Some links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
Yes, AI growth can increase oil demand indirectly, but electricity—not oil—is the direct energy input for data centres. More AI computing means more electricity use; the added power can come from renewables, natural gas, nuclear energy and other sources, depending on the region. Oil may enter through backup generators, construction and transport, equipment supply chains, or wider economic growth. The International Energy Agency (IEA) has not published a standalone estimate of how much oil demand AI causes, so claims that AI has a specific oil-demand footprint go beyond the available figures.
How does AI use oil if AI runs on electricity?
AI uses electricity to run processors, memory, cooling systems and the other equipment inside data centres. That electricity is generated by a local power system or, in some cases, by onsite equipment. Oil is not the inherent fuel for an AI query: it can be involved further back in the chain, depending on how power is generated and how facilities and equipment are built and operated.
As an Amazon Associate I earn from qualifying purchases.
The chain is best understood as a sequence: more AI use can mean more computing; more computing can mean more data-centre electricity demand; meeting that demand can require new generation or grid capacity; and the resulting construction, fuel choices and economic effects can influence demand for several energy sources, including oil. Each link matters, but none makes every data centre an oil-powered facility.
The IEA captures the two-way relationship this way: “There is no AI without energy; at the same time, AI has the potential to transform the energy sector.” The statement appears in its 2025 report, Energy and AI. It describes both the energy needed to expand AI and the potential for AI applications to change how energy is produced and used.
#1 Best Overall
Where oil can enter the chain
- Backup power: A facility may use diesel or another liquid fuel in backup generators. This is a possible operational pathway, not evidence that oil is its routine source of electricity.
- Construction and logistics: Building data centres, power lines and generation capacity requires materials, equipment and transport. Those activities can use oil-derived fuels, although the IEA figures cited here do not isolate their oil use for AI projects.
- Equipment and materials: Data-centre hardware and infrastructure depend on supply chains that include petrochemical products. This is an indirect material connection, not a measure of oil burned to perform computation.
- Economic activity: If AI raises output or income, the resulting activity could increase energy use across the economy. The size and composition of any increase depend on what grows and how efficiently it is powered and supplied.
How much electricity do data centres use, and how fast is demand growing?
The IEA estimated that data centres worldwide consumed about 415 terawatt-hours (TWh) of electricity in 2024, or 1.5% of global electricity use. Their electricity consumption grew by about 12% per year from 2017 to 2024. These are data-centre figures, not a direct measurement of electricity used only for AI: facilities also run cloud computing, storage, networking and other workloads.
The IEA’s 2025 Energy and AI base case projected data-centre electricity consumption of 945 TWh in 2030 and 1,200 TWh in 2035. A separate IEA update in 2026 gives a 485–950 TWh range for data-centre electricity from 2025 to 2030, and reports that global data-centre electricity demand rose 17% in 2025. It also reports 50% growth in electricity consumption from AI-focused data centres that year. These figures come from different publications and forecast periods; they should not be spliced into a single continuous forecast or treated as oil-demand estimates.
| Measure | Reported figure | What it describes |
|---|---|---|
| Data-centre electricity use in 2024 | About 415 TWh | Global data-centre electricity consumption, IEA 2025 |
| Share of global electricity in 2024 | 1.5% | Data centres’ share, IEA 2025 |
| Regional shares in 2024 | United States 45%; China 25%; Europe 15% | Shares of global data-centre electricity consumption, IEA 2025 |
| 2025 data-centre demand growth | 17% | Global growth reported by the IEA in its 2026 update |
| 2025 AI-focused data-centre electricity growth | 50% | Growth reported by the IEA in its 2026 update |
| 2030 base-case projection | 945 TWh | Data-centre electricity consumption in the IEA’s 2025 base case |
| 2035 base-case projection | 1,200 TWh | Data-centre electricity consumption in the same IEA 2025 base case |
| 2025–2030 projection | 485–950 TWh | Data-centre electricity projection in the IEA’s 2026 update |
The United States, China and Europe together represented the large majority of data-centre electricity use in 2024, but their shares do not tell you which fuels supplied the power at a particular site. National and regional grids have different generation mixes, and those mixes can change over time.
Will data centres make oil consumption rise, or will other fuels supply the power?
For electricity demand directly associated with data centres, the IEA’s 2025 projections point to substantial additional generation from renewables, natural gas and nuclear power. The projected additions indicate that natural gas can be an important source of new supply, especially in constrained markets, but they do not show oil becoming the principal fuel for data centres.
Rank #2
| Generation source | IEA 2025 projection for additional data-centre-related generation to 2035 | How to read it |
|---|---|---|
| Renewables | More than 450 TWh | Projected additional generation serving data-centre demand |
| Natural gas | 175 TWh | Projected additional generation; can be a major marginal source in some markets |
| Nuclear | About the same amount as natural gas | Projected additional generation; the available data does not specify a more precise amount |
“Marginal” generation is the supply added or increased to meet the next increment of demand. It can matter more than the average grid mix when asking what new data-centre load causes a system to build or dispatch. But the marginal source is market-specific: it may be gas in one place, while another region adds renewable, nuclear or other generation. The IEA’s global projections are not a claim that every new data centre will draw power from the same fuel.
Oil can still be part of an electricity system, but the evidence here does not identify it as the main source meeting data-centre load. Nor does a rise in electricity demand automatically translate into a corresponding rise in oil demand: electricity and oil are different energy carriers, and generation choices vary by location.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Why might some data centres install onsite natural-gas generation?
