Everybody is buying AI stocks. That is understandable. The money, attention and momentum have all been in models, chips and cloud platforms. But Meta’s C$13 billion Alberta data-centre announcement suggests the more interesting trade may be moving underneath the technology layer entirely. AI is no longer just a software race. It is becoming a power race, and that means the companies behind electricity, fuel, generation and grid infrastructure may be more important than the market wants to admit.
The conventional view is that Meta is building a Canadian AI campus because it needs more compute. True, but incomplete. The better interpretation is that Meta is building an energy-backed industrial facility. The planned Sturgeon County campus is expected to start at one gigawatt and scale toward 1.8 gigawatts, making it Meta’s first data centre in Canada and its largest outside the United States.[1] That is not a normal office park. That is an industrial-scale electricity load. Once AI demand reaches that size, the old assumption that technology companies can simply plug into the grid and wait for power becomes obsolete.
Meta’s Alberta project quietly created a list of infrastructure winners. The market sees Meta. We see the power stack underneath it.
The first company to watch is Capital Power. The company announced a long-term agreement to provide 250 megawatts of capacity and energy in support of Meta’s Sturgeon County data centre.[2] That may not sound as exciting as GPUs or model training, but it goes directly to the most important constraint in the AI economy: time. AI cannot wait five years for the perfect power plant, the perfect transmission upgrade or the perfect regulatory framework. It needs electricity before the full buildout arrives. Existing reliable generation suddenly has a customer that cannot afford downtime and cannot afford delay.
That changes how investors should think about a company like Capital Power. It is not simply selling electrons into a familiar utility market. It is becoming part of the bridge between today’s grid and tomorrow’s AI infrastructure. In an economy where hyperscalers need massive amounts of firm electricity, the assets that already exist may be worth more than the market expected.
The second company is Pembina Pipeline. Most investors still put Pembina in a traditional midstream box: pipelines, processing, liquids, gas transportation. That box is becoming too small. Pembina, alongside Morgan Stanley Infrastructure Partners and Kineticor Asset Management, reached a positive final investment decision on the Greenlight Electricity Centre, a 932-megawatt gas-fired combined-cycle power project in Sturgeon County designed to serve a major data-centre customer.[3] That project is not just adjacent to the AI story. It is part of the AI story.
This is where the market is still behind. A pipeline company involved in dedicated power generation for a hyperscale AI campus is no longer just moving fuel from point A to point B. It is helping convert natural gas into compute capacity. If every major hyperscaler starts asking the same question—where can we get reliable power at scale, quickly?—then midstream infrastructure starts to look less like yesterday’s energy business and more like tomorrow’s AI supply chain.
The third winner is not one stock. It is natural gas infrastructure as a whole. That may sound like a cop-out, but it is the biggest point. People keep asking what the next Nvidia is. The better question is who powers the next hundred Nvidias. Meta is not spending billions in Alberta because natural gas is fashionable. It is doing it because reliability beats ideology when a company’s AI strategy depends on power every hour of every day.
The next important development did not come from Alberta. It came from New York, where Governor Kathy Hochul signed an executive order creating the country’s first statewide moratorium on new hyperscale data centres. The order pauses state environmental permits for up to one year while New York develops a regulatory framework focused on ratepayers, the environment, the energy grid and local communities.[4]
At first glance, that sounds like New York saying no to AI. We do not think that is the real story. New York is not rejecting artificial intelligence. It is rejecting the idea that massive data-centre loads should arrive first and have the cost conversation later.
That distinction matters. Data centres use extraordinary amounts of electricity, but the problem is not just the electricity itself. It is the infrastructure required to deliver it: substations, transformers, transmission lines, backup generation, cooling systems, interconnection upgrades and planning capacity. When a hyperscale facility arrives with a load larger than many towns, local communities are right to ask whether the benefits justify the costs.
AP reported that New York’s order is designed to give the state time to create rules around environmental impact, energy demand, water usage and other concerns tied to large data centres.[5] Hochul’s own framing was even more direct: progress should not arrive with higher utility bills, strained water supplies or local pollution.[5] That is not anti-technology. It is cost allocation.
