Land Is the Cheapest Input in the AI Power Stack
Land is the cheapest input in the AI power stack. The scarce resources are the pieces that determine whether a data-center project can actually become energized capacity.
The binding constraints are tightening:
- New heavy-duty gas-turbine orders run about three years. GE Vernova has roughly 10 GW of slots left across 2029 and 2030 combined.
- – Large power transformers now exceed 160 weeks of lead time, up from about 50 weeks in 2021.
- – PJM projects that generators coming online in 2025 averaged more than seven years from queue entry to operations.
That timeline changes the investment question. A developer can release acreage with a signature, but a signature cannot create a turbine, a transformer, or an interconnection. The projects that matter are the ones with secured equipment, power, and a credible path to energization.
For investors, the implication is to look beyond the headline size of AI demand. The durable opportunity sits with the suppliers and infrastructure owners that control the bottlenecks: grid equipment, generation, interconnection, and the land tied to usable power.
