{"course":{"slug":"ai-actually-explained","title":"AI, Actually Explained","url":"https://addaly.com/learn/ai-actually-explained","level":"beginner","track":"start"},"lesson":{"slug":"what-a-query-costs","title":"What a query costs, in rupees and watt-hours","url":"https://addaly.com/learn/ai-actually-explained/what-a-query-costs","position":67,"of":92,"minutes":9,"answer":"A page in and a page out costs a few paise on a small model and about a rupee on a frontier one, and around a third of a watt-hour of electricity: small per query.","keyIdea":"A page in and a page out costs a few paise on a small model and about a rupee on a frontier one, and around a third of a watt-hour of electricity: small per query, and the total is not small."},"body":{"format":"markdown","text":"## The unit is the token\n\nThe chatbot module told you the model's cost is measured in tokens, not words. This lesson puts prices on it, because knowing the order of magnitude changes how you use the thing and how you read the headlines about it.\n\nA page of English prose (about five hundred words) is roughly six hundred and fifty to seven hundred tokens. Hindi in Devanagari runs two to three times as many tokens per word, for the reason the module on being weaker in your language gave. Keep that page in mind as the unit.\n\n## The money\n\nLabs sell access to developers by the million tokens, and they charge more for the tokens a model writes than for the ones it reads, because writing is the sequential, slow part the chip lesson described. In 2025 the prices fell into two bands.\n\nThe **small models** (the ones that answer the free tiers and most business chatbots) cost around ten to fifteen cents of a US dollar per million tokens read and forty to sixty cents per million written. A page in and a page out on one of those is about five hundredths of a US cent: four or five paise.\n\nThe **frontier models** cost roughly twenty times more: a few dollars per million read, ten to fifteen per million written, and considerably more for the models that reason at length before answering. A page in and a page out is about one and a quarter US cents: around a rupee.\n\nA subscription of twenty dollars a month buys, at frontier prices, something like a thousand to two thousand such exchanges, which is why the subscription is the better deal for a heavy user and a poor one for most people, and why the free tier exists at all.\n\n## The electricity\n\nHere the numbers are more contested, and it is worth knowing why.\n\nFor two years the figure quoted everywhere was about three watt-hours per chatbot query, ten times a web search. It came from a 2023 estimate built on the hardware of the time and the assumption that every query ran on a large model. In 2025 two of the largest labs published their own measurements: a median text prompt of about a quarter of a watt-hour on one and a third on the other, including the idle capacity and cooling around the chip. The fall is real (smaller models and better chips) and so is the caveat: the figures are company-reported, cover a median text query, and exclude training.\n\nTo place a third of a watt-hour: a phone battery holds about fifteen watt-hours, so around fifty text queries is one full charge. A ten-watt bulb burns it in two minutes.\n\nImages cost more. A 2023 measurement across many models found generating one image took on the order of a few watt-hours: one image, roughly ten text queries. Video costs more again: minutes of a data-centre chip per clip, tens of watt-hours. And the reasoning models that write thousands of tokens of working before answering cost, per question, ten to a hundred times a plain reply, because the cost is in the tokens written.\n\n> Figure: Electricity per thing produced, order of magnitude. Bars (watt-hours): One median text reply 0.3; One generated image 3; One answer with long written reasoning 6; Ten seconds of generated video 30. The text figure is the labs' own 2025 measurement, including cooling and idle capacity, and excluding training. A phone battery holds about fifteen watt-hours, so roughly fifty text replies is one charge. The per-query number is small; a billion queries a day is not, and both statements are true at once.\n\nWater follows electricity: the same 2025 reports put a text query at a fraction of a millilitre, a few drops, mostly for cooling.\n\n## Small each, large together\n\nA third of a watt-hour is nothing. A billion queries a day is roughly three hundred thousand kilowatt-hours a day for text alone, before images, video, reasoning and training. Data centres as a whole drew about one and a half per cent of the world's electricity in 2024, and the agency that tracks it expects the figure to roughly double by 2030, with AI the fastest-growing part. The final module of this course puts that in its place. For this lesson, the honest position is both halves at once: your query is cheap, and the system is not.\n\n## What the training cost adds\n\nSpread the electricity of a frontier training run (tens of gigawatt-hours, from the history module) across the billions of queries the model then answers, and it adds a fraction of a watt-hour to each. Training is a large fixed cost that inference dilutes; within a year of a popular model's release, serving it has used more power than making it did.