Who Pays for the Grid? The New Cost-Allocation Fight Around AI Data Centers
The AI power debate is moving beyond whether enough electricity can be built. The next question is who pays for the infrastructure required to deliver it. As hundreds of megawatts of new demand drive generation, transmission and distribution investment, utilities and regulators are building new rules around cost causation, minimum payments, collateral, stranded investment and ratepayer protection.
A 500 MW data center can change the system built to produce and deliver electricity. Generation, transmission, substations, transformers and reserves may all need investment years before the customer reaches full demand.
The next power constraint is becoming a cost-allocation question: which infrastructure belongs to the data center, and which infrastructure belongs to the grid?
One campus can create costs across the power system
01
Generation
Incremental energy and peak capacity
02
Transmission
Regional reinforcement
03
Substation
Transformation and switching
04
Distribution
Local service upgrades
05
Reserves
Capacity and reliability
Cost causation meets system benefit
Dedicated facilities are relatively easy to assign. Regional upgrades are harder: a project may trigger them while the completed infrastructure improves reliability, reduces congestion or supports later customers. Regulators increasingly have to reconcile both principles.
Interactive allocation toolkit
What problem does each mechanism solve?
Cost causation
Direct Assignment
Assign identifiable customer-specific facilities directly to the load.
Dedicated substations and site extensions
Shared network upgrades can benefit others
Virginia's GS-5 framework illustrates the emerging toolkit
14 years
Minimum contract
85%
T&D demand floor
Up to 60%
Minimum-charge collateral
GS-5
Separate rate class
Customer-specific cost versus system cost
More directly attributable
Customer infrastructure
- Dedicated substation and site extension
- Customer equipment and direct connections
- Cancellation and stranded-cost exposure
Potentially shared
Broader system infrastructure
- Regional transmission and generation
- Reliability upgrades with multiple beneficiaries
- Capacity that supports future growth
Flexibility can reduce the bill before allocating it
Non-firm service
Accept lower service priority
Onsite generation
Reduce constrained-hour withdrawals
Battery storage
Shape short-duration peaks
Compute flexibility
Shift selected workloads
Bottom line
The nominal electricity rate is becoming less informative. Minimum demand, collateral, contract duration, buildout reimbursement, transmission riders and curtailment rights can determine the real commercial exposure.
The AI power question is no longer simply whether the grid can serve the load. It is increasingly who pays for the grid required to serve it.
Verified sources
Federal Energy Regulatory Commission — Large Load Show Cause Orders
June 18, 2026. Six regional proceedings addressing consumer protection, reliability and transparency.
FERC Commissioner Rosner — Large Load Remarks
Cost Recovery Agreements, ratepayer protection, transparency and flexible-service concepts.
Virginia State Corporation Commission — Data Center Initiatives
GS-5 rate class, 14-year commitment, minimum charges and collateral provisions.
AEP Ohio — Schedule DCT
Data Center Tariff minimum-demand and collateral requirements.
FERC — PJM Large Load Show Cause Order, Docket EL26-67
Visibility into large-load-driven network upgrades, costs and allocation.
Robert Dizon
Expert insights from the Nistar team on energy infrastructure and hyperscale development.