AI and data centers: the risks of the natural gas rush
AI’s energy demand is pushing major operators toward natural gas, but costs, supply chains and volatility can turn the shortcut into a strategic risk.
The race for energy to power artificial intelligence is pushing large digital operators toward an industrial choice that is far from neutral: building or financing gas plants dedicated to data centers. The logic is simple: guarantee continuous power, fast deployment times and less dependence on the traditional electricity grid. But the simplicity of the solution hides a series of economic, operational and reputational risks.
For those leading infrastructure, investment or technology strategy, the question is not whether AI consumes a lot of energy: it already does. The real question is how sustainable it is to tie AI growth to a fossil source subject to volatility, industrial bottlenecks and increasingly strong regulatory pressure.
Why gas has become the fastest answer
Data centers dedicated to AI require continuity, high energy density and procurement timelines compatible with the speed of the market. Renewables remain central to the transition, but on their own they do not always guarantee immediate reliability at industrial scale. Gas therefore appears as a pragmatic bridge: modular plants, stable production and the ability to locate generation close to the load.
The problem is that the bridge risks becoming a permanent structure. When an infrastructure is built to meet a demand peak but takes years to complete, forecasting errors can turn into locked-in costs, underused assets or dependence on inflexible contracts.
The hidden bottlenecks in the supply chain
The first limit is not only the fuel, but the industrial chain needed to turn it into electrical capacity. Turbines for large plants are already under pressure, with long delivery times and rising prices. In practice, even those with the capital may not be able to bring the plant online when they want.
This creates an execution risk typical of markets in strong euphoria: everyone chases the same solution, at the same time, with the same technology. The result is competition for scarce assets that pushes toward premiums, delays and tactical choices that are hard to correct once contracts are signed.
Cost volatility and business impact
Even with dedicated supply agreements, gas does not remove market exposure. The price of electricity tends to reflect that of the fuel, and contracts can protect only partly against rises, seasonal bottlenecks or climate shocks. For an AI operator, this means greater uncertainty about the total cost of ownership of the infrastructure.
If the strategy relies on thin margins or an aggressive expansion roadmap, even moderate changes in energy costs can erode profitability. The risk is not only the energy bill: it also concerns the ability to plan investment, pricing and compute availability over reliable time horizons.
Systemic risk: competition with other essential uses
When gas demand grows rapidly to feed digital loads, conflicts emerge with sectors that depend on that same resource and have no immediate alternatives. This applies especially to industrial activities with high thermal intensity, which are harder to electrify and more exposed to supply interruptions.
There is also a question of social resilience: during winter peaks or extreme weather events, system priorities can become politically explosive. If available capacity is not enough, the question is no longer technical but strategic: who is served first, and who bears the cost of scarcity?
Three implications for those deciding today
The choice of gas to support AI should not be read as a definitive solution, but as a high-risk trade-off. Reducing that risk requires energy planning, contractual flexibility and architectures that do not depend on a single energy source.
- Design for modularity: avoid rigid dependence on a single source or a single plant.
- Assess the total cost: include timelines, turbines, fuel risk, regulation and reputation.
- Favor energy hybridization: combine gas, grids, storage and renewables wherever possible.
- Stress-test scenarios: test demand peaks, a harsh winter and rising prices.
- Align energy and AI strategy: compute growth must follow a credible energy roadmap.