The Race to Control AI Begins With the Data Centre

The Conversation That Started This Investigation

I was not searching for a reason to oppose artificial intelligence. I was listening to Erin Brockovich speak with Theo Von about data centres, the communities living around them and the concerns people were reporting about water, electricity, noise and decisions being made before residents understood what was coming.

Watch or listen to Erin Brockovich with Theo Von

I thought the appearance did what it was meant to do. It brought a growing citizen issue into a conversation that people outside technology and environmental policy could understand. It also left me with a larger question—not because the discussion failed, but because no single podcast appearance could reasonably explain the full system behind the conflict.

Why are these facilities being built so quickly? Why are technology companies, infrastructure funds and governments willing to commit such large amounts of money, electricity, land, water and political support to them? What is being created that makes control of computing capacity so important?

Brockovich has since built a public reporting map that distinguishes operational, proposed, under-construction and community-reported locations. The project itself acknowledges an important limitation: a community report is evidence that a concern has been submitted, not automatic proof that a particular facility caused the reported harm. That is the standard this series will follow. Concerns deserve investigation. Allegations still require verification.

Review the Brockovich AI Data Center Reporting initiative and watch her June 2026 PBS interview.

This Is Not a Left-Wing or Right-Wing Question

The debate is often forced into political categories before the facts are understood. That is a mistake. People can support innovation, private enterprise, national competitiveness and a strong technology sector while still demanding transparent contracts, environmental review and protection from unfair utility costs.

People can also support environmental protection and corporate accountability while recognizing that data centres can strengthen scientific research, domestic computing capacity, employment, medical technology, accessibility, cybersecurity and national defence.

Citizens across political lines have reasons to care about property rights, public subsidies, electricity reliability, water security, government surveillance, foreign dependence, local control and democratic oversight. The questions may begin in different political traditions. They still lead to the same building.

The Cloud Is a Building

Artificial intelligence is commonly presented as an invisible service. A person types a request, waits a few seconds and receives an answer, image, video, analysis or song. The word cloud makes the process sound weightless.

It is not weightless. The service depends on physical machines operating inside physical facilities connected to physical energy and communications systems.

Inside and around that facility are servers, graphics processing units and other accelerators, networking equipment, batteries, uninterruptible power systems, cooling equipment, fibre connections, transformers, security systems and backup generation. Some cooling systems depend heavily on water. Others use air, closed loops, reclaimed water or combinations that change according to local conditions.

The International Energy Agency estimates that data centres consumed about 415 terawatt-hours of electricity in 2024, around 1.5% of global electricity consumption. Its base case projects approximately 945 terawatt-hours by 2030—just under 3% of global demand. That global percentage matters, but it can also hide the local problem: data centres tend to cluster, while new transmission, generation and grid equipment can take much longer to build than the facilities requesting power.

Read the International Energy Agency analysis of AI and data-centre electricity demand.

Why Governments and Companies Are Building So Quickly

Rising public use is part of the answer. More people are using generative AI. Businesses are integrating it into software, customer service, research, design, logistics and decision-making. New agent systems may remain active longer and complete more steps than a single chatbot response. Training increasingly capable models requires large clusters of advanced chips, while serving millions of users requires continuing inference capacity after training ends.

But consumer demand alone does not explain the language used by governments and companies. They increasingly describe computing capacity as strategic infrastructure tied to economic power, technological independence and national security.

In January 2025, OpenAI and SoftBank announced Stargate, a company intended to invest up to US$500 billion over four years in American AI infrastructure, beginning with an intended US$100 billion deployment. The announcement connected the project to American AI leadership, national security and the development of artificial general intelligence. It also identified power, land, construction and equipment as parts of the required buildout.

The United States is not alone. Canada has established a Sovereign AI Compute Strategy intended to expand domestic public and commercial capacity. The European Union is building AI Factories and planning larger AI Gigafactories. The United Kingdom has created AI Growth Zones intended to improve access to power and planning support for AI-enabled data centres. India’s national compute program is designed to broaden access for researchers, startups, public institutions and industry.

These programs have different ownership structures, rules and political goals. They should not be treated as one coordinated project. What they share is the recognition that access to advanced computing capacity may influence which countries can develop domestic AI, protect sensitive data, support local research and avoid dependence on foreign providers.

Read the original Stargate announcement.

Reaching Advanced AI Is Only Part of the Race

Public discussion often treats the race toward artificial general intelligence as though it ends when a laboratory produces a breakthrough system. That is only one stage.

Designing a breakthrough model is like designing a powerful new engine. Making that engine useful to millions of people requires fuel, factories, roads, maintenance, operators, replacement parts and a supply chain capable of keeping it running.

A capable model still requires continuing access to advanced chips, data-centre space, electricity, cooling, fibre, engineers, replacement hardware, security and capital. It may need new infrastructure to train future versions. It may need more capacity as usage expands. It may need separate systems for governments, regulated industries, scientific institutions or military applications.

Sustained access can also be interrupted. Export controls can restrict advanced chips. Grid connections can be delayed. Transformers and other electrical equipment can be scarce. Water or land approvals can be challenged. Financing can fail. Cyberattacks, contract disputes, government action or international conflict can change who has access.

