Saying “please” and “thank you” to your AI assistant is wasting millions of dollars in electricity costs

The CEO of OpenAI estimated that the company has likely lost tens of millions of dollars due to users’ polite interactions with AI models. The significant power consumption required for AI interactions is straining electricity grids, and projections suggest that data centres could account for a considerable increase in electricity demand in the US and globally.

When a user on X, formerly Twitter, asked “how much money OpenAI has lost in electricity costs from people saying “please” and “thank you” to their models”, OpenAI CEO and billionaire Sam Altman replied that it probably cost his AI-company “tens of millions of dollars”.

A US-survey from last year revealed that around 67 percents of the respondents are being nice to their AI assistants and chatbots — presumably saying “please” and “thank you”. Of those, around 55 percent said they say are being nice “because it’s the right thing to do,” while the other 12 percent did it — jokingly, or not — in order to appease the AI in case there ever was an AI uprising.

AI assistants and chatbots use a huge amount of power to respond to their user’s questions. And clearly, a lot of people are being nice to their AI chatbots and writing longer and “unnecessary” prompts, which unfortunately are costing AI companies a lot and putting a huge strain on the electricity grid — and ultimately, the environment.

For example, one ChatGPT query consumes ten times the electricity needed to process a Google search engine query. And an investigation by the Washington Post, which was done with the help from researchers from the University of California, revealed the hidden environmental costs of using AI chatbots. It showed that sending one AI-composed email requires 0.14 kilowatt-hours (kWh) of electricity, which is enough to power 14 LED lights for an hour. And if you were to send an AI-composed email per week over the course of a year, that would require 7.5 kWh, which is roughly the same amount of electricity consumed in one hour by nine households in Washington DC.

All that computing power quickly adds up, big-time.

Growth in data centres is driving electricity demand in USA

According to projections by the Electric Power Research Institute, data centres could consume as much as 9.1 percent of all electricity in the US by the end of the decade. Mind you, those are for all data centres in the US, and not just specifically AI data centres. But much of this growth is expected to come from data centres powering the increasing use of AI.

Eric Schmidt, billionaire and former CEO of Google, has previously said at a hearing at the House Energy and Commerce Committee that “we [the AI sector] need energy, and the numbers are profound,” and that “we need energy in all forms. Renewable, nonrenewable, whatever. It needs to be there, and it needs to be there quickly.”

At the same hearing, Manish Bhatia, EVP of Micron Technology, highlighted projections that warned that due to the development of the AI sector, overall energy consumption in the US could increase by up to 15 percent within the next five years. Energy consumption in the US has traditionally increased by about 0.5 percent per year over the past several decades. Electricity usage in the US jumped by about 2 percent just last year due to the increased demand from new data centres, according to data by EIA.

That’s a huge and significant difference.

And Donald Trump wants all those AI data centres in the US to be powered by dirty coal — or by America’s “beautiful clean coal resources”, as the clown himself has said. “You know, we need to do the AI, all of this new technology that’s coming on line,” Trump said during a signing ceremony for the executive orders. “We need more than double the energy, the electricity, that we currently have.”

That’s why Trump has used his emergency authorization powers to relax environmental regulations on coal, and signed a series of executive orders that are designed to encourage the use of coal and help the declining US coal industry — which have seen a sharp decline from 2011 when coal represented nearly half of all US energy supply to only accounting for about 15 percent of all energy today.

Data centres are partly to blame for increased electricity demand globally

We are currently seeing a rising demand for electricity worldwide. Globally, electricity demand rose by 4.3 percent in 2024, and it will continue to grow at close to 4 percent annually during the next couple of years, according to a recently released report by the International Energy Agency. “We see one very clear trend: electricity growth,” said Fatih Birol, the IEA’s executive director.

Most of that electricity demand comes from China, India, and other developing countries in Southeast Asia.

In China, the increased electricity demand is coming from their huge and growing electric-vehicle market and heavy industries. It’s hard to comprehend the absolutely huge numbers we are seeing in China. But for example, last year about 300 terawatt-hours worth of electricity was used in China just to produce solar modules, batteries, and electric vehicles — that’s as much electricity as Italy uses in a whole year. That’s mind-boggling.

As such, a large part of the increase in electricity globally is due to the general electrification of everything. But a large part also comes from increased usage of air conditioning due to record high temperatures and extreme heat waves in India and China.

But new data centres still plays a role in driving up the demand globally. According to IEA’s 2024 World Energy Outlook projections, data centres are expected to account for about 10 percent of the total growth in global electricity demand between now and 2030.

While data centres are partly driving electricity demand globally, they and the AI industry clearly plays a bigger role in driving up the demand in developed nations — such as the US and European countries where electricity demand has largely been flat for the past decade, partly thanks to energy efficiency improvements.

Which energy source will power future data centres?

While Trump is pushing for coal and tech billionaires are looking at nuclear energy to feed the ever-growing electricity demand by the AI industry, the question is what energy source will power the future of AI and our digital life?

Luckily, the IEA says that around 80 percent of new electricity generation globally last year came from renewable energy or nuclear power. “If we want to find the silver lining, we see that there is a continuous decoupling of economic growth from emissions growth,” said Laura Cozzi, who was lead author of IEA’s 2025 edition of the Global Energy Review.

But by just adding renewables or low-carbon energy sources to meet the growing electricity demand isn’t enough to combat climate change. In order to actually make a dent in our greenhouse gas emissions, we need to quickly grow our clean energy sources so that they actually start replacing the existing fossil fuel infrastructure. Nuclear energy is simply just too costly and takes too long to build to be able to do that. Only renewable energy sources can be quickly and cost-effectively implemented on such a large scale that we can actually start to replace our dirty and polluting energy sources.

Until the transformation of our energy system actually happens, we all need to be considerate of how AI affects both our energy systems and climate, and take meaningful steps to limit that impact. A few years back, it was popular to have a small text in your email signature that urged the reader to “please consider the environment before printing this email”. It seemed meaningless in recent years, because who prints emails these days? But maybe we need something similar for this new AI web we are increasingly moving towards?