Meta broke ground on its first Canadian data center: a 1-gigawatt facility in Sturgeon County, Alberta, north of CAD $13 billion (roughly $9 billion USD). It's Meta's 33rd data center globally, and the first deliberate step of its AI buildout across the US border — arriving alongside Meta's Q1 2026 disclosure that it expects to spend $125-145 billion on infrastructure in 2026 alone, primarily for AI.
Why Alberta
Alberta offers power infrastructure and land at a scale that's increasingly hard to secure in saturated US markets like Northern Virginia. Meta says the campus will match all its electricity consumption with clean and renewable energy — as much a practical power-sourcing decision as a sustainability commitment, given how much scrutiny hyperscale AI facilities face over grid demand. The project is expected to support 3,000+ construction jobs at peak and 300+ permanent operational roles — modest permanent headcount, which is exactly why regions compete for these projects anyway, for the tax base and grid investment they bring. Alberta's oil-and-gas-era energy infrastructure expertise gives it a workforce arguably better prepared to support a gigawatt-scale industrial power consumer than most regions without that legacy.
The scale of the bet
Read this single facility against Meta's full 2026 capex guidance: $125-145 billion, primarily AI infrastructure, with $600 billion-plus already committed inside the US. That places Meta alongside Microsoft, Google, and Amazon in a race where the binding constraint isn't chip supply or capital anymore — it's power availability and construction timelines. A 1GW facility is roughly a mid-sized power plant dedicated entirely to running servers, and it's one of dozens of similar projects across the industry right now.
The power crisis is the real headline
Meta choosing to build outside the US, specifically where power procurement is more available, is itself a data point in 2026's data center power crisis. As US grid capacity in traditional hubs gets harder to secure at the speed AI companies want to build, expect more hyperscalers to follow Meta's lead toward Canada and other jurisdictions with available grid capacity and renewable procurement options. It raises a longer question for US policymakers too: if the fastest-growing category of industrial power demand increasingly locates outside US borders, the construction jobs, tax revenue, and operational employment go with it.
What it means if you just consume cloud AI
More capacity distributed across more regions generally supports more stable pricing and better latency over time — but providers with the fastest, most geographically diversified buildouts (Meta, Microsoft, Google, Amazon) will likely hold a durable capacity advantage over smaller or regionally-constrained competitors. For multi-year cloud and AI commitments, knowing which providers are actually securing new power and land at this scale — versus relying on existing, strained capacity — is a legitimate vendor-selection factor now.
Practical takeaways
- Track hyperscaler capex guidance as a leading indicator of where capacity, and eventually pricing power, will concentrate.
- Consider geographic diversification in your own architecture as new regions like Alberta come online.
- Watch power availability, not chip announcements — that's the real 2026 constraint on AI infrastructure growth.
- Factor renewable-matched facilities into vendor evaluation if your org has its own ESG reporting tied to cloud selection.
- Check data residency implications if Meta's Canadian infrastructure ends up hosting workloads relevant to your compliance posture.
AI infrastructure spending has moved past "buy more GPUs" into "secure power and land at hyperscale, wherever it's available." How aggressively your providers are expanding physical infrastructure is no longer a background detail — it's a direct input into your future capacity, pricing, and reliability.