\n\n## How to use the numbers\n\nUse the small model for the reliable-column tasks; the frontier gains you little on a rewrite and costs twenty times more. Send the page, not the book, when the page is what matters. Do not ask a reasoning model to fix a comma. And when a headline gives a single figure for \"the cost of AI\", ask the questions this lesson gave you: per query or per day, text or image, which model, who measured it, and does it include training. The number will usually be true for one of those and reported as if it were true for all of them."},"check":{"question":"A newspaper reports that a single chatbot query uses ten times the electricity of a web search. Why should you not take the figure at face value?","options":[{"text":"Because chatbots do not use electricity in any meaningful amount.","correct":false,"tempting":"Overcorrects to nothing; a query does draw measurable power, and the aggregate is large."},{"text":"Because it rests on a 2023 estimate for a large model on old hardware; 2025 lab measurements are about a tenth of it.","correct":true},{"text":"Because web searches use far more electricity than the comparison assumes, so the ratio is inverted and the chatbot is the cheaper of the two.","correct":false,"tempting":"Moves the error to the wrong side; the search figure is old but not the source of the tenfold gap."},{"text":"Because electricity use depends only on the length of the answer, not on which model produced it.","correct":false,"tempting":"Ignores the model; the same answer costs twenty times more from a frontier model than a small one."}]},"prev":{"slug":"running-one-on-your-own-machine","title":"A model on your own laptop or phone","url":"https://addaly.com/learn/ai-actually-explained/running-one-on-your-own-machine"},"next":{"slug":"the-assistant-inside-your-apps","title":"The assistant inside your email and spreadsheet","url":"https://addaly.com/learn/ai-actually-explained/the-assistant-inside-your-apps"},"authorship":"Founder and series editor: Dr Mustafa Mun. Written by AI models to a written standard. Exam keys checked blind by a second model: 3,665 of 3,666 agreed. Academic review invited.","review":{"status":"Draft, open for academic review"},"cite":"Addaly (2026) 'What a query costs, in rupees and watt-hours', in AI, Actually Explained. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/ai-actually-explained/what-a-query-costs (Accessed: 13 October 2026).","licence":{"name":"CC BY-NC-SA 4.0","url":"https://creativecommons.org/licenses/by-nc-sa/4.0/"},"reader":{"schema":1,"title":"What a query costs, in rupees and watt-hours","courseTitle":"AI, Actually Explained","module":{"index":7,"slug":"putting-it-to-work","title":"Putting it to work without getting burned"},"answer":null,"keyIdea":"A page in and a page out costs a few paise on a small model and about a rupee on a frontier one, and around a third of a watt-hour of electricity: small per query, and the total is not small.","hero":{"kind":"figure","figure":{"t":"figure","label":"Figure","figure":{"type":"bars","unit":"watt-hours","title":"Electricity per thing produced, order of magnitude","series":[{"label":"One median text reply","value":0.3,"highlight":true},{"label":"One generated image","value":3},{"label":"One answer with long written reasoning","value":6},{"label":"Ten seconds of generated video","value":30}],"caption":"The text figure is the labs’ own 2025 measurement, including cooling and idle capacity, and excluding training. A phone battery holds about fifteen watt-hours, so roughly fifty text replies is one charge. The per-query number is small; a billion queries a day is not, and both statements are true at once."},"alt":"Figure: Electricity per thing produced, order of magnitude. Bars (watt-hours): One median text reply 0.3; One generated image 3; One answer with long written reasoning 6; Ten seconds of generated video 30. The text figure is the labs’ own 2025 measurement, including cooling and idle capacity, and excluding training. A phone battery holds about fifteen watt-hours, so roughly fifty text replies is one charge. The per-query number is small; a billion queries a day is not, and both statements are true at once."}},"blocks":[{"t":"h2","id":"the-unit-is-the-token","runs":[{"text":"The unit is the token"}]},{"t":"p","runs":[{"text":"The chatbot module told you the model’s cost is measured in tokens, not words. This lesson puts prices on it, because knowing the order of magnitude changes how you use the thing and how you read the headlines about it."}]},{"t":"p","runs":[{"text":"A page of English prose (about five hundred words) is roughly six hundred and fifty to seven hundred tokens. Hindi in Devanagari runs two to three times as many tokens per word, for the reason the module on being weaker in your language gave. Keep that page in mind as the unit."}]},{"t":"h2","id":"the-money","runs":[{"text":"The money"}]},{"t":"p","runs":[{"text":"Labs sell access to developers by the million tokens, and they charge more for the tokens a model writes than for the ones it reads, because writing is the sequential, slow part the chip lesson described. In 2025 the prices fell into two bands."}]},{"t":"p","runs":[{"text":"The "},{"text":"small models","b":true},{"text":" (the ones that answer the free tiers and most business chatbots) cost around ten to fifteen cents of a US dollar per million tokens read and forty to sixty cents per million written. A page in and a page out on one of those is about five hundredths of a US cent: four or five paise."