When grid connections, transmission capacity or reliable power are difficult to secure, a data-centre developer may consider generating electricity onsite. The IEA’s 2026 update estimates that 15–27 gigawatts (GW) of onsite natural-gas capacity could serve US data centres by 2030. That is a possible capacity estimate for the United States, not a count of operating plants or a global forecast.
Reliability requirements can also affect the amount of equipment a project needs. The IEA says reliable onsite gas generation requires 30%–70% overbuilding relative to critical and variable load. In practical terms, installed capacity may have to exceed the load the facility expects to serve, so that available generation can cover fluctuations and reliability needs. This does not mean the plant runs at full capacity all the time, nor does capacity in gigawatts tell you annual fuel consumption.
Onsite gas is not the same as oil-powered computing. It is one response to power-supply constraints, and it is specific to the markets and projects where developers choose it. The estimate concerns potential US capacity; it should not be generalized to data centres worldwide.
Could more efficient AI reduce energy demand?
It could reduce the energy required for some tasks, but lower energy per task does not guarantee lower total electricity use. The IEA reports that energy use per AI task has fallen by at least an order of magnitude annually in recent years. At the same time, AI adoption, the number of tasks and the complexity of workloads can grow quickly enough to increase total demand.
Not all AI work has the same energy profile. The IEA says simple text queries use less electricity than running a television for the same period. Video generation, reasoning and agentic tasks can use hundreds or thousands of times more energy per query than simple text queries. The comparison is about the energy for a task, not a universal rating for all AI services: workload, model, hardware and operating conditions matter.
This is a rebound effect: as a service becomes cheaper or more efficient to use, people and businesses may use more of it or apply it to additional tasks. Efficiency gains can be real while total consumption rises. To judge the overall effect, distinguish energy per task from task volume and from the mix of workloads being run.
Rank #4
Can AI applications also reduce energy or oil use?
Yes. AI can be used to improve the operation of energy systems and industrial processes, potentially reducing fuel use or waste. The IEA identifies applications including industrial optimisation, renewable-power forecasting, grid fault detection, transmission management, oil-and-gas maintenance, leak detection and methane reduction.
Those uses create a potential counterweight to the energy needed for computing. Better forecasting may help manage variable renewable generation; improved grid operations may help make better use of existing infrastructure; and detecting leaks can reduce wasted fuel and methane emissions. The existence of an application, however, does not establish the size of its net savings across the economy.
Recommended Free Tools
The balance depends on adoption, how much energy each application actually saves, whether efficiency leads to more activity, and policy choices. AI may reduce energy use in one process while its broader adoption increases electricity demand elsewhere. The available IEA figures do not settle the net oil effect.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How should you assess claims that AI will drive oil demand?
Check what the claim measures and how it connects AI to oil. A claim about rising data-centre electricity use is not itself a claim about oil consumption; it needs evidence about the generation mix or another oil-using pathway.
Best Value
- Direct use or indirect effect? Separate liquid fuel burned in backup generation from oil demand associated with construction, transport, materials or economic growth.
- Electricity or total energy? A TWh figure for electricity is not a barrel figure and does not identify the fuel used to generate that electricity.
- Which place? Ask whether the claim concerns a specific site, a constrained regional grid, the United States or the world. The answer can change with local generation and infrastructure.
- Which time horizon? A near-term buildout through 2030 is different from a post-2030 scenario. Identify the report year and whether a number is historical, a projection or a range.
- Which workload? A simple text query is not interchangeable with video generation or an agentic task. Aggregate demand depends on both energy per task and how many tasks are performed.
- What is the counterfactual? If a study attributes growth to AI, ask what it compares against: a no-AI trend, an alternative grid plan or a different level of economic activity.
- Are constraints included? Grid connections, transmission, turbines, chips and financing can affect how quickly new capacity is built and which options are available.
A useful claim should identify its boundary: the facilities or economy counted, the fuels included, the region, and the period. Without those details, an electricity projection may be mistakenly presented as a precise oil forecast.
What does the evidence establish—and what remains uncertain?
The IEA’s figures establish that data-centre electricity demand is growing, that AI-focused data-centre electricity use grew particularly quickly in 2025, and that meeting expected demand will require additional electricity supply. Its projections include significant renewable generation and additional natural-gas and nuclear generation. The 2026 update also identifies potential onsite gas capacity for US data centres.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →They do not establish a standalone quantity of oil demand caused by AI. Oil-related pathways exist, but the evidence summarized here does not add them up into a global or regional AI-oil total. It also does not show that all data centres use gas, or that falling energy use per AI task will necessarily outweigh growth in the number and intensity of tasks.
The careful conclusion is therefore directional, not numerical: AI expansion can contribute to oil demand indirectly, but its immediate energy requirement is electricity, and the size of any oil effect depends on power systems, supply chains, backup arrangements and broader economic responses. A credible oil-demand number would need to measure those pathways explicitly rather than convert data-centre electricity use into barrels by assumption.
Frequently Asked Questions
Can you convert data-centre electricity use into barrels of oil?
Not from the electricity total alone. A conversion would need assumptions about where the electricity is generated, which fuels supply it, and how much oil is used in the other indirect pathways; the IEA figures cited here do not provide that calculation.
Does data-centre electricity use tell me how much pollution AI causes?
No. Electricity consumption is an energy quantity, while emissions depend on the generation mix and other parts of the supply chain. The electricity figures alone do not provide an emissions total for AI.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Repair Windows errors before they cause bigger problems3Fix the driver behind crashes, sound loss and screen glitchesDoes AI’s growth mean oil prices will rise?
The evidence summarized here does not establish a price effect. Oil prices reflect many influences, and a possible indirect demand contribution from AI cannot be translated into a specific price change without a separate market analysis.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
