Meta Alberta AI data center

The headline is a $13-billion Meta campus, but the load-bearing part of this deal is a separate 932 MW natural gas plant - the Greenlight Electricity Centre - built and owned not by Meta but by a partnership of Pembina Pipeline, Morgan Stanley Infrastructure Partners and Kineticor ^[1]. Meta buys the power under contract rather than operating the station itself, a structure that lets it secure gigawatt-scale, dedicated generation without carrying a power utility on its books. The tell is in the timing: the partnership locked in a final investment decision on the plant on July 2, roughly six days before Meta's public groundbreaking on July 8 ^[1]. The power path was settled before the data center was even announced.

That sequencing is the whole story of how the AI buildout has changed. As one industry analyst put it, "the hyperscaler race is moving from site acquisition to power-path control" ^[2]. Meta coordinated with AESO and Alberta utilities years in advance, and reporting notes that power conversations for gigawatt-scale projects now start three to five years out, versus a 6-to-12-month window not long ago ^[2]. Alberta's pitch - internally branded 'bring your own power' - is explicitly designed to attract this pattern, and the province says 60 data centre proponents are already in discussions. Meta is the proof point, and the gas plant, not the servers, is the scarce asset everyone is racing to lock up.

Meta says it will fund new generation and grid infrastructure to match 100% of the site's annual electricity with clean energy and to improve reliability across the Alberta grid ^[3]. The skeptics' objection is about what gets built now versus later. The Pembina Institute argues Alberta's framework favours natural gas over cheaper low-carbon alternatives and could push power prices up for ordinary consumers; its electricity director framed the arrangement bluntly: Alberta and Meta "have alternatively agreed to lock in gas today and build renewables later" ^[4]. Developers in Minnesota and Texas are pairing new AI load with wind and solar hybrids, while Alberta is anchoring a decade-scale facility to a gas plant that targets an in-service date in the second half of 2030 ^[4].

That is where the economics get contested. The province projects roughly $250M in annual benefits from royalties, taxes and fees, and Meta is putting about CAD $60M into local roads and water ^[5]. But the community argument - loudest on Reddit's Alberta and Canada forums - is that those figures understate the externalized costs: grid strain, emissions, road wear, water, and eventual reclamation, set against only about 300 permanent jobs after roughly 3,000 construction jobs disappear ^[5]. The dominant grassroots reaction was that the province is giving away long-term value for a resource-heavy project, with several threads demanding Meta cover the full external cost rather than socializing it onto ratepayers.

Meta's most repeated reassurance is environmental: a closed-loop, liquid-cooled system with dry cooling that it says eliminates operational water use, limiting water to domestic use, fire protection and equipment maintenance ^[2]. The company frames annual water use as less than a typical Alberta golf course or a 50-acre canola farm ^[6]. On paper this directly addresses the fear that data centers drain scarce water in a drought-prone region.

The claim has not landed cleanly. Mainstream Canadian news coverage on YouTube and television repeatedly returned to the Pembina Institute as a recurring skeptic, questioning both the 100%-renewable-match promise and the gas strategy's effect on bills. On Reddit, the water reassurance ran into open distrust, with worldnews commenters flagging water use in a drought-prone region as a reason for doubt rather than comfort. Part of the credibility gap is comparative memory: communities elsewhere have pushed back hard, with Hamilton imposing a data-centre moratorium and Manitoba rejecting a large proposal. In that context, a corporate water figure - however favourable - reads to hostile audiences as marketing, and the burden of proof has shifted onto Meta to demonstrate the closed loop performs as advertised over a multi-decade operating life.

Underneath the boosterism is a wager that AI demand justifies a purpose-built, decade-scale gas plant. TD Economics notes Alberta's available grid capacity for new data centers is limited - the first phase of the AESO large-load integration program offers just 1.2 GW - and that gas turbine shortages plus higher capital costs could slow new generation ^[2]. That scarcity is exactly why the dedicated-plant model exists, but it also concentrates risk: a 932 MW asset with an in-service target in the second half of 2030 is being committed against demand nobody can fully underwrite.

That is the fear animating the harshest community reaction. Reddit's Alberta forum surfaced explicit stranded-asset framing - what happens to a dedicated gas plant and its reclamation liability if the large-language-model buildout cools before the plant even switches on. The skepticism has moved beyond commentary into organizing, with a federal House of Commons petition (e-7550) calling to cease AI data-centre incentives, and Greenpeace Canada demanding a moratorium until legislated environmental and human-rights protections exist ^[5]. The bull case is that Meta's contract de-risks the plant for its builders; the bear case, voiced loudest by the people who live next to it, is that Alberta has tied a long-lived fossil asset and a consumer-facing power market to a single hyperscaler's capex cycle.