The AI Efficiency Survey

SambaNova commissioned a survey in December 2025 that polled over 2,500 adults across the US and the UK, exploring how people feel about AI’s impact on energy use, electricity costs, and national control of AI infrastructure.

Key Findings Include:

• There is great concern about AI’s energy appetite; its potential to drive up bills is now firmly in consciousness, with 75% of respondents fearing AI data centers could lead to higher household energy bills in their area.

• People are becoming pro‑efficiency, with 83% believing AI companies should prioritize energy efficiency, even if it slows down technological development.

• There is a widespread perception that AI will strain national energy grids, with 71% agreeing that AI data centers will strain their country’s power grid.
  
  

FULL KEY FINDINGS

AI is mainstream

• 90% of respondents are at least somewhat familiar with AI services such as chatbots, image generators, or AI assistants.
• 41% say they are “very familiar – I use them regularly,” and a further 49% “use them occasionally.”
• Only 0.3% say they are “not familiar at all” with AI services.

Interpretation: AI is no longer niche. Any public debate about AI’s energy use or regulation is about a technology people already know and use.

Consumers are already aware of AI’s energy appetite

• 75% were aware before the survey that AI data centers use large amounts of electricity (47% “very aware,” 28% “somewhat aware”).
• Only 13% say they were “not aware at all” of this issue.

Interpretation: Public awareness of AI’s energy demand is already high; this is not a niche or expert-only concern.

Concern about energy demand, prices, and grid strain

• 76% are concerned that AI data centers will drive up electricity demand and prices (“somewhat” or “very” concerned).
• 75% are concerned that AI data centers could increase household energy bills in their area.
• 71% agree that “AI data centers will strain the electricity grid in my country.”

Interpretation: The public now sees AI as a factor in national energy security and household costs.

Consumers oppose ‘AI at any cost’

• If new AI data centers were expected to increase household electricity bills, 64% say they would oppose further AI expansion; only 7% would support it.
• 74% support limits or temporary pauses on high‑emissions “dirty AI" data centers until cleaner power is available.
• 81% agree with the statement: “I don’t want technology companies to roll out more AI services if doing so increases my household energy bills.”

Interpretation: People do not oppose AI itself, but they strongly oppose business‑as‑usual if it drives up bills and emissions.

Who should pay for AI‑driven grid upgrades?

• 80% say technology companies should pay for upgrades to the electricity grid needed to support new AI data centers.
• 12% say governments should pay; just 1% say consumers/households.
• 7% say the cost should be shared equally.

Interpretation: There is a strong cross‑class consensus that big tech – not bill‑payers – should shoulder the cost of AI infrastructure.

Strong demand for transparency and ‘energy labels’

• 85% support requiring AI services to show “energy labels” revealing how much electricity they use, similar to nutrition labels on food.
• 81% say those labels would influence which services they choose (31% “a lot,” 50% “somewhat”).

Interpretation: There is a powerful mandate for simple, accessible disclosure of AI energy use that would directly shape consumer behavior.

Pro‑efficiency, pro‑renewables, not anti‑AI

• 83% say AI companies should prioritize making AI more energy‑efficient, even if that slows down how fast new AI features are released.
• 82% support requiring large AI companies to use mostly or entirely renewable energy to power their services.

Interpretation: The public wants AI to move forward, but they expect the next phase of growth to be built on efficiency and clean power.

Strong support for sovereign, locally governed AI

• 91% say it is important that their country has its own AI systems, rather than relying mostly on foreign companies; 67% call this very or extremely important.
• 66% would feel more comfortable using AI services if they knew those services were hosted and governed within their own country or region; only 3% would feel less comfortable.

Interpretation: Sovereign AI is not just a policy slogan; it is a mainstream consumer expectation.

Consumers want AI investment – done responsibly

• 73% believe their government is investing about the right amount or more than enough in AI to keep their country competitive.
• 27% say their government is investing too little or far too little.
• Interesting regional difference: 77% in the US vs 68% in the UK.

