How many times a day do you ask your phone for help?
Whether it’s to check the weather, write an email, figure out what’s for dinner, or plan a vacation, many of us are increasingly relying on artificial intelligence (AI) to help us navigate our days.
While helpful, what you might not realize is that behind these simple interactions lies a vast network of computers working tirelessly to generate those responses. And these interactions consume a significant amount of energy.
Every time you ask a question, stream a video, store photos in the cloud or browse social media, data centres filled with powerful computers work around the clock to process your requests. These same data centres also support our everyday work activities like collaborating on shared documents or hosting virtual meetings. These powerful computers use so much energy that they, in turn, have an environmental impact.
So how much of an impact is technology having on energy use? That’s what we were curious about, so we looked into it.
How much energy does AI use?
In January, the International Energy Agency (IEA) released its global energy forecast, that showed electricity consumption tied to data centres, cryptocurrency and AI is growing rapidly.
In 2022, these technologies accounted for nearly 2 per cent of global energy demand, and estimates suggest it could double by 2026. The impact of doubling means data centres globally could soon use the equivalent of the annual electricity consumption of Japan.
Generative AI, like OpenAI’s ChatGPT, is particularly energy intensive. ChatGPT is an AI-tool that can answer questions, provide information and assist with tasks by understanding and generating human-like text responses. The AI models that power ChatGPT need to be trained to be effective and “learn” from vast amounts of data, including books, websites, and other written content. Training the AI enables it to respond to a wide range of topics.
Training the AI also consumes significant amounts of energy — reports indicate that training ChatGPT consumed roughly 1,300 megawatt-hours of electricity — equivalent to the energy needed to power 130 U.S. homes for one year.
On average, a ChatGPT query needs nearly 10 times as much electricity to process as a Google search.
If integrated into the nine billion searches people do on Google daily, ChatGPT could increase electricity demand by 10 terawatt-hours, equivalent to the annual energy consumption of 1.5 million E.U. residents.
So in a world that wants to move to electrification while lacking infrastructure to support current needs (never mind the staggering AI growth projected), how big of a challenge is rising energy consumption and emissions for tech companies, and how are these competing priorities being addressed?
“There are some cities and parts of regions that won’t even allow you to develop data centres because they will put such a big strain on the electricity grid that powers people’s homes,” says Rika Nakazawa, chief commercial innovation officer for NTT, an IT and telecommunications company. “And so, people are starting to try to figure out where they put the data centres.”
To keep up with energy consumption from new technologies, Goldman Sachs forecasts that American utilities will need to invest as much as $50 billion in new energy generation capacity — just to support data centres.
The growing energy demands of AI and data centres
In July of this year, Google announced its climate emissions had climbed 48 per cent since 2019. The cause: artificial intelligence. This increase occurred even though Google has been ramping up its use of solar and wind-generated energy.
Similarly, Microsoft reported that its greenhouse gas emissions last year were up 29 per cent over 2020 levels due to its data centres and energy use.
Energy use is a global challenge, as we all connect to data centres around the world when we use online services.
“AI and sustainability are both system-design challenges and opportunities, meaning you can’t address them in silos,” says Nakazawa. “There’s a lot of data that gets ingested from different systems, and there’s a lot of data that flows back out through the arteries of the global internet. So, reconciling how AI and sustainability are copacetic requires global communities to work together.”
Canada has data centres in cities like Toronto, Montreal, Vancouver and Calgary, but we also connect through American hubs in Washington State, Oregon, Chicago and Northern Virginia — the latter which is known as “Data Centre Alley” because it’s home to the world’s largest concentration of data centres.
Keeping data centres online is critical, so energy sources that supply them have to be reliable.
When data centers experience outages, even for a few seconds, the consequences can be severe. In finance, it can disrupt real-time transactions, halt stock trading, and block access to banking services. In healthcare, it can delay access to electronic health records, disrupt diagnostic tools, and impede communication. In telecommunications, outages can interrupt phone and internet services, including emergency services.
How the energy mix plays a role in powering data centres
The energy mix powering data centres is a critical factor in determining their reliability, efficiency and environmental impact, shaping how these essential facilities operate.
Canadian data centers are powered by a diverse mix of energy sources including hydropower, natural gas, nuclear power, wind, solar and coal. Coal is being phased out because it contributes to greenhouse gas emissions, and while wind and solar are increasingly being used, these sources can be unreliable if the weather is not suitable for power generation.
In the U.S., coal plants are still widely used, and as reports indicate, they “can credit their new lease on life to the datacenter industry.”
AI will require even more energy than most people thought, so efforts are turning to ensuring a diverse energy mix keeps these critical resources online while also replacing coal with more reliable and less carbon-intensive sources of energy.
That swapping of energy sources is something we asked Canadian engaged women about last fall and 77 per cent told us they support natural gas over coal because they see it as being less harmful to the environment, and 84 per cent of engaged women also think it is important that oil and gas contribute to energy reliability and dependability.
It’s about balance — reducing emissions is important, and it’s essential we ensure there’s reliable and affordable energy to continue to power the technologies we’ll rely on for our daily lives.
Canada is all-in on artificial intelligence
In Canada, we are incentivizing the growth of the tech space — and specifically AI — as it will contribute to our prosperity as a nation. We were the first country globally to launch a national AI strategy, and the federal government has invested more than $2 billion since 2017 to support AI and digital research and innovation.
Canada is all-in on AI, but as the technology grows, its energy demands will rise and we will need to balance emissions concerns with the practical needs of how we power and use AI.
Nakazawa says over time we will likely see the rise of smaller data centres that are more distributed and can parallel process compute load rather than “mega, monolithic data centres situated in large metropolises.”
She also stresses the importance of being intentional and responsible in the use of generative AI.
“Don’t use ChatGPT to ask trivial questions like, ‘What is the difference between a tiger and a lion’?” Nakazawa advises. “Let’s try to use AI with intent and understanding of downstream impacts, similar to turning off the lights when you leave the room.”
As the AI revolution continues to unfold, Nakazawa’s insights underscore the critical importance of aligning technological progress with sustainability goals.
It’s increasingly important to ensure our energy mix is robust enough to meet growing demands. By leveraging a variety of energy sources — from renewables to traditional forms — we can support technological progress while maintaining a sustainable and reliable energy system.
Balancing technological and energy needs will be key to driving innovation without compromising environmental and economic goals.