The uncomfortable truth for AI investors is that New York probably will not be the last jurisdiction to ask these questions. Data-centre moratoriums or restrictions have already been proposed in multiple places, even if most have not advanced as far as New York’s statewide pause.[5] The political question is obvious: if AI companies want unlimited power, who pays for the grid required to deliver it?
That is why Alberta and New York belong in the same article. Alberta shows what happens when a region says yes to AI by pairing data-centre development with dedicated energy infrastructure. New York shows what happens when a region says not yet because the cost, power and community framework is not ready. These are not opposing stories. They are two sides of the same thesis. AI demand is now large enough to force energy policy into the centre of technology policy.
The White House just made the same point nationally. Its Ratepayer Protection Pledge says companies should pay for new power delivery infrastructure upgrades required to serve their data centres, including network upgrade costs, so those expenses are not passed on to ordinary households.[6] It also says companies should negotiate separate rate structures and pay for power and related infrastructure brought online to serve their data centres, whether they use that electricity or not.[6]
That is a major shift in the AI trade. For months, investors asked how much money AI could make. The better question is now becoming who pays to power it.
Reuters reported that the White House has been rallying utilities and data-centre developers around this ratepayer-protection framework as AI demand puts new pressure on electricity systems.[7] Earlier reporting noted that companies including Google, Microsoft, Meta and Amazon signed onto the pledge to bear the cost of new electricity generation for data centres, in part to address concerns that Big Tech demand could push up bills for households and small businesses.[8]
That matters because once government starts asking who pays, investors start asking who profits. If hyperscalers are expected to fund their own generation, grid upgrades and reliability solutions, that is bullish for the companies that build and operate those systems. Utilities with available generation matter. Natural gas matters. Pipelines matter. Transformers, switchgear, substations and transmission matter. The AI story becomes less like a pure software story and more like a full industrial buildout.
This does not mean every utility or pipeline stock automatically wins. It means the market has to separate companies that merely sit near power demand from companies that can actually deliver power at scale, under contract, in regions that want the development. Capital Power’s 250-megawatt agreement with Meta matters because it solves a timing problem.[2] Pembina’s role in Greenlight matters because it solves a fuel-to-power problem.[3] Alberta’s posture matters because it solves a permitting and political problem. New York’s moratorium matters because it shows what happens when those problems remain unresolved.[4]
The controversial conclusion is simple: AI may accelerate natural gas infrastructure, not because Big Tech has abandoned clean-energy goals, but because reliability has become a competitive advantage. Meta and other hyperscalers can still buy renewables, fund offsets, sign clean-power agreements and pursue lower-carbon strategies. But the physical reality remains. Large AI campuses need firm electricity, and firm electricity requires infrastructure that exists in the real world, not just on a slide deck.
The AI trade is changing. It began with chips. It moved to cloud capacity. Now it is moving to power. The investors still only looking for the next software winner may be missing the quieter opportunity beneath the entire boom.
If the next decade of AI depends on who can secure reliable electricity first, are the biggest winners really going to look like tech companies—or are they going to look like boring infrastructure businesses?
[1] https://www.reuters.com/world/americas/meta-build-c13-billion-alberta-data-center-its-first-canada-2026-07-08/
[2] https://www.capitalpower.com/powering-whats-next-capital-power-signs-long-term-250-mw-agreement-with-meta/
[3] https://www.pembina.com/media-centre/news/details/3403fb77-2257-466b-a8e2-f043f4cd4650
[4] https://www.governor.ny.gov/news/first-statewide-moratorium-new-hyperscale-data-centers-launched-governor-kathy-hochul
[5] https://apnews.com/article/new-york-data-centers-moratorium-ai-c1e05b74208a6c570eec7c658ac8f187
[6] https://www.whitehouse.gov/releases/2026/03/ratepayer-protection-pledge/
[7] https://www.reuters.com/legal/litigation/white-house-rally-utilities-data-centers-over-ai-power-costs-2026-07-13/
[8] https://www.theguardian.com/us-news/2026/mar/04/us-tech-companies-energy-cost-pledge-white-house
Your personal details are strictly for our use, and you can unsubscribe at any time.