Behind The Meter (BTM): Solving the Interconnection Queue
Grid connection timelines are pushing 4+ years. We analyze the viability of co-locating training clusters directly with nuclear and solar generation assets.
Global Scale Research
The single biggest bottleneck in AI today is not chips; it is transformers and substations. Connecting a new 100MW load to the public grid can take 3-5 years due to regulatory studies, transmission upgrades, and supply chain delays for high-voltage equipment. The solution for the impatient hyperscaler? Go Behind The Meter (BTM).
What is BTM?
In a BTM scenario, the data center is built physically adjacent to a power generation asset (a power plant, solar farm, or wind farm). It connects directly to the generation busbar, effectively bypassing the public transmission grid. The data center consumes the power before it ever hits the utility meter.
The Nuclear Renaissance
Nuclear is the holy grail for BTM. It provides 24/7 baseload power, carbon-free, at massive scale. The recent acquisition of the Cumulus data center by AWS — which is connected directly to the Susquehanna nuclear plant — proved the model.
This deal sent shockwaves through the market. We are currently tracking multiple nuclear sites in the US and Europe actively seeking data center tenants for adjacent land. The proposition is compelling: 960MW of potential capacity with zero carbon footprint and zero grid queue delay.
The Risks & Challenges
Regulatory Fury: Grid operators and regulators (FERC in the US) are concerned. If baseload power is diverted to data centers, it removes stable capacity from the public grid, potentially raising rates for consumers and threatening reliability. Expect fierce regulatory battles ahead.
Complexity: Structuring a BTM deal is infinitely more complex than a standard utility connection. It involves PPA (Power Purchase Agreement) structures, backup power considerations (what if the reactor trips?), and massive land use negotiations.
Despite the hurdles, for gigawatt-scale ambition, BTM is the only path that aligns with the AI deployment velocity.