How much energy does AI use UK businesses 2026 | Kilowatt Energy guide

How Much Energy Does a Single AI Query Use?

The most important context for understanding AI energy consumption is starts at the individual level before scaling up.

A single text query to a large language model like ChatGPT uses a surprisingly modest amount of electricity.

Sources: OpenAI, Epoch AI February 2025, Google Gemini environmental disclosure August 2025.

According to the International Energy Agency, if all conventional internet searches were performed with simple AI text queries, it would consume less than 4 terawatt-hours of electricity annually, equivalent to less than 1% of total data centre consumption today. IEA

The per-query figure is therefore not the issue. The issue is what happens when those queries are multiplied across millions of users, thousands of businesses, and every department within those businesses, running continuously every working day.

What Happens When You Scale AI Across a Business

One query is a rounding error. An entire organisation using AI tools across multiple departments every day is a different calculation entirely.

Consider a medium-sized UK business with 50 staff using AI tools routinely:

136 watt hours per day across the whole business is approximately 50 kilowatt hours per year from AI usage alone. At a current electricity rate of 27p per kWh, that is approximately £13.50 per year in direct AI energy cost for a 50-person business. Negligible in isolation.

But that calculation only captures the electricity consumed by the devices querying AI systems. The electricity consumed by the data centres running those AI models is paid for by the technology companies and is factored into the cost of wholesale electricity that every UK business pays.

The Bigger Picture: AI Data Centre Energy Demand in 2026

This is where the numbers become genuinely significant for UK businesses.

Global data centre electricity consumption is set to more than double to around 945 terawatt hours by 2030, slightly more than Japan’s total electricity consumption today. AI is the most important driver of this growth, alongside growing demand for other digital services. IEA

The global electricity demand of data centres grew by 17% in 2025, in line with IEA projections. Electricity consumption from AI-focused data centres grew even faster, surging 50% in 2025. IEA

Sources: IEA Energy and AI Report 2025, IEA Key Questions on Energy and AI 2026.

What This Means for UK Electricity Costs

The UK has approximately 500 to 600 operational data centres, with plans for a further 100 already established. Around half of existing centres are concentrated in and around London.

Data centres currently use less than 10 terawatt hours of energy in the UK, of the 319 TWh total consumption, marking around 3%.

However, data centre electricity demand is expected to rise to as much as 71 TWh between 2025 and 2050, putting additional strain on the grid. In London, 29 known data centres account for nearly one fifth of the energy consumed. Energycosts

For UK business electricity customers, there are two ways AI data centre growth affects what you pay:

Effect 1: Wholesale price pressure

Data centres buy electricity on the same wholesale market that energy suppliers use to price your business tariffs. When a large new data centre comes online, it adds demand to that market. More demand, all else being equal, puts upward pressure on wholesale prices.

In the United States, data centres account for nearly half of electricity demand growth between now and 2030. The UK trajectory, while smaller in scale, follows the same direction. IEA

Effect 2: Non-commodity charge increases

Cornwall Insight has forecast that by 2026, non-commodity charges will make up nearly 60% of a typical business electricity bill, driven by rising transmission costs, new bill components and support schemes for energy-intensive users.

If AI data centres add pressure to transmission and distribution investment, the effect may not show up only in the wholesale unit rate. It could also appear through higher non-commodity charges, especially on electricity contracts where third-party costs are passed through. Energycosts

This is the less visible but more immediate channel through which AI growth affects your bills. TNUoS charges already rose by approximately 60% from April 2026 as grid upgrade costs are recovered from all electricity consumers.

The grid upgrades needed to connect large AI data centres to the network add to the investment programme that future TNUoS charges will need to recover.

AI Energy and the Carbon Footprint Question

For businesses with sustainability commitments or ESG reporting obligations, the carbon footprint of AI tools is an increasingly relevant consideration.

Google published a detailed methodology in August 2025 showing the median Gemini text prompt consumes approximately 0.24 watt-hours and produces 0.03 grams of CO2 equivalent. On a per-query basis, that is a very small number. 

At global scale, research published in the journal Patterns in December 2025 estimated AI systems could produce between 32.6 million and 79.7 million tonnes of CO2 in 2025.

The carbon intensity of AI depends almost entirely on how the electricity powering data centres is generated.

Estimates based on 0.3 watt hours per query at stated carbon intensities.

Google’s total greenhouse gas emissions rose 51% since 2019, with AI a key driver, while its data centres consumed 30.8 million megawatt hours of electricity in 2024, more than double the amount in 2020. 

Microsoft’s emissions rose 23.4% since 2020 despite pledges to be carbon negative by 2030, primarily due to AI and cloud computing energy demands.

For businesses reporting under SECR or targeting net zero commitments, understanding the emissions profile of your digital operations and your energy supplier’s fuel mix is increasingly relevant.

What Landlords and HMO Operators Need to Know

Landlords and HMO operators face electricity cost pressures from two directions in 2026: the market-wide effects of AI data centre growth described above, and the more immediate challenge of EPC compliance requirements.

For standard buy-to-let landlords:

If you hold the energy account for your rental properties, the electricity rate you pay is affected by the same wholesale market pressures that AI data centre growth is contributing to. 

The July 2026 price cap rise of 13%, driven primarily by a 27.7% increase in the wholesale gas rate, demonstrates how quickly costs can move when market conditions shift.

Landlords holding energy accounts on residential properties who have not reviewed their tariff in the last 12 months should compare the market. 

The difference between a competitive fixed rate and an out-of-contract rate remains 40 to 60%, and rising market conditions make securing a fixed rate now more strategically sensible than waiting.

