When AI and energy redraw the global infrastructure map

The Governor of Texas has ordered an audit of 1,800 data center projects, believing that the AI industry has dug its own grave in the face of public rejection. At the same time, Iran announces a gas field of over 7.5 trillion cubic feet, and investment funds are buying up American utility companies.

Key Points

• Iran claims the discovery of over 7.5 trillion cubic feet of gas in Fars province, of which approximately 5,700 billion cubic feet would be extractable, a "sweet" gas that would reduce processing costs.
• Nigeria positions itself as an alternative natural gas supplier in the face of the war in the Middle East, according to statements by its Minister of State for Petroleum Resources.
• A Brazilian construction company, facing the commoditization of logistics, is diversifying into data centers to restore its margins, a sign of a shift in investments.
• Startups are exploring underwater volcanoes off the coast of Oregon to produce continuous geothermal electricity, while AI companies are considering the ocean as a location for data centers.
• Hungary has brought its Paks nuclear power plant back to full capacity after a European drought had reduced its output, illustrating the fragility of energy infrastructures.

Context

For two decades, digitalization was presented as dematerialization. But the training and inference of AI models consume massive amounts of electricity and water. Data centers, once discreet, are becoming objects of contention. Opposition has emerged in the United States, particularly in Texas where 1,800 projects are underway. In parallel, energy geopolitics is becoming brutal again: the war between the United States, Israel, and Iran has put natural gas markets under strain, pushing Nigeria to seize the opportunity. This context recalls the oil shocks of the 1970s, but with a new variable: technology itself has become a player in energy demand.

Key Players

Local governors and regulators, like that of Texas, are demanding environmental audits and threatening to curb permits. Tech giants, facing local residents' anger and rising energy prices, are seeking refuges: oceans, regions rich in renewables, or even their own power plants. Investment funds, attracted by the profitability of utilities boosted by AI demand, are buying up electricity assets, with the risk of locking in fossil sources. Hydrocarbon-producing states, such as Iran, Nigeria, or Oman, are trying to monetize their reserves in a volatile market. Finally, laboratories and startups, like the one developing underwater geothermal generators, are proposing technological alternatives that are not yet proven on a large scale.

Data and Figures

Available institutional data show that the share of renewable energies in French electricity production locally reaches unprecedented levels: 44.97% for solar in Brittany, 47.01% for total renewables in Pays de la Loire, and 46.15% for bioenergies in Corsica. In Turkey, hydropower and other renewables account for 47.22% of electricity production. These figures, exceeding an alert threshold of 15%, indicate a green surge but also volatility: hydraulic production in Centre-Val shows a 47.33% decrease over one year, illustrating dependence on climatic hazards. In the United Kingdom, an indicator of electricity demand sensitivity to emissions stands at 42.86%, signaling a tension between growth and decarbonization. In light of this, Iran claims 7.5 trillion cubic feet of gas, of which 5,700 billion are extractable, an announcement that weighs on global prices. Texas, for its part, has 1,800 data center projects currently under audit.

Analysis of Issues

In the short term, local opposition risks delaying or canceling data center projects, particularly in arid areas or those already under electrical strain. Electricity prices could continue to rise, fueling inflation and political anger. Cyberattacks, such as an AI agent hacking development tickets, remind us that securing these infrastructures is critical. The construction sector, for its part, is rapidly reorienting: a Brazilian builder is partially abandoning logistics warehouses for data centers, a risky strategy if demand reverses.

In the medium term, natural gas geopolitics will toughen. Nigeria, with its abundant reserves, could become a major supplier for Europe and Asia, provided its infrastructures are secured. The recommissioning of Hungary's Paks nuclear power plant shows that nuclear power remains an adjustment variable, but its fragility in the face of droughts is a warning. Scientific work on the non-linear relationship between AI development and environmental pollution suggests that the net effect is not unequivocal: efficiency gains can be canceled out by the rebound effect. A study on lower-cost electricity trajectories in Saudi Arabia and another on alternative energy vectors for carbon neutrality in India indicate that developing countries will seek optimized solutions rather than renunciations.

This interpretation contains uncertainties: Iranian gas discovery figures must be confirmed by independent audits, and local renewable rates say nothing about the ability to meet peak AI demand. Furthermore, opposition to data centers varies across regions and political cultures; it could wane if tangible economic benefits are demonstrated.

Prospective Hypotheses

• Scenario 1: Regulation and refocusing (estimated probability 40%). Under electoral pressure, states like Texas impose moratoriums or strict conditions (energy efficiency, heat reuse). Data centers migrate to countries with decarbonized and stable electricity, such as Nordic countries or Canada. Indicators: number of permits refused, data center vacancy rates, electricity prices.
• Scenario 2: Oceanic and geothermal innovation (probability 35%). Pilot projects for underwater data centers and volcanic geothermal energy succeed technically and attract investments. Continuous and clean electricity production develops, reducing the footprint of data centers. Indicators: levelized cost of geothermal energy, commercial deployment, adoption by hyperscalers.
• Scenario 3: Energy shock and AI slowdown (probability 25%). A military escalation in the Middle East or a prolonged drought causes gas prices to surge and power outages. Governments ration electricity for data centers, slowing AI expansion. Indicators: natural gas prices, grid load alerts, generalized moratoriums.

Why This Matters

This is not an expert dispute: electricity bills, the availability of digital services, and job security are at stake. American workers, concerned about their jobs due to AI, observe how their employers communicate; broken trust can lead to failed adoption. Every citizen directly or indirectly consumes the energy of data centers. The open question: will the race for AI accelerate the energy transition or lock in a new fossil cycle?