Interpretation: People broadly support continued AI investment by governments, but want that investment aligned with efficiency, transparency, and national control.

REGIONAL DIFFERENCES

Comparing UK to US findings:

• Concern that AI data centers will drive up electricity demand and prices: 77% in the US vs. 75% in the UK.
• Concern that AI data centers could increase household energy bills in their area: 76% in the US vs. 74% in the UK.
• Agreement that AI data centers will strain the national grid: 73% in the US vs. 68% in the UK.
• Support for limits on high‑emissions “dirty AI” data centers until cleaner power is available: 73% in the US vs. 76% in the UK.
• Belief that tech companies should pay for grid upgrades: 79% in the US vs. 81% in the UK.
• Support for AI energy labels highlighting energy usage. The labels would indicate pressure on the energy grid and the amount required to generate a response: 85% in the US vs. 86% in the UK.
• Support for requiring renewables for large AI companies: 79% in the US vs. 87% in the UK.
• Perception that the government is investing enough in AI: 77% in the US vs. 68% in the UK.
• Importance of having domestic AI systems: 90% in both countries, with very/extremely important at 69% in the US and 63% in the UK.
• Feeling more comfortable with locally hosted and governed AI: 66% in US and 68% in UK.

Interpretation: US respondents are slightly more worried about grid strain, while UK respondents are slightly more demanding on renewables. Support for sovereign, efficient AI is strong on both sides of the Atlantic.

DEMOGRAPHIC DIFFERENCES

Age differences – young vs. old:

• Regular AI use (“very familiar – I use them regularly,” Q10) is highest among younger respondents: 47% of under‑40s, 42% of 40–59‑year‑olds, and 33% of those aged 60+.
• Opposition to further AI expansion if bills rise is strongest among younger respondents: 69% of under‑40s would oppose expansion, compared with 61% of 40–59‑year‑olds and 53% of those 60+.
• Sovereign AI matters most to older respondents: 83% of those 60+ say it is very or extremely important that their country has its own AI systems, versus 70% of 40–59‑year‑olds and 57% of under‑40s.
• Comfort with locally hosted and governed AI follows the same pattern: 81% of those 60+ would feel more comfortable, versus 67% of 40–59‑year‑olds and 56% of under‑40s.

Interpretation: Younger people are the heaviest AI users but also the quickest to push back if AI drives up bills, while older people are the strongest advocates for sovereign, locally governed AI.

Income differences – lower vs. higher income:

• Opposition to further AI expansion if bills rise is highest among lower‑income respondents: 65% of lower‑income, 63% of middle‑income, and 58% of higher‑income respondents would oppose expansion.
• Across all income bands, majorities say technology companies should pay for grid upgrades: 79% of lower‑income, 79% of middle‑income, and 82% of higher‑income respondents.
• Support for making AI more energy‑efficient, even if it slows feature roll‑outs, is consistently high: around 81–83% across all income bands.
• Support for AI energy labels is also high across the board: 85% of middle‑income, 85% of lower‑income, and 80% of higher‑income respondents.
• Sovereign AI is widely supported: around 71% of higher‑income, 70% of middle‑income, and 64% of lower‑income respondents say domestic AI systems are very or extremely important.

Interpretation: Pro‑efficiency, pro‑renewables, and pro‑sovereign AI sentiment is shared across income groups. The differences are relatively modest, probably not newsworthy.

QUALITATIVE INSIGHTS FROM OPEN-ENDED RESPONSES

In addition to structured questions, the survey included a single open-ended item: "In a sentence or two, what is your biggest worry about AI’s impact on energy, the environment, or society?”

This question generated hundreds of rich responses. We coded each answer into themes, allowing multiple themes per response. The analysis below converts those themes into quantitative signals.

1. Bills and climate dominate consumer worries

Around 46% of respondents explicitly mention energy use, electricity or bills when describing their biggest worry, and roughly 23% mention climate change, emissions, pollution or environmental damage.

