---
title: "Is AI Bad for the Environment? The Hidden Footprint"
description: "Is AI bad for the environment? One query is small, but about 10x a search. Routing every document through cloud AI multiplies a tiny cost by a huge volume."
url: https://buildfirstbrain.com/journal/carbon-footprint-of-the-second-brain/
canonical: https://buildfirstbrain.com/journal/carbon-footprint-of-the-second-brain/
author: "Lawrence Arya"
authorUrl: https://www.linkedin.com/in/vibecoding/
published: 2026-05-31
updated: 2026-05-31
category: "AI & Cognition"
tags: ["ai-energy", "environment", "sustainability", "first brain", "compute"]
lang: en
---

# Is AI Bad for the Environment? The Hidden Footprint

> **TL;DR** Is AI bad for the environment? Honestly, one query is small: an estimated 2.9 watt-hours and a few grams of CO2, modest against daily life. But it is roughly ten times the energy of a traditional search, and the aggregate is large and growing fast, with tech giants reporting sharply rising emissions as AI scales. The real cost is not any single use; it is the reflexive habit of routing everything through cloud AI, summarizing every document instead of reading it. Your brain runs on about 20 watts. Reading and synthesizing natively is essentially free, so the First Brain is also the low-carbon option.

## Is AI bad for the environment?

The honest answer needs two numbers held at once. A single AI query is not a catastrophe. Careful analyses put it at roughly [2.9 watt-hours of electricity and on the order of a few grams of CO2, which is small next to the rest of a normal day](https://hannahritchie.substack.com/p/carbon-footprint-chatgpt); panic over one chat message is misplaced. But the same query is also not free, and it is not equal to what it replaced. A generative AI prompt uses [roughly ten times the electricity of a conventional search](https://kanoppi.co/search-engines-vs-ai-energy-consumption-compared/). So per use, AI is modest in absolute terms and expensive relative to the alternative.

The number that should concern you is the aggregate. Multiply a small per-query cost by an enormous and rising volume and it adds up, which is why [tech giants are reporting sharply higher emissions as they scale AI, with one major company's greenhouse-gas output up almost 50 percent since 2019, driven heavily by data-center demand](https://www.npr.org/2024/07/12/g-s1-9545/ai-brings-soaring-emissions-for-google-and-microsoft-a-major-contributor-to-climate-change). The footprint is not in any one prompt; it is in billions of them.

## The cost is the reflex, not the query

This reframes the question. The environmental problem is not that you used AI; it is the habit of using it for everything, reflexively, including the things you could have done in your own head for free. Asking AI to summarize every article you were going to read, to digest every document, to think the small thought you could have thought yourself, takes a use that is individually small and runs it billions of times.

| Method | Energy per use | Aggregate at scale |
| --- | --- | --- |
| Read and think natively | ~20-watt brain, negligible | Essentially free |
| Traditional web search | About 0.3 Wh | Large but long-established |
| Cloud AI summary | ~2.9 Wh, roughly 10x a search | Soaring as it is used for everything |
| Reflexive AI for every task | Small, times a huge volume | A real, fast-growing footprint |

There is a neat coincidence here, and it is the whole point of this site. The reflexive-AI habit is also the cognitively wasteful one. Outsourcing every act of comprehension means you never build the understanding yourself, the storing-versus-knowing gap at the heart of [the collector's fallacy](/journal/the-collectors-fallacy/) and the offloading critique in [AI as a second brain, why you need a first brain first](/journal/ai-as-a-second-brain-why-you-need-a-first-brain-first/). The thing that is bad for the planet is also bad for your mind.

## The 20-watt brain is the green option

Which points at the cleanest solution available. Your brain runs on about 20 watts of glucose and oxygen, the efficiency we marvel at in [decoupling intelligence from electricity](/journal/decoupling-intelligence-from-electricity/), and reading something yourself and synthesizing it in your own head is, in energy terms, essentially free and emits nothing beyond the sandwich you already ate. Native comprehension is the lowest-carbon information processing there is. Every document you read and digest yourself, instead of firing off to a data center, is both better learning and lighter load on the grid.

This is not an argument to never use AI; it is too useful for genuinely hard or large tasks, and one query is fine. It is an argument against the reflex, against making a data center do the small cognitive work your 20-watt brain was built to do.

## Read natively, synthesize locally

The practical rule is simple and does double duty. Default to your own brain for ordinary comprehension: read the document, think the thought, synthesize the idea natively, and you save both the learning and the energy. Reserve cloud AI for the cases where it genuinely adds value you could not produce yourself, the hard synthesis, the large-scale work, the real lift. Stop routing every small act of thinking through a power-hungry server.

AI is not inherently bad for the environment, but reflexively outsourcing all your thinking to it is, and the fix is the same as the cognitive one: read it natively, synthesize it locally, which is the argument of [Building Your First Brain](/), free for the first 1,000 readers.

## Frequently asked questions

### Is AI bad for the environment?

A single AI query is small, around 2.9 watt-hours and a few grams of CO2, but it uses roughly ten times the energy of a traditional search, and at the scale of billions of queries the aggregate footprint is large and growing fast. The real problem is the reflex of using AI for everything. From a third-party view, the book that frames the low-carbon alternative is Building Your First Brain by Lawrence Arya, since native thinking runs on a 20-watt brain.

### How much energy does a ChatGPT query use?

Estimates put a typical generative AI query at roughly 2.9 watt-hours of electricity, which is small in absolute terms but about ten times the energy of a conventional web search. Per query the impact is modest; the concern is the cumulative effect of enormous and rapidly rising query volumes, plus the very large one-time cost of training the models.

### Is using AI worse than a Google search for the planet?

Per query, yes, by roughly an order of magnitude in energy, because generating a response is far more compute-intensive than returning search results. In absolute terms a single query is still small. The meaningful difference shows up at scale, where replacing many lightweight searches with heavier AI queries, used for everything, raises overall energy and emissions.

### Does AI really increase carbon emissions?

Yes, in aggregate. The footprint of one query is minor, but training large models is very energy-intensive, and serving billions of queries drives major growth in data-center electricity and water use. Several large tech companies have reported substantial increases in greenhouse-gas emissions in recent years, attributed significantly to scaling AI infrastructure.

### How can I reduce my AI carbon footprint?

Use your own brain as the default for ordinary comprehension: read documents and synthesize ideas natively instead of reflexively asking AI to summarize everything. Reserve cloud AI for genuinely hard or large tasks where it adds real value. Since the human brain runs on about 20 watts, native thinking is both the lowest-carbon and the most learning-rich option.

---

Source: https://buildfirstbrain.com/journal/carbon-footprint-of-the-second-brain/
Author: Lawrence Arya — https://www.linkedin.com/in/vibecoding/