The Sovereignty Argument: How Data Centers Became a Question of National Control

In September 2026, Canadian Prime Minister Mark Carney stood beside Saskatchewan Premier Scott Moe and Bell Canada CEO Mirko Bibic to announce the largest private capital investment in Saskatchewan's history: a $52 billion AI data center expansion, scaling from 300 megawatts to 1.2 gigawatts .

The framing was telling. Carney called it "the largest sovereign AI infrastructure network in Canada." Moe invoked the transcontinental railway, calling it "nation-building" for the digital age. "Do we build the infrastructure necessary to keep Canadian data in Canada?" Moe asked. "Or do we leave our data sovereignty in the hands of other countries?" 

This is the new political grammar of AI infrastructure. Data centers are no longer just warehouses for servers. They are instruments of national strategy, and the competition to host them is being reframed as a contest over who controls the physical layer of the digital economy.

The Sovereignty Turn

Fawaz Sheikh says The Canadian announcement is part of a broader pattern. Governments are increasingly treating AI compute capacity the same way they once treated oil refineries or semiconductor fabs: as strategic assets that should not be left entirely to market forces or foreign ownership.

Saskatchewan's data center framework, released in August 2026, explicitly prioritizes "Canadian ownership, support for Canadian data and AI sovereignty, the creation of Saskatchewan jobs, self-supplied power generation, and a centralized provincial intake process" . The Bell project was approved under these principles.

The logic is straightforward. If AI systems will mediate everything from healthcare to financial services to national security, the infrastructure that runs them should not be wholly dependent on foreign jurisdictions or foreign companies. Data sovereignty—the principle that data is subject to the laws and governance structures of the nation in which it is collected or stored—requires physical infrastructure within that nation's borders.

But sovereignty comes with costs. Bell must supply its own power for the expansion, a requirement that reflects the province's "Bring Your Own Power" principle . The first 300 megawatts will come from the provincial grid; the additional 900 megawatts must be self-generated. This is a significant burden, but it's also a political compromise: the province gets the jobs and the sovereignty narrative, while ratepayers are shielded from shouldering the infrastructure costs of a private AI project.

The Backlash

Not everyone is celebrating. Across Canada, municipalities have begun pushing back against data center projects. Oakville, Ontario, passed a temporary moratorium in August 2026 . The concerns are familiar: water consumption, electricity prices, noise, and the sense that communities are being asked to absorb the environmental and infrastructural costs of an industry whose benefits flow primarily to shareholders and distant users.

Federal AI Minister Evan Solomon responded by announcing voluntary principles for data center construction: projects must "create lasting local benefits," cannot shift electricity costs to Canadians, must minimize water use, and must "bring strategic value to Canada" . The principles are aspirational rather than binding, but they signal that the political tolerance for unfettered data center growth is narrowing.

The Southeast Asian Gambit

The sovereignty logic is not limited to wealthy Western nations. In Southeast Asia, Malaysia has positioned itself as a data center hub, attracting RM95.8 billion (approximately $23.6 billion) in approved data center and cloud investment in the first half of 2026 alone—nearly 44% of the country's total approved investment .

OpenAI's recent deal with Firmus to anchor two Malaysian AI data centers reflects this momentum. The facilities will deploy Nvidia's Vera Rubin systems and Firmus's HyperCube design, integrating liquid cooling and electrification . But the deal also reveals the tension between ambition and reality: Firmus has not disclosed how many megawatts OpenAI has contracted, where the facilities will be located, or when they will begin serving workloads .

Malaysia's Data Centre Task Force now clears projects only when developers can demonstrate secured electricity and water supplies and green compliance . In Johor, the data center pipeline has reached 8,542 MW, yet the colocation vacancy rate is just 0.7%—underscoring the gap between announced projects and operational capacity .

The Sovereignty Paradox

The sovereignty argument contains a paradox. The nations most eager to assert control over AI infrastructure are often the least equipped to build it independently.