For HMO landlords specifically:

HMO operators who include energy costs within the rent face a more acute exposure than standard landlords. If your energy allowance was calculated in 2024 or early 2025, it is almost certainly below what the property now costs to run given the July 2026 price cap increases.

Estimates based on typical HMO consumption profiles and the July 2026 market movement. Actual figures depend on contract rates and consumption.

For a portfolio of 5 HMOs at the 5 bedroom benchmark, the cumulative monthly shortfall between a 2024 calculated energy allowance and 2026 actual costs is approximately £335 per month, or £4,020 per year. This is before any further market movement.

HMO landlords should review their energy tariffs and allowance calculations immediately. Our landlord energy switching guide covers the full process including VAT entitlements, portfolio procurement and void period management.

The EPC and AI connection for landlords:

The Government’s commitment to EPC Band C for all rental properties by 2030 is being accelerated partly by the same energy security concerns that are driving AI data centre policy. 

The UK AI Growth Zone strategy, which aims to have nationally significant AI campuses capable of serving 500MW or more of demand by 2030, requires grid infrastructure investment that creates upward pressure on non-commodity charges for all electricity customers, including landlords.

Landlords who invest in insulation, smart heating controls and energy-efficient appliances now are not just meeting future compliance requirements. They are reducing their exposure to electricity cost increases that will compound as grid investment charges rise through the 2020s.

AI as an Energy Management Tool for Businesses

The relationship between AI and energy is not exclusively about AI consuming more power. AI is also increasingly used to reduce energy consumption in commercial buildings and industrial settings.

For landlords and HMO operators, AI-powered smart heating controls and communal lighting systems with occupancy detection represent some of the most immediately accessible and cost-effective applications. 

Smart thermostats with AI scheduling can reduce heating costs in shared houses by 15 to 25% without affecting tenant comfort, and the payback period at current energy prices is typically 18 to 36 months.

The Grid Infrastructure Challenge

The UK Compute Roadmap says the Government wants a group of nationally significant AI Growth Zone sites, each capable of serving at least 500 megawatts of demand by 2030, with at least one AI Growth Zone scaling to more than 1 gigawatt. Energycosts

Connecting these sites to the national grid requires significant transmission infrastructure investment. That investment is recovered through TNUoS charges paid by all electricity consumers, including every UK business.

The interaction between renewable energy connection, AI data centre growth and grid infrastructure investment is the structural reason business electricity non-commodity charges are rising and will continue to rise regardless of what happens to wholesale prices.

What UK Businesses Should Do Right Now

AI energy demand is a macro trend you cannot control. What you can control is how your business is positioned on the energy market that this trend is helping to reshape.

Our commercial energy comparison service accesses live rates from 30-plus suppliers simultaneously at no cost to your business.

FAQ: How Much Energy Does AI Use?

Q: How much energy does a single AI query use?
A typical text-based AI query uses approximately 0.3 to 0.34 watt-hour of electricity, according to figures from Epoch AI and OpenAI. More complex tasks such as image generation use 1 to 3 watt-hours. 

On a per-query basis, these are small amounts. The significant energy demand comes from scale, with hundreds of millions of queries processed daily across global AI systems.

Q: How does AI energy use compare to a Google search?
A standard Google search uses approximately 0.03 watt-hour of electricity. A typical text-based AI query uses approximately 10 times more energy at 0.3 watt-hours. 

However, AI systems are becoming more efficient as hardware and software improve, and the gap has narrowed significantly from earlier estimates that circulated before 2024.

Q: Does AI affect my business electricity bill directly?
Not directly. AI tools do not appear as a separate line on your electricity bill. However, the growth of AI data centres adds demand to the wholesale electricity market and to the grid infrastructure that all electricity consumers pay for through non-commodity charges. 

Cornwall Insight has forecast that non-commodity charges will make up nearly 60% of a typical business electricity bill by 2026. The AI data centre build-out is one contributing factor to this trend. Energycosts

Q: What is the carbon footprint of AI tools?
It depends primarily on how the electricity powering the data centre is generated. Google has reported that a median Gemini text prompt produces approximately 0.03 grams of CO2 equivalent. 

At global scale, AI systems could produce between 32.6 million and 79.7 million tonnes of CO2 in 2025, according to research published in the journal Patterns in December 2025. 

For businesses with net zero or ESG targets, the fuel mix of your energy supplier affects the carbon intensity of your AI tool usage indirectly.

Q: How much will AI data centres affect UK electricity costs by 2030?
Global electricity consumption from AI-focused data centres grew 50% in 2025 and is projected to continue growing rapidly through 2030. In the UK, data centre electricity demand could rise from under 10 TWh today to as much as 71 TWh by 2050. 

The primary effect on UK business electricity costs will be through non-commodity charges rather than wholesale prices, as grid infrastructure investment to connect large AI campuses is recovered from all electricity consumers through transmission and distribution charges. IEA

Q: Should landlords and HMO operators be concerned about AI energy demand?
Yes, in the sense that AI data centre growth is one of several factors contributing to upward pressure on electricity costs in 2026 and beyond. 

For HMO landlords who include energy costs within the rent, the combination of the July 2026 price cap rise of 13% and anticipated continued non-commodity charge increases means energy allowances calculated in 2024 are likely already below actual costs. Reviewing your energy tariff and recalculating allowances is recommended immediately.

Q: Can AI help my business reduce its energy costs?
Yes. AI-powered energy management systems can reduce consumption in commercial buildings by 10 to 30% depending on the application. For HMO landlords, AI smart heating controls and occupancy sensing lighting can reduce communal area energy costs by 15 to 40%. 

The payback period for most AI energy management tools at current electricity prices is 18 to 48 months. Our forensic energy audit service also uses data analysis to identify billing errors and overcharges that have gone undetected.