy systems.
View originalAI Index Report 2025
The annual AI Index, useful for comparing technological and economic indicators with the previous edition.
View originalDigital Progress and Trends Report 2025: Strengthening AI Foundations
A global assessment of AI readiness covering connectivity, computing capacity, locally relevant data and digital skills.
View originalOECD Digital Economy Outlook 2024, Volume 1
Digital technology developments and policy questions across the economy.
View originalGen-AI: Artificial Intelligence and the Future of Work
Analysis of AI exposure, complementarity and distributional effects. Exposure estimates are different from observed job displacement.
View originalAI Index Report 2024
Cross-cutting reference on AI research, performance, investment and policy.
View original