Canada has approximately 337 megawatts of operational AI data center capacity, but projects accounting for another 20 gigawatts are "under planning or development" . The gap between ambition and execution is vast. The Bell expansion alone, if completed, would nearly quadruple the nation's operational capacity.

Similarly, Malaysia's investment figures are impressive in ringgit terms, but the country's ability to deliver power, water, and connectivity at the scale AI demands remains unproven. The Johor pipeline's 8,542 MW is a statement of intent, not a measure of achievement.

The sovereignty narrative also sits uneasily with the reality of supply chains. Nvidia chips are designed in California and fabricated in Taiwan. Liquid cooling systems, power transformers, and grid equipment come from global supply chains. A "sovereign" AI data center is sovereign only at the margin; its core components are thoroughly international.

What Sovereignty Actually Buys

The sovereignty argument may be most useful as a political tool rather than an economic one. It gives governments a language for justifying subsidies, expediting permits, and overriding local opposition. It transforms data centers from controversial industrial projects into symbols of national self-determination.

For the companies building them, sovereignty offers access. Bell gets provincial approval, grid connections, and political cover. OpenAI gets a foothold in a strategically important region. In both cases, the sovereignty frame smooths the path for infrastructure that might otherwise face insurmountable local resistance.

For citizens, the benefits are less clear. Sovereignty does not guarantee lower electricity prices, cleaner water, or better jobs. It does not ensure that AI systems built on "sovereign" infrastructure will serve public interests rather than corporate ones. The Bell project promises 500 permanent jobs and 3,000 ancillary positions—meaningful, but modest relative to a $52 billion investment .

The sovereignty argument, in other words, is necessary but not sufficient. It answers the question of where data centers should be built. It does not answer the question of what they should be for.

The Coming Test

The next phase of the sovereignty debate will play out in the details. Will "Bring Your Own Power" requirements become a template, forcing AI developers to internalize the infrastructure costs they currently externalize onto grids and ratepayers? Will sovereignty principles translate into enforceable standards for water use, community benefits, and data governance? Or will they remain rhetorical cover for projects that would have been built anyway?

The Bell expansion in Saskatchewan will be watched closely. If it delivers on its promises—jobs, sovereignty, self-supplied power—it will strengthen the case for treating AI infrastructure as a public-private strategic project. If it becomes a cautionary tale of overpromised benefits and underdelivered results, the sovereignty argument will lose its political force.

For now, the race is on. Governments want to host AI infrastructure; companies want to build it; communities want to know what they get in return. The sovereignty frame is the current answer to that question. Whether it survives contact with reality is the test ahead.

When the Grid Becomes the Bottleneck: What Texas's Data Center Interconnection Pause Really Means

In August 2026, Texas Governor Greg Abbott did something that sent shockwaves through the energy world: he ordered a pause on new data center grid interconnection applications until a full audit of the queue could be completed.

The reason was simple, and the numbers were staggering. The interconnection queue at ERCOT, Texas's grid operator, had reached 474 gigawatts of total requested capacity—roughly five times the state's all-time peak demand. About 90% of that came from companies building AI data centers.

The absurdity of the figure lies here: Texas's total electricity consumption in 2026 is projected to be around 761 terawatt-hours. If all 474 GW were built, data centers alone would multiply the state's grid demand several times over. And Texas is already one of the most open, permitting-aggressive energy markets in the United States.

Texas Isn't Alone in the Queue

Denmark's situation is equally severe. In March 2026, Danish transmission system operator Energinet suspended signing new interconnection agreements because the queue had reached roughly 60 gigawatts—while Denmark's entire peak demand is only about 7 GW. Of that, 14 GW came from data centers.

The Danish government's response was symbolic: it announced legislation to establish a statutory priority order for grid access. Healthcare, defense, and households applying to connect EV chargers or heat pumps would be prioritized, while large data centers would be placed at the back of the line.

This is not an isolated case. Globally, data centers are competing with green hydrogen producers, battery factories, and low-carbon steel makers for the same grid interconnection tickets. And the appetite of AI data centers—gigawatt-scale for a single project—is making that competition profoundly unequal.

