Grid interconnection vs Behind-the-meter generation
Grid service dates come from the host utility's studies, agreements and network upgrades, and they differ by project. On-site generation runs on other clocks: turbine delivery, fuel supply, air permits and protection studies. Earlier power is worth a lot: SemiAnalysis's contested estimate puts AI-cloud contract revenue at ~$12–13B/GW-yr of contracted critical IT, gross, and ~90 GW announced of behind-the-meter generation has been announced.
| Axis | Grid interconnection | Behind-the-meter generation |
|---|---|---|
| Speed to power | host-utility studies, agreement, network-upgrade scope, power rights, and firm-service milestone are project-specific | Enverus's market estimate for a new aeroderivative turbine is ~40 mo (the OEM's dated quote governs a real order); site, fuel, emissions, protection and operating approvals run on separate clocks |
| Cost certainty | known tariffs and PPA structures; basis and shape risk are manageable | fuel price exposure for the asset's life, and the hedging is yours |
| Regulatory risk | large-load rules are changing mid-queue (ERCOT's Batch Zero audit pause; FERC's rejection of PJM's curtail-first interim service for new large loads); budget for rule changes as well as delay | air permits (~60 days-12 months (Region 8 indicative estimates)), co-location disputes at FERC, local opposition |
| Reliability | grid plus backup generation, the classic stack | islanded operation is a serious engineering program (Chapter 4.8), not a procurement line |
| Carbon | can contract 24/7 carbon-free energy and nuclear PPAs (up to ~13 GW announced contracted) | gas keeps combustion and fuel-chain emissions; small-modular-reactor targets need project-specific delivery evidence |
| Endgame | the destination: cheap, firm, and someone else's spinning reserve | usually a bridge: island now, interconnect later, redeploy the turbines |
Quantitative cells are the guide's canonical figures — each is date-stamped and sourced in the numbers register and derived in the chapters below.
How the decision falls
Apply for grid service first: the grid is the cheap, firm endgame. Add behind-the-meter generation as a bridge when it can energize the site years before the grid and the contracts let you redeploy or resell the turbines once the grid arrives. That grid-plus-bridge or co-located hybrid is the structure winning current deals, so underwrite the portfolio rather than a single supply (Chapter 3.4).
What would flip it: A grid date close enough to erase the bridge's head start, or sustained high gas prices, flips it back to grid-only. A utility that cannot serve the load at all turns the bridge into a long-term island, which is an engineering program in its own right (Chapter 4.8).
Model this fork with your own numbers: Project-finance IRR calculator →
Full derivations, worked examples, and the numbers behind this matrix: Energy supply strategy: the four structures (Ch 3.4) · On-site and bring-your-own-power generation (Ch 3.5) · The power-bound era (Ch 16.1)