This does not prove that AGI is imminent, inevitable or controlled by one organization. It does explain why the infrastructure race can continue even if predictions about AGI are wrong. The same facilities can still support business software, scientific modelling, government systems, media, cloud services and less ambitious forms of AI.

When Computing Capacity Becomes Power

Stanford’s 2026 AI Index reports that private industry produced more than 90% of the notable frontier models identified in 2025. It also describes a hardware system concentrated around a limited number of leading chip and infrastructure providers.

That concentration matters because independent access to computing capacity can determine who is able to train frontier systems, test safety claims, conduct advanced research and compete with established companies. Smaller businesses, universities and countries may possess talent and ideas without possessing the infrastructure needed to operate at the same scale.

Compute can therefore become a source of corporate and national influence. It may affect which countries depend on foreign providers, which researchers can test powerful models, which businesses receive affordable access, which governments gain intelligence or military capabilities and which institutions set the acceptable rules of use.

This remains a supported inference, not proof that a single company already controls advanced intelligence. Ownership is also more complicated than it first appears. The landowner, building operator, electricity customer, cloud provider, chip owner, model developer, financial investor and government partner may all be different organizations.

Review the Stanford 2026 AI Index.

Is the Manhattan Project Comparison Fair?

The comparison is partly useful. Both involve strategic competition, national-security concerns, scarce technical expertise, government intervention, major physical construction and a fear that falling behind could change the balance of power.

The scale of the current AI buildout may also become broader and more enduring than a single historical project because it is distributed across many companies, governments, financial institutions, energy systems and countries. Data-centre construction can continue long after any individual model breakthrough.

But the comparison becomes misleading when treated as literal equivalence. The Manhattan Project had a defined wartime weapons objective and a central government chain of command. AI development is multinational, commercial and fragmented. It includes civilian and creative uses alongside security and military applications. No single government currently controls the entire global effort.

The analogy helps explain the urgency and strategic mobilization. It should not be used to replace the facts.

The AI Race Is Being Built With Public Resources

A private company may own the servers, but a major data-centre project can still depend on public systems, public approvals and resources shared with the surrounding community.

These contributions are not automatically unjustified. A data centre can create years of construction work, skilled-trade opportunities, tax revenue, infrastructure investment, scientific capacity and domestic access to technology. In some communities, those benefits are substantial and residents support the projects.

The stronger question is not whether data centres create any value. They do. The question is how much value, for whom, for how long and at what public cost.

The next articles in this series will examine who pays for the electricity infrastructure, how water figures should be measured, what the public receives for subsidies and who ultimately owns and controls the computing power created.

What the Evidence Supports—and What It Does Not

Finding Your Sound Also Means Finding Your Position

I cover AI music because I believe in what music can do for an individual. A song can help someone understand an emotion, name an experience, strengthen faith, recover confidence or feel less alone.

Music can also do something for the public. It can carry an idea beyond a technical report. It can make people feel why an issue matters. It can inspire people to speak, question, create and act.

That is why building a creator brand cannot stop at choosing colours, writing a biography or developing a recognizable sound. A creator also has to develop a voice and learn how to form an informed position on the subjects affecting the people they hope to reach.

I use AI. I teach it. I believe it can make meaningful creative work possible for people who previously lacked the money, access, training or confidence to begin. That belief does not require silence when the infrastructure behind it raises questions about public resources, concentrated power and accountability.

Read Deeper: Four Books Behind This Investigation

This article contains Amazon.ca affiliate links. I may earn a commission from qualifying purchases at no additional cost to you.

These books do not all reach the same conclusions. Together, they offer different ways to understand the semiconductor supply chain, the physical foundations of AI, the governance problem and the institutions driving the race.

Questions Citizens Should Begin Asking

Later articles will turn these into deeper checklists. For now, these questions can help citizens move beyond a simple argument over whether a project is good or bad.

1. What will the facility be used to operate?
2. How much electricity has it requested, and when?
3. Who pays for the required generation, transmission and grid equipment?
4. Which water source and cooling system will it use?
5. What is the difference between permitted, peak and expected water use?
6. What tax incentives, grants, land arrangements or discounted services have been offered?
7. How many construction jobs and permanent jobs are expected?
8. Who owns the land, building, chips, model and computing capacity?
9. What information must be reported publicly after operations begin?
10. What happens if the project expands, closes, changes owners or fails to meet its commitments?

Primary Sources Used for This Introduction

• OpenAI and SoftBank: The Stargate Project
• International Energy Agency: Energy Demand From AI
• Stanford HAI: 2026 AI Index Report
• Brockovich AI Data Center Reporting
• Government of Canada: Sovereign AI Compute Strategy
• European Commission: AI Factories

The Intelligence May Be Digital. The Power Behind It Is Physical.

The race for artificial intelligence is usually described through models, benchmarks, applications and promises of superintelligence. But none of it operates without a physical place where chips receive electricity, systems are cooled, data is processed and access is controlled.

AI infrastructure can create real public benefits. Governments and companies have legitimate reasons to expand computing capacity. But citizens should not be asked to accept long-term commitments based only on broad promises about innovation and jobs.

Before we can understand the water, public money, surveillance capacity, ownership or ideology surrounding advanced AI, we have to understand the structure connecting them.

The investigation begins with the data centre because the power being built there will not remain inside the building.