Why AI Data Centers Are So "Greedy"

The power demand of a single hyperscale AI data center can easily rival a mid-sized city. Nvidia's latest GB200 NVL72 rack draws over 120 kilowatts per cabinet. A standard AI training cluster often requires hundreds of such racks running simultaneously.

More troublesome is the load profile. Traditional data centers have relatively steady loads, while AI training clusters can surge from near-zero to full power in seconds. NERC (North American Electric Reliability Corporation) warned in its July 2026 report that multiple incidents in 2025 involved more than 1,000 megawatts of computing load unexpectedly dropping off the grid during faults, creating new risks to system frequency stability.

This means AI data centers aren't just "large users"—they're "difficult users." The grid must reserve capacity for their peak demand, yet their behavior patterns make it hard for dispatchers to predict.

The Attempt to Bypass the Grid

Facing interconnection queues, tech giants' strategies are diverging.

Microsoft has chosen the "behind the meter" approach—building generation directly on-site or contracting with dedicated power plants that don't rely on the public grid. In Texas, the company has partnered with energy firms to develop gas-fired generation co-located with data centers, bypassing the ERCOT queue entirely.

Google is pursuing a different path: signing long-term power purchase agreements with nuclear and geothermal developers, betting that next-generation baseload technologies can scale fast enough to meet its 24/7 carbon-free energy goals.

Amazon, meanwhile, has gone on a buying spree—acquiring existing power plants and securing dedicated supply contracts, including a major nuclear power deal in Pennsylvania.

The Political Fallout

The Texas pause carries significant political weight. Data centers have become a rare bipartisan target: conservatives criticize them for driving up electricity prices for ordinary ratepayers, while progressives attack their water consumption and carbon footprint.

ERCOT's own projections show that data center demand could add $10 to $20 per month to residential electricity bills by 2028 if current growth continues. In a state where the grid is already strained by extreme weather, that's a politically explosive number.

Abbott's pause is being framed as a "pause for review," not a moratorium. But the signal is clear: even in the most business-friendly energy jurisdiction in America, the era of unlimited grid access for AI is ending.

What This Means for the AI Industry

Fawaz Sheikh says, The implications extend far beyond Texas.

First, location strategy is being rewritten. Proximity to power—not proximity to users or fiber—is becoming the primary siting criterion. This favors regions with surplus generation, stranded assets, or rapid renewable buildout.

Second, the cost of power is becoming the dominant variable in AI economics. Companies that locked in long-term contracts early are advantaged. Those that didn't will face rising prices and queue delays.

Third, the "bring your own power" model is becoming standard. Whether through on-site gas, nuclear PPAs, or dedicated microgrids, hyperscalers are increasingly expected to solve their own supply problems rather than relying on the public grid.

The Broader Question

The Texas pause raises a fundamental question: should data centers have the same right to grid access as households and hospitals?

Denmark's answer is no—at least not at the front of the line. Texas's answer is still being negotiated. But the direction of travel is clear. As AI's power demand grows, the social contract between data centers and the communities that host them is being renegotiated in real time.

The companies that thrive in this new era won't just be those with the best models or the most GPUs. They'll be those that figured out how to secure power in a world where the grid is no longer a given—and where the patience of regulators, utilities, and the public has limits.

AI and Data Centers: The Grid Is the New Frontier

The AI industry has a new bottleneck, and it isn't chips. It's electrons.

Microsoft CEO Satya Nadella put it bluntly in a recent podcast: "The biggest problem we face now isn't a surplus of compute resources, but whether power can be built fast enough where the data facilities are. If we can't, we may have a pile of chips sitting in warehouses that can't be plugged in."

That admission captures the defining tension of the AI era's second act. The constraint that matters most isn't fabrication capacity or algorithmic ingenuity—it's whether the physical infrastructure can deliver reliable, affordable electricity at the scale AI demands.

The Numbers Behind the Squeeze

Gartner's latest forecast projects global data center electricity consumption will reach 565 terawatt-hours in 2026, a 26% jump from 447 TWh in 2025. By 2030, that figure could exceed 1,200 TWh. For context, the world's data centers would consume more electricity annually than Japan uses in a year.

