· 8 minute field note
The obligation behind the megawatt
September's strongest AI-infrastructure signal was not the amount of power companies requested. It was the arrival of rules and contracts that ask who pays when that demand is late, smaller than forecast, or never connects.
- 01LegitimateIs the request specific enough to deserve scarce queue capacity?
- 02AllocatedWho pays for generation, transmission, and distribution upgrades?
- 03SecuredWhat collateral survives cancellation or a smaller-than-promised load?
- 04ApprovedWhich regulator, lender, or counterparty still has to say yes?
- 05DeliveredWhat still separates the obligation from energized, accepted power?
01 / The queue
A request consumes a scarce resource before it consumes electricity.
Massachusetts now directs utilities to eliminate speculative data-center projects from interconnection queues. For projects above 25 MW, the state points to fees, deposits, or other requirements that can establish whether the request is legitimate.
This matters because an interconnection study has an opportunity cost. A placeholder project can delay housing, factories, distributed energy, and another data center with committed customers and capital. Queue position is not delivered power, but it is already a claim on engineering time.
02 / The obligation
California turned a forecast into a minimum payment.
California's enacted AB 2383 requires generation tariffs for covered data centers. The payment mechanism must run for at least 10 years and include non-completion and early-termination fees, upfront collateral or prepayment, and a minimum payment based on the projected load in the interconnection application.
The important shift is not punitive. It is informational. A developer that accepts those obligations is sending a stronger demand signal than one that can reserve a large block of future capacity without carrying the downside.
03 / The subscriber
New supply is more credible when the beneficiary funds it.
Georgia Power says Google's proposed subscription will support approximately 96 MW of uprates at the existing Vogtle and Hatch nuclear plants. The utility expects the structure to create approximately $900 million of customer benefits over the life of the units and protect nonparticipating customers from incremental uprate costs.
The agreement does not create 96 operating megawatts today. The Hatch work and the tariff remain subject to Georgia Public Service Commission approval. What it does provide is a named subscriber, a proposed cost-allocation mechanism, existing generating assets, and a regulatory path.
04 / The financing
Committed capital clears one gate, not every gate.
VIRTUS completed a £2.45 billion financing package from a 13-bank group, including a £1.2 billion green capex facility. That is stronger evidence than a financing target: the funding framework exists.
The package supports expansion including the proposed 78 MW Saunderton campus. It does not make those 78 MW operational. Planning, power delivery, construction, commissioning, customer acceptance, and utilization remain separate evidence.
05 / The public grid
Selected funding is still waiting to become infrastructure.
The U.S. Department of Energy selected 31 SPARK grid projects across 26 states and estimates the portfolio could make more than 23 GW of additional capacity available. Its own timetable places expected awards between October 2026 and January 2027.
Selection is not an executed award. An award is not a completed line. A completed line is not a data-center interconnection. Public funding can improve the path to power without collapsing those stages into one headline number.
06 / What to watch next
The next evidence is contractual and physical.
- Queue deposits, study milestones, site control, and withdrawal rules.
- Tariff approval, minimum bills, collateral, and termination payments.
- Named subscribers, financing close, and notice to proceed.
- Executed awards, equipment orders, construction, and energization.
- Accepted load, utilization, reliability, and who absorbs any shortfall.
A balance sheet does not replace a substation. It tells the rest of the system whether the customer is prepared to pay while the substation becomes real.