S06.6 · Energy, Power & Climate
Development and operation of renewable electricity generation assets, where hyperscaler power-purchase agreements have become a major buyer category.
Renewable power generation and development covers roughly 4,700 GW of installed global capacity in 2025 (IEA Renewables 2025), on annual investment of $670-700B (BloombergNEF). Capacity has grown 13-15% a year, the IEA forecasts a rough doubling to about 9,800 GW by 2030, and China accounts for over half of global renewable capacity additions. The scale numbers are impressive and mostly beside the point for any single developer. What has changed commercially is the buyer: hyperscaler power-purchase agreements have become a major offtake category, reshaping who a developer sells to well before AI touches how development itself is done.
The US, EU and India follow China in capacity additions. Ownership is fragmented — the top 10 developers, NextEra, Iberdrola, EDF and Chinese state majors among them, hold under 20% of global capacity combined — far less concentrated than conventional generation, a legacy of a decade in which new entrants rather than incumbent utilities drove the buildout. Developers sell power through power-purchase agreements to utilities, corporates and traders, and compete directly with conventional generation for grid dispatch.
The economics are capex-heavy, with 60-80% of revenue contracted under PPA and rising merchant exposure beyond that, dependent on tax credits and subsidies (the US Inflation Reduction Act, the EU's Renewable Energy Directive framework), and gated by multi-year interconnection queues in both the US and EU. The queue, not capital or turbine supply, is the binding constraint on how fast new capacity actually reaches the grid in most developed markets. Developers accordingly compete on queue position and permitted land as much as on project economics.
Storage co-location earns its place defensively, managing intermittency and curtailment; green hydrogen and transmission development are the expansionary reach into new offtake markets. Curtailment — power produced but not delivered because the grid or the market cannot absorb it — is a growing drag on merchant revenue in high-penetration markets, which is a large part of why storage co-location has moved from nice-to-have to standard project feature in new development.
Where AI applies to the segment's own work — automating site selection, resource assessment and interconnection-queue modeling — it compresses the development-stage engineering cost line. That is real but secondary. The demand-side effect of hyperscaler PPA buying is already underway, and it is new-net demand for contracted renewable output, not a reallocation from existing utility and corporate buyers. Nor does it spread evenly: the dollars accrue disproportionately to developers with capacity already through the interconnection queue rather than to the segment broadly. AI-native site-development tools are expected to reach broader adoption within 2-5 years.