The International Energy Agency offers an even starker projection: global data center electricity consumption could more than double from 2024 levels to roughly 945 TWh by 2030, with the U.S. and China accounting for 80% of that demand. The United Nations University's 2026 report frames it differently: if data centers were a country, they would already rank as the world's eleventh-largest electricity consumer, ahead of Saudi Arabia and behind France.

Power capacity tells a parallel story. Global data center power demand is expected to reach 132 gigawatts in 2026, up from 104 GW in 2025, and is projected to more than double to 290 GW by 2030. A single AI-optimized server rack now draws 30 to 100 kilowatts—compared to 5 to 15 kW for traditional racks.

Why AI Workloads Are Different

Fawaz Sheikh says: The shift isn't just quantitative. It's architectural.

Traditional data centers were built for predictable, relatively steady workloads. AI facilities are fundamentally different beasts. Training large models requires dense clusters of GPUs running at maximum capacity for extended periods, while inference—the process of actually answering user queries—now accounts for 80 to 90% of total AI energy consumption.

The scale of individual deployments has become staggering. OpenAI's first "Stargate" data center in Abilene, Texas, will require up to 1.4 GW of power—enough to supply roughly a million homes—and will house more than 400,000 GPUs. Amazon's Indiana data center campus, costing $15 billion, will consume 2.2 GW.

Power purchase agreements have scaled accordingly. TotalEnergies signed two PPAs to deliver 1 GW of solar capacity to Google's Texas data centers over 15 years—the company's largest renewable PPA in the U.S. to date. ENGIE structured a three-party deal to supply 48 MW of solar to QTS data center operations in Irving, Texas, combining renewable generation with retail supply.

The Interconnection Bottleneck

The problem isn't just generating power. It's getting it to where the servers are.

Grid interconnection queues—the waiting lists for new energy projects to connect—have stretched to as long as seven years in some regions. Meanwhile, AI data centers can be built in two to three years. That mismatch routinely stalls projects across every major market.

Ditlev Engel, CEO of Energy at DNV, noted in a World Economic Forum piece that getting a new data center connected to the grid can take four to ten years in many regions—far longer than construction itself.

This gap has real consequences. In early 2026, at least 75 data center projects across the United States were halted or delayed, representing $130 billion in investment. The reason wasn't capital or chips. It was power.

Capital Flows Follow the Electrons

PwC's Global Data Center Outlook, released in September 2026, projects $31.6 trillion in cumulative data center investment through 2050. The report identifies power as "the binding constraint" in every region, the single most influential variable directing where capital lands.

The United States is expected to capture nearly half of all global investment—$15.1 trillion in cumulative spending. Asia Pacific follows at $8.2 trillion, led by China and India. Europe is projected at $5.6 trillion, though the report notes the region is "punching below its economic weight" due to power constraints and planning friction.

Annual data center capital expenditure is projected to climb from about $800 billion in 2026 to $1.8 trillion by 2050. Critically, information and communications technology equipment—servers, GPUs, networking gear—will rise from 70% of total spending today to 93% by 2050, as hardware refreshes every four to six years force continuous reinvestment.

The Geographic Shift

The migration of AI infrastructure toward power sources is reshaping the global compute map.

In the United States, West Virginia has emerged as an unlikely AI hub. Anthropic reportedly agreed to spend approximately $45 billion over six years for computing capacity at Nscale's Mason County campus, which will deploy NVIDIA's next-generation Vera Rubin technology and operate under a state-certified microgrid framework.

China's "East Data, West Computing" initiative has similarly pushed compute capacity toward regions with abundant energy. Ulanqab, a city in Inner Mongolia with fewer than 2 million residents, has signed 89 data center projects with total investment exceeding 500 billion yuan. Huawei, Apple, Alibaba, ByteDance, and DeepSeek have all established operations there. By mid-2026, the city's planned capacity reached approximately 12.5 GW—more than ten times its operating scale in 2025.

The logic is straightforward: AI workloads are being pushed toward where the power is, not where the users are.