}]},{"t":"p","runs":[{"text":"The "},{"text":"frontier models","b":true},{"text":" cost roughly twenty times more: a few dollars per million read, ten to fifteen per million written, and considerably more for the models that reason at length before answering. A page in and a page out is about one and a quarter US cents: around a rupee."}]},{"t":"p","runs":[{"text":"A subscription of twenty dollars a month buys, at frontier prices, something like a thousand to two thousand such exchanges, which is why the subscription is the better deal for a heavy user and a poor one for most people, and why the free tier exists at all."}]},{"t":"h2","id":"the-electricity","runs":[{"text":"The electricity"}]},{"t":"p","runs":[{"text":"Here the numbers are more contested, and it is worth knowing why."}]},{"t":"p","runs":[{"text":"For two years the figure quoted everywhere was about three watt-hours per chatbot query, ten times a web search. It came from a 2023 estimate built on the hardware of the time and the assumption that every query ran on a large model. In 2025 two of the largest labs published their own measurements: a median text prompt of about a quarter of a watt-hour on one and a third on the other, including the idle capacity and cooling around the chip. The fall is real (smaller models and better chips) and so is the caveat: the figures are company-reported, cover a median text query, and exclude training."}]},{"t":"p","runs":[{"text":"To place a third of a watt-hour: a phone battery holds about fifteen watt-hours, so around fifty text queries is one full charge. A ten-watt bulb burns it in two minutes."}]},{"t":"p","runs":[{"text":"Images cost more. A 2023 measurement across many models found generating one image took on the order of a few watt-hours: one image, roughly ten text queries. Video costs more again: minutes of a data-centre chip per clip, tens of watt-hours. And the reasoning models that write thousands of tokens of working before answering cost, per question, ten to a hundred times a plain reply, because the cost is in the tokens written."}]},{"t":"p","runs":[{"text":"Water follows electricity: the same 2025 reports put a text query at a fraction of a millilitre, a few drops, mostly for cooling."}]},{"t":"h2","id":"small-each-large-together","runs":[{"text":"Small each, large together"}]},{"t":"p","runs":[{"text":"A third of a watt-hour is nothing. A billion queries a day is roughly three hundred thousand kilowatt-hours a day for text alone, before images, video, reasoning and training. Data centres as a whole drew about one and a half per cent of the world’s electricity in 2024, and the agency that tracks it expects the figure to roughly double by 2030, with AI the fastest-growing part. The final module of this course puts that in its place. For this lesson, the honest position is both halves at once: your query is cheap, and the system is not."}]},{"t":"h2","id":"what-the-training-cost-adds","runs":[{"text":"What the training cost adds"}]},{"t":"p","runs":[{"text":"Spread the electricity of a frontier training run (tens of gigawatt-hours, from the history module) across the billions of queries the model then answers, and it adds a fraction of a watt-hour to each. Training is a large fixed cost that inference dilutes; within a year of a popular model’s release, serving it has used more power than making it did."}]},{"t":"h2","id":"how-to-use-the-numbers","runs":[{"text":"How to use the numbers"}]},{"t":"p","runs":[{"text":"Use the small model for the reliable-column tasks; the frontier gains you little on a rewrite and costs twenty times more. Send the page, not the book, when the page is what matters. Do not ask a reasoning model to fix a comma. And when a headline gives a single figure for “the cost of AI”, ask the questions this lesson gave you: per query or per day, text or image, which model, who measured it, and does it include training. The number will usually be true for one of those and reported as if it were true for all of them."}]}],"check":{"question":[{"text":"A newspaper reports that a single chatbot query uses ten times the electricity of a web search. Why should you not take the figure at face value?"}],"options":[{"text":[{"text":"Because chatbots do not use electricity in any meaningful amount."}],"correct":false,"tempting":[{"text":"Overcorrects to nothing; a query does draw measurable power, and the aggregate is large."}]},{"text":[{"text":"Because it rests on a 2023 estimate for a large model on old hardware; 2025 lab measurements are about a tenth of it."}],"correct":true,"tempting":null},{"text":[{"text":"Because web searches use far more electricity than the comparison assumes, so the ratio is inverted and the chatbot is the cheaper of the two."}],"correct":false,"tempting":[{"text":"Moves the error to the wrong side; the search figure is old but not the source of the tenfold gap."}]},{"text":[{"text":"Because electricity use depends only on the length of the answer, not on which model produced it."}],"correct":false,"tempting":[{"text":"Ignores the model; the same answer costs twenty times more from a frontier model than a small one."}]}],"order":[3,1,0,2],"why":null},"prev":{"slug":"running-one-on-your-own-machine","title":"A model on your own laptop or phone","minutes":10},"next":{"slug":"the-assistant-inside-your-apps","title":"The assistant inside your email and spreadsheet","minutes":9}}}