The Next Global Crisis Could Start in a Data Center

Civilization is quietly building a new foundational layer, and few governance systems have caught up to it.

Artificial intelligence is often discussed as a software innovation, a productivity tool, or the latest phase of technological disruption. Those descriptions are increasingly insufficient. AI is no longer merely an application layer built on top of digital systems. It is rapidly evolving into critical infrastructure, a foundational capability upon which governments, economies, industries, and societies are beginning to depend.

Historically, societies have treated certain systems as strategically indispensable because their disruption would produce cascading consequences. Electricity, telecommunications, transportation, water systems, and financial institutions all became critical infrastructure because modern civilization could not function without them. AI is entering this category too, not because it replaces these systems, but because it is becoming embedded within them.

The Common Misunderstanding

The most common misunderstanding about AI is treating it as pure software, something that exists abstractly in the cloud with no physical footprint. In reality, advanced AI systems require enormous physical resources: compute infrastructure, semiconductor fabrication, hyperscale data centers, energy generation, cooling systems, and rare earth supply chains form the material foundation beneath every frontier model. This is not a minor technical detail. It is the reason AI infrastructure is now a geopolitical asset in the same category as oil reserves or naval dominance, and it is why the actual numbers behind this buildout are worth taking seriously.

The Physical Scale of the Buildout

The dollar figures involved are no longer abstract. The four largest US hyperscalers, Amazon, Microsoft, Meta, and Alphabet, are projected to spend between roughly $630 billion and $725 billion combined on capital expenditures in 2026, up sharply from an already record $388 to $410 billion in 2025 (Futurum Group; Tom’s Hardware, 2026). The large majority of that spending goes toward AI data centers, specialized chips, and the power infrastructure needed to run them.

Energy is where this physical dependency becomes most concrete. The International Energy Agency found that global data center electricity demand grew 17 percent in 2025, with electricity consumption specifically from AI-focused data centers surging 50 percent that same year. Data centers overall consumed roughly 415 terawatt hours in 2024, about 1.5 percent of global electricity, and the IEA projects that figure will roughly double to around 950 terawatt hours by 2030, close to 3 percent of global demand, with AI-specific consumption growing fastest of all (International Energy Agency, 2026). Training and operating frontier models is, in a very literal sense, an energy policy question now, not just a computing one.

Why Semiconductor Concentration Is a Strategic Vulnerability

Semiconductor manufacturing illustrates the concentration risk clearly. Advanced AI systems depend on specialized chips built for high-performance parallel processing, and a strikingly small number of companies dominate that supply chain. Taiwan Semiconductor Manufacturing Company alone controls roughly 72 percent of the global pure-foundry market, with its nearest competitor, Samsung, holding only about 7 percent, and Taiwan is responsible for producing something like 90 percent of the world’s most advanced chips (TSMC quarterly filings, 2026). A disruption to that supply chain, whether from natural disaster, geopolitical conflict, or export policy, would not simply raise prices. It would constrain the physical capacity for AI development worldwide almost overnight.

This is precisely why governments increasingly treat semiconductor policy as a national security matter rather than a purely economic one. Export controls, domestic manufacturing incentives, and industrial policy around AI hardware reflect a recognition that compute capacity has become a strategic asset in its own right.

The Rise of Sovereign AI

Governments are responding to this concentration risk with a wave of what is now commonly called sovereign AI investment, building domestic compute, chip, and model capacity specifically to reduce dependence on foreign providers. France has committed roughly €109 billion toward domestic AI infrastructure, including a Fluidstack-backed supercomputer project intended to host 500,000 next-generation chips, and the broader European Union has launched a €20 billion AI gigafactory initiative to fund large-scale compute facilities across the bloc (raiseSUMMIT; McKinsey, 2026). India’s IndiaAI Mission had deployed roughly 34,000 subsidized GPUs to startups, researchers, and government agencies by mid-2026, with a stated target of 100,000 public GPUs by year’s end. Gulf states, led by the UAE’s G42 and Saudi Arabia’s HUMAIN, have announced combined AI infrastructure commitments exceeding $100 billion. The global sovereign AI market itself is estimated at roughly $40 billion in 2025, projected to grow toward $148 billion by 2032 (MarketsandMarkets, 2026). None of this is hypothetical policy discussion. It is capital already being deployed at a national scale.

Concentration, Cybersecurity, and Systemic Risk

The concentration of AI capability within a small number of firms raises a related governance concern distinct from semiconductor supply. A handful of corporations currently control substantial portions of the world’s cloud infrastructure, advanced AI research, and foundational models, creating a level of infrastructural centralization in private hands that influences economic activity, communications, and information access well beyond what any single regulator currently oversees. Disruptions affecting a major cloud provider or foundational model platform could cascade across multiple industries simultaneously, which makes that concentration simultaneously an efficiency advantage and a resilience liability.

Cybersecurity reinforces AI’s status as critical infrastructure in a similar way. As AI systems integrate into essential services, they become attractive targets for espionage, sabotage, and disruption, and an attack on AI infrastructure specifically may not simply compromise data. It may degrade the operational decision-making that increasingly depends on these systems. AI systems can also amplify vulnerability internally, through automation errors, biased outputs, model manipulation, or simply opaque decision-making that is hard to audit under pressure.

What This Means for Sovereignty and Governance

Historically, nations sought energy independence, military independence, and industrial independence to reduce strategic vulnerability. Sovereign AI investment reflects the same logic applied to a new domain: a recognition that dependence on foreign-controlled AI infrastructure could create long-term strategic exposure comparable to energy dependence in earlier eras.

At a societal level, this raises questions that go beyond infrastructure economics. Electrical grids transformed industrial society. Telecommunications reshaped social interaction. The internet reorganized information access. AI is positioned to alter how individuals work, communicate, and make decisions by increasingly mediating all three. As dependence deepens, the line between tool and infrastructure erodes. Infrastructure, in the end, is defined less by whether it is physical and more by whether society can function without it. Once that threshold is crossed, a system becomes infrastructural regardless of whether it is made of steel, electricity, or code.

Conclusion

Governance has not caught up to this reality. AI policy discussion still often focuses narrowly on content moderation and consumer applications, while underestimating the broader strategic implications of a physical, energy-intensive, geographically concentrated infrastructure layer that a small number of companies and countries now substantially control. The question is no longer simply whether AI can improve productivity. It is whether governments, corporations, and societies are building a foundational layer of modern civilization without fully accounting for its dependencies, its concentration risk, and its vulnerabilities, and whether they will recognize that in time to act on it rather than after the fact.

W3 Evidence Index™

W3 Evidence Index™ Score: 7.1/10
Confidence Level: High Confidence

Full category breakdown, assessment, and limits available below for paid subscribers.