The Water Dimension

Electricity is only half the resource equation.

Google consumed 10.9 billion gallons of water in 2025, a 34% increase from 2024 and more than double its 2021 level. U.S. data centers directly consumed roughly 17.4 billion gallons in 2023, more than triple the 5.6 billion gallons consumed in 2014.

A single standard data center can consume roughly 110 million gallons annually—comparable to the yearly water needs of a town of 50,000 residents. Large hyperscale facilities can draw up to 5 million gallons per day.

The Stanford AI Index estimates GPT-4o's annual inference water use alone may exceed the drinking water needs of 12 million people. By 2030, UNU-INWEH projects the water footprint associated with data center electricity could reach 9,300 billion liters annually—equivalent to the domestic water needs of 1.3 billion people in sub-Saharan Africa.

The trade-off between water and power is thorny: air-cooled chillers avoid on-site water use but shift the burden toward higher electricity consumption, which relocates the environmental cost to the indirect water embedded in power generation. Reducing on-site water often just inflates the off-site footprint.

The Emerging Response

The industry's answer is increasingly to bring power generation on-site, behind the meter.

Nscale's West Virginia campus operates under a state-certified microgrid framework, with legislation designed to prevent infrastructure costs from being shifted onto existing residential and business ratepayers. Fifty percent of tax revenue generated through the microgrid system is dedicated to reducing the state income tax.

In Texas, Nexus Data Centers secured a $16 billion project-finance package for a 2,900-acre facility in Hubbard that will have its own gas-fired power plant, with Anthropic as primary tenant.

The natural gas and nuclear resurgence is directly tied to AI's 24/7 power needs. Data centers cannot rely solely on intermittent renewables; they need baseload power that runs when the sun doesn't shine and the wind doesn't blow.

What Comes Next

The AI buildout has entered a phase where software ambition collides with physical limits. The industry's ability to secure reliable, affordable, and increasingly low-carbon electricity will determine not just where data centers get built, but which companies can scale and which regions capture the economic benefits.

PwC's report identifies five factors directing investment flows: power, connectivity, security, policy certainty and community consent, and GPU access. Power sits at the top of that list.

The coming years will test whether the digital economy can decouple its growth from its energy appetite—or whether the two will remain inseparable, with electricity as the ultimate arbiter of AI's reach.

AI and Data Centers: From "Compute Hunger" to a Symbiotic Relationship

The explosive growth of artificial intelligence is redrawing the map of global data centers. When OpenAI used more than ten thousand AI agents to crack a 90-year-old unsolved math problem in 88 hours, the compute it consumed was enough to make a mid-sized city's power grid tremble. This reveals an increasingly clear core contradiction: AI's capability ceiling is increasingly determined by whether data centers can secure enough electricity, water, and—more importantly—smarter ways of operating.

The Physical Cost of Runaway Compute

Stanford University's 2026 AI Index Report reveals a set of sobering numbers: the United States has 5,427 AI data centers, more than ten times that of any other country; AI data center power capacity has climbed to 29.6 gigawatts, roughly equivalent to New York State's entire peak demand.

Water consumption is equally striking. GPT-4o's annual inference water use is estimated to exceed the drinking water needs of 12 million people. In a white paper jointly released by China's CAICT and Grundfos, it is projected that by 2030, U.S. data centers alone will add between 697 and 1,451 million gallons of daily water supply demand just to meet cooling needs.

"The mismatch between water use efficiency and the pace of compute expansion further amplifies local water supply-demand conflicts," the report notes. In the AI era, the core competitive edge has shifted from simply comparing compute scale to a comprehensive contest of compute, energy, and water resource coordination.

When AI Starts Managing Itself

Facing this predicament, the industry's answer is surprisingly elegant: use AI to optimize AI's own infrastructure.

At NTT DATA's data center in Bonn, Germany, a hybrid AI-based cooling optimization system is already running. The system combines physical models with data-driven AI algorithms to create a "digital twin" of the cooling system, simulating and optimizing chiller operations in real time. Initial results show a 19.1% reduction in cooling system energy consumption, with projected annual savings of up to 25%.

