Two Countries, Two Answers to the Data Center Question

Meta's 1-gigawatt Alberta data centre, announced this week at C$13 billion and scalable to 1.8 GW, is the first hyperscale investment to land on a structure no American jurisdiction has built: a federal enabling floor for AI compute ^[1]. The structure is not subtle. Canada's Treasury Board committed $2 billion to Telus for three data centres in British Columbia under a program whose name — "Enabling Large-Scale Sovereign AI Data Centres" — states its purpose ^[2]. Mark Carney's $2.3 billion "AI for All" strategy, unveiled in June, sets a federal compute capacity target of 850 MW by 2030, seeds a $700 million Compute Access Fund, and backs a sovereign AI Factory in Rimouski, Quebec, that opened last September and sold out immediately ^[3][4]. The federal pitch deck from Innovation, Science and Economic Development Canada claimed over 20 GW of capacity under development, enough to rank Canada second in the G7 — though officials later called that a "high-level snapshot" and conceded most proposed capacity will not proceed ^[1]. The United States has nothing on this axis. Federal engagement with the data center boom runs in two directions, neither of which chooses where facilities go. The first is restriction: AOC and Bernie Sanders introduced a bill to pause all new data center construction nationally ^[5]. The second is deflection: Trump's Ratepayer Protection Pledge tells tech companies to build their own power plants, pushing them off-grid rather than creating a siting framework ^[5]. The Federal Energy Regulatory Commission has ordered six grid operators to reform large-load tariffs, and PJM, the grid operator covering the mid-Atlantic, launched an emergency 15 GW generation plan ^[6][7]. But FERC's mandate is cost allocation and grid capacity. Where a data center physically sits remains a local zoning question, and the federal government has no position on it.

how each federal government answers 'where should AI data centers go?'

Canada: Enable: Federal billions committed: $2B to Telus (Treasury Board program) ^[2], $2.3B "AI for All" strategy with 850 MW national compute target by 2030 ^[3]. Sovereign AI Factory in Rimouski sold out ^[4]. Bell Canada sovereign infrastructure partnership ^[4]. Federal pitch deck courting hyperscalers with 20+ GW pipeline ^[1].

United States: Restrict or Deflect: AOC-Sanders national moratorium bill ^[5]. Trump Ratepayer Protection Pledge: "build your own power plant" ^[5]. FERC orders grid tariff reform (cost allocation, not siting) ^[6]. Virginia: $600M/year energy tax ^[8]. Texas Governor Abbott flags data center regulation for 2027 ^[9]. Multi-state moratorium push: NY, NM, MN, MO, KY ^[10]. County-level patchwork: Limerick Township tightens zoning, Fort Worth rejects restrictions — same week ^[9].

The structural asymmetry produces a pattern visible only when you line up a dozen separate stories. In the US, every level of government can say no to a data center but no level can say yes and make it stick. Counties veto: Limerick Township, Maine, unanimously tightened zoning with height limits, 1,000-foot setbacks, and 110% decommissioning funds, while Fort Worth's zoning commission split 7-4 the other way — but the City Council still votes August 11, so even the yes is provisional ^[9]. States tax: Virginia, the world's largest data center hub, imposed a $600 million annual energy consumption tax in July, with water, noise, and emissions controls layered on top ^[8]. The industry lobby's response captured the mood — the Data Center Coalition warned that "Virginia is no longer a reliable partner" ^[8]. The federal government either bans or steps aside. New York enacted a statewide moratorium; New Mexico, Minnesota, Missouri, and Kentucky have lawmakers pushing restrictions of their own ^[10]. Even Texas, which leads the nation in data center growth and whose grid operator reports data centers comprise 87% of 410 GW in large-load connection requests, is moving toward regulation, not enabling ^[9][11]. Canada's model does not eliminate local resistance. It relocates the veto from the county board to the premier's office. Manitoba Premier Wab Kinew blocked a proposed hyperscale data center southeast of Winnipeg, citing community impacts and energy demands ^[12][1]. An Angus Reid survey found 68% of Canadians oppose building large data centers near their homes, matching the NIMBY patterns driving US zoning fights ^[12]. In Saskatchewan, the RM of Sherwood unanimously approved Bell Canada's 300 MW facility amid a chaotic meeting where more than 100 protesters — including members of George Gordon First Nation alleging a failure in the duty to consult — banged on windows and were ejected from the gallery ^[13]. The Canadian debate is about how to enable, not whether to restrict. The Conservative shadow minister for AI, Ben Lobb, argued the government should "get out of the way" ^[2] — a critique of the enabling mechanism, not the enabling itself. Alberta is the one province whose government explicitly aligned with the federal yes. Two Alberta provincial officials, the technology minister and the premier, claimed deliberate governance as the cause of a corporate siting decision. Minister Nate Glubish said the Meta investment was no accident ^[1]. Premier Danielle Smith said the province had created the right conditions ^[1]. Whether that framing accounts for the whole decision is another matter. Alberta's deregulated electricity market delivered an 18% price drop in the first quarter of 2026, landing at $32 per megawatt-hour, driven by oversupply from natural gas and renewables ^[14]. That is a concrete economic pull that has nothing to do with governance architecture. And Meta is simultaneously investing $600 billion in US AI infrastructure over three years, with sites in Childress, Texas, and Warrenton, Missouri ^[15][1]. The Alberta deal is a parallel expansion, not a redirection from the United States. What the comparison does establish is that Canada has built something the US has not: a federal government that can say yes and back the word with billions, a national compute target, and a pipeline of sovereign infrastructure projects from Rimouski to Sturgeon County. The US federal government can say no — or it can tell companies to go generate their own power — but it cannot say yes in a way that binds the county zoning board two states away. Whether that governance gap becomes the variable that decides where the next hundred billion in AI infrastructure goes is a question one deal cannot yet answer. The Meta announcement is the first data point. The next test is whether a second hyperscaler picks Canada for reasons a federal pitch deck can credibly claim.