Typical comments blend cost of living and climate into a single worry:

• “Increased energy usage, causing more climate change and higher cost of electricity.”

• “I worry how this will increase the cost of living for average people.”

• “I wouldn’t want AI to increase anyone’s electricity bill.”

Taken together, the open responses confirm the central tension identified in the quantitative data: People like AI, but they do not want it to drive higher bills or undermine climate progress.

2. Jobs and economic impact leak into an energy question

Even though the question asked specifically about energy, the environment and society, around 21% of respondents still bring up job loss, unemployment or broader economic disruption.

Examples include:

• “AI could take over many jobs and that’s what I worry about most.”

• “Number one is job loss, number 2 is nobody has to think for themselves anymore.”

• "My biggest worry is not AI in the energy sector, it is losing my job to AI. I do however want my energy bills to stay down.”

This suggests people experience AI as both potentially inflationary (through higher energy and infrastructure costs) and disruptive to jobs – a backdrop that strengthens the case for highly efficient AI that delivers real economic value.

3. The hidden ‘water cost’ of AI is emerging

About 8% of respondents explicitly mention water use or water scarcity when describing their biggest worry. For a relatively technical issue, this is a high level of concern.

Examples include:

• “It will use so much water that we will eventually have a water crisis because of lack of water.”

• “It’s a tremendous problem but from what I’ve read, most of the environmental concerns are about energy. The water issue is something else though.”

• “I think it will be the downfall of humanity… drives up costs of electricity, decimates our water supply, etc.”

This provides an additional narrative thread for the AI Energy Reckoning: Beyond electricity and emissions, citizens are starting to worry about the water footprint of AI data centers.

4. Society and human behavior: laziness, dependence and inequality

Around 16% of respondents focus their biggest worry on how AI will change society. They talk about people becoming “lazy” or over-dependent on AI, or widening gaps between those who can harness AI and those who cannot.

Examples include:

• “That we will depend way too much on it, and that it will take our jobs and make us lazy.”

• “People will get lazy.”

• “My biggest worry is the effect on people – that casual use will make people more dependent on AI and that artists and models will lose work to Gen AI content.”

These concerns support a framing of AI that enhances and augments human capability, rather than eroding skills or agency.

5. Misuse, misinformation and loss of control

Roughly 10% of respondents explicitly reference misuse, danger, “downfall,” deepfakes or misinformation when asked about AI’s impact on energy, the environment or society.

Illustrative comments include:

• “I worry that society is placing too much emphasis on AI technology and not placing enough safeguards on the malicious uses of AI.”

• “I worry about it getting into the hands of people who would use it to spread misinformation. We’ve already seen that here with vaccines and elections.”

• “I think AI is being used to spread misinformation across social media. I think it is harmful for the environment as well as society.”

Even in an energy-focused survey, people link AI’s physical impacts (energy, emissions) with its digital harms (misinformation, social polarization).

6. Fairness: Who benefits, who pays?

A recurring thread across responses is fairness: A sense that AI’s benefits accrue to large technology companies, while ordinary people are left carrying the cost in higher bills, higher prices or local environmental impacts.

Examples include:

• “I worry that the data centers placed in rural areas could hurt those areas more than helping them.”

• “We’re investing way too much, way too fast, and it’s going to push energy prices and other electronics way up which is very frustrating as a consumer.

This dovetails with the quantitative finding that 80% of consumers say technology companies – not households – should pay for grid upgrades needed for AI, reinforcing a clear public mandate for "clean, fair AI."

7. A small minority express no major worries

Only around 2% of respondents say they have no real worries about AI’s impact on energy, the environment or society, or give a response that is essentially neutral. The overwhelming majority articulate at least one concrete concern.

This does not mean the public is anti‑AI – usage and familiarity are high – but it does mean that citizens expect AI to be delivered in a way that is managed. Efficient, renewable‑powered, sovereign AI speaks directly to this.