Digital Realty deployed Phaidra's AI platform at its Northern Virginia data center. The company's 2025 impact report shows that while its data center portfolio grew 34%, water consumption increased by only 3%. This "huge divergence" is attributed to AI-driven optimization systems that monitor pumps, filters, and the full infrastructure stack in real time.

AWS has gone further. Its network operations now incorporate AI agents that automatically correlate telemetry from multiple monitoring systems, completing root-cause identification in seconds rather than the traditional minutes. For routine operational issues, the AI reviews tickets, compares historical patterns, and in many cases resolves problems without human intervention.

A New Paradigm for Grid Flexibility

The real breakthrough may come from the concept of "flexibility." Emerald AI's Conductor software represents the most radical approach: letting data centers actively sense grid needs and modestly reduce compute load during power stress—without affecting critical task execution.

In one test, Conductor took control of 256 Nvidia A100 GPUs—hardware consuming roughly as much power as about 170 U.S. homes. When the grid came under pressure, the system reduced chip power by 25% for three hours while maintaining acceptable compute performance.

This "speed-to-power" model is drawing attention from utilities. In Hillsboro, Oregon, Aligned Data Centers agreed to install a 31-megawatt battery system to reduce draw during grid congestion. This flexibility measure, combined with others, allowed local utility Portland General Electric to add 80 megawatts of supply capacity for surrounding data centers without building new power plants.

Google has been shifting processing loads from high-demand regions to less stressed facilities since 2023, and has signed agreements with five utilities adding up to 1 gigawatt of flexibility.

The Battle Over Cooling Technologies

The choice of cooling method directly determines a data center's resource footprint. According to the CAICT report, open-loop cooling systems rely on evaporation to remove heat, consuming large amounts of water; closed-loop dry cooling systems circulate coolant through pipes, reducing evaporative water consumption to extremely low levels.

The reason northern China has become a compute hub—Ulanqab in Inner Mongolia, Zhongwei in Ningxia, Qingyang in Gansu—is not only abundant green power but also climatic conditions that allow low-water cooling. But an easily overlooked contradiction is that these regions are precisely the most water-scarce in the country. A data center's "absolute water consumption" may not be large, but its "relative pressure on water resource carrying capacity" cannot be ignored.

The value of liquid cooling extends beyond mere cooling efficiency. Grundfos technical experts point out that liquid cooling systems can return water at temperatures of 40-50°C, making it possible to heat surrounding buildings. "The significance of liquid cooling isn't just high cooling efficiency—it's that it allows data centers to transform from pure energy consumers into resource nodes that can export thermal energy. "

Investment Frenzy and Rational Concerns

Capital continues to pour in. Bell Canada announced it would expand its AI data center in Saskatchewan from a planned 300 megawatts to 1.2 gigawatts, increasing investment to 52 billion Canadian dollars—the largest private capital investment in the province's history. Yet Wall Street has begun weighing the possibility of a slowdown in AI development. Anthropic CEO Dario Amodei's proposal to slow the pace of frontier model development directly triggered a selloff in AI infrastructure stocks—GE Vernova fell nearly 9%, Vertiv nearly 8%.

This tension reveals a deeper problem: the current data center construction boom is, to a large extent, a bet on sustained compute demand from a handful of frontier AI companies. Any change in the pace of model development transmits rapidly through the supply chain.

Meanwhile, a new safety concern is emerging. The evolution of AI model architectures—such as "recurrent depth" techniques—is migrating more reasoning processes into a "latent space" unobservable to humans. This means the "chain-of-thought" window developers use to monitor whether models are engaging in deception or reward hacking may gradually narrow.

The relationship between AI and data centers is evolving from a simple "demand and supply" dynamic into a complex symbiosis. Data centers use AI to manage their own energy and water consumption, while AI's progress in turn defines the shape of data centers. The ultimate test of this transformation may be this: between compute hunger and resource constraints, can the industry find a path that sustains capability growth without overdrafting environmental and social trust?