Alberta's $3.2 Billion Gas Plant for Data Centers: Who Pays if the Load Never Shows Up?

PowerMag reported on July 2 that Aecon's consortium won a C$1.7 billion (about $1.2 billion USD) engineering, procurement, and construction contract to build the first phase of the Greenlight Electricity Centre, a C$4.6 billion (about $3.2 billion USD) natural gas combined-cycle plant in Sturgeon County, Alberta.^[1] The project is sponsored by Pembina Pipeline, Morgan Stanley Infrastructure Partners, and Kineticor Asset Management, and is explicitly designed to power a major data-center development.^[5] Completion is targeted for the second half of 2030.^[5]

This is the architecture of socialized risk masquerading as private investment. The record shows a built-for-purpose gas plant chasing a load, not a load driving rational infrastructure. Alberta is betting that announced data-center projects, Kevin O'Leary's Wonder Valley (7.5 GW claimed), Data District's multi-center proposal (€8 billion investment [about $8.7 billion USD] described)^[5], will materialize on schedule and sign firm offtake agreements. But the public disclosures tell us almost nothing about what is actually contracted versus announced. Which data centers have signed binding power-purchase agreements, for how long, at what demand ratchets, and with what exit and assignment clauses? That information, if it exists, is buried in confidential commercial agreements between the sponsors and the load.

Alberta's regulatory framework does not yet mirror the tariff protections that US states have begun deploying for hyperscaler loads. Virginia's GS-5, Ohio's AEP large-load deal, and Oregon's Schedule 96 all impose minimum-demand ratchets (typically 60, 85% of contracted capacity), collateral requirements (Virginia: roughly $1.5 million CAD per MW), and multi-decade term locks to force the load class, not ratepayers, to absorb stranded-capacity risk if the load underperforms or exits early.^{[4 research library]} Alberta has no such standing large-load tariff. Instead, the jurisdiction will rely on project-level special agreements, which are opaque and typically exclude protective collateral and ratchets.

The data-center pipeline itself invites skepticism. Announced and speculative data-center MW in North America have grown five-fold in the past three years, according to grid-planning reports, yet historical build rates for announced projects have consistently fallen short of forecast.^{[2 research library]} The sponsors (Pembina, Morgan Stanley, Kineticor) have clear incentives to claim the load will materialize; they are betting corporate capital and seeking returns on construction and fuel-supply commitments. But they are not on the hook for stranded power if the load does not show. If the plant runs at 50% nameplate capacity because the Wonder Valley project stalls or a major customer rewrites its compute strategy, the transmission and distribution network upgrades that enabled that 932 MW base load remain. Alberta's ratepayers, not the data-center consortium, will carry that cost in perpetuity.

The protective move is immediate: Alberta's Utilities Commission should require that any special contract binding data-center load to this plant disclose (or, if redacted, summarize for the regulator's record) the key risk-allocation terms: (1) the contracted MW and minimum-demand ratchet; (2) the term of the offtake agreement versus the life of the asset; (3) collateral and exit-fee provisions covering unamortized investment; (4) cost-allocation rules for network upgrades; and (5) whether the load can assign its contract or walk early without triggering liability. In parallel, Alberta should codify a large-load tariff mirroring Virginia's model, isolating data-center cost responsibility from residential and small-business customers. Both moves are within regulatory scope and deployable before shovels break ground.