Artificial Intelligence is being promoted as the latest crisis in a world seemingly hungry for a continuous diet of crises (to which I alluded in this recent article for BIG Media). Considering that news media label every situation with potential negative impacts as a “crisis”, is the reported hysteria surrounding the rapid rise of AI just another media tactic, or do we really have something to worry about?
Let’s start with the basics – what is Artificial Intelligence, and how big is it?
AI is a term used in many ways – a “label that has been slapped on technologies ranging from self-driving cars to facial recognition, chatbots, to fancy Excel,” according to the MIT Technology Review. Trying to define AI more precisely, the same article called it: “technologies that make computers do things we think need intelligence when done by people.” I don’t think that really narrows it sufficiently.
Artificial Intelligence has for decades played important roles in science fiction, from Asimov-era robots and ship-borne computers, through Arthur C. Clarke’s HAL in 2001: A Space Odyssey, to galaxy-spanning AIs in Neal Asher’s space operas of the early 2000s. One can argue, however, that artificial intelligence only hit the real-world mainstream more recently when people discovered they could download ChatGPT, Claude, Grok, Gemini, or other large-language-model chatbot, and use them for important everyday tasks such as “enhancing communication” (i.e., trying to convince people you are really smart and knowledgeable when in actuality you have trouble stringing together two thoughts in a sentence), or having them write your assignments and take-home exams.
Even if you are like me – not the biggest fan of AI – you get it offered to you every time you make a Google search or log in to a video meeting.
Whether or not one thinks these applications are important, there is little doubt that immense human and societal gains can be made by applying AI to fields ranging from health care to industrial efficiency. The desirability of these gains, and what we are paying for them in terms of jobs and social structures, are widely debated issues.
AI and energy
Leaving those debates to others, I am going to look at AI from my favourite viewpoint – energy.
AI is a profligate user of electricity, not only to power the chips, circuits, and servers making roughly 100 septillion (100,000,000,000,000,000,000,000,000) calculations per day, but to keep them cool so they do not melt down when tasked with writing a first-year political science essay, finding the world’s best chocolate chip cookie recipe, or applying thousands of hours of video to train robots to do simple tasks.
So why is that a problem? Many policymakers are backing “electrification” of our economy – using electricity for things including transportation, heating buildings, and industrial processes – assuming that electricity is more efficient, less wasteful, and produces fewer emissions than burning fossil fuels.
But AI is not a substitution – instead, it adds a whole new class of electricity consumption to what we already consume. Instead of firing up neurons powered by a bit of coffee and sugar to write an essay, or employing people to move boxes in a warehouse, we are tasking a bunch of high-powered central processing units (CPUs) with performing electricity-gobbling calculations to get those jobs done.
Where does the AI magic and massive electricity consumption happen? At AI data centres, which are described on CBC News as, “much larger and more energy-intensive than traditional data centres.” Said Anne Pasek, Canada Research Chair in Media, Culture and the Environment at Trent University: “For an older-style conventional cloud facility, we might have a couple of megawatts (MW) moving into that building, but for AI facilities, it’s much more common to see hundreds of megawatts.” Why are there suddenly so many data centres planned in Canada? Your questions answered
And that is hundreds of MWs being consumed all the time, not just at peak hours. How much electricity is that? The entire Canadian province of Alberta runs most of the time on 9,000-11,000 MW – so one new 100 MW data centre represents a major additional industrial load (Current Supply Demand Report).
But the issue is bigger than just one new data centre – the same CBC article reports that there are only five data centres with >50 MW capacity in Canada today, but 116 new centres, each with capacity of 100 MW or greater, proposed or under construction.
In Alberta alone, total capacity of announced or under-construction data centres is more than 12,500 MW – which if built would essentially double provincial electricity demand. Not all will be built, of course – but Canada (and many other western countries) faces rapidly escalating demand for electricity after many years of static (pun intended) or even falling demand.
Think about that. For the past 20 years, policymakers have expended massive efforts and dollars on building or incenting solar and wind electricity generation, and battery-electric storage systems in order to reduce greenhouse gas emissions. Yet consumption of oil, gas, and coal – and their attendant emissions – continues to rise with increasing global population and living standards. Now, even as we see the limits to low-emissions generation growth looming – critical mineral supply chains, building in areas with poorer solar and wind resources, challenges in grid balancing and transmission tie-ins – AI is ramping up electricity demand.
How is all that new electricity going to be generated? Who is going to pay for it? How will increased demand affect consumer prices? What are the implications for emissions?
All good questions with no simple answers. Some jurisdictions have encouraged data centre builders to bring their investments, good-paying jobs, and local economic support, but are facing pushback from voters concerned about electricity prices and environmental issues. While some concerns – land use, noise, water consumption – are common to many industrial developments, electricity supply and pricing are bigger and more difficult to address.
One response is to tell developers to “bring their own power” – that is, reduce their demands on the grid by generating their own electricity (also called behind-the-meter generation). Lots of ideas have been floated – in gas-rich regions such as Alberta and Texas in the southern United States, developers propose building gas-fired generators. Big Data firms including Meta and Google have signed deals to take over or re-start nuclear power plants. I am in conversation with a firm looking to power a new data centre with geothermally generated electricity.
Whatever the individual solutions, it is safe to say that much more electrical generation is going to be built in the near future. Given that global emissions rose continuously before the AI boom, they are certain to rise more quickly with faster generation buildout.
The future for AI and energy
Summing up, several competing factors are in play:
-
- AI growth promises to be broad-based and explosive
- New data centres will demand huge new supplies of electricity
- Existing electrical grids are already challenged to incorporate more diverse and widely distributed generation sources
- Electricity prices are rising already, adding cost burdens to consumers and industry
- Voters in many jurisdictions are doing the NIMBY (not in my backyard) thing and fighting against data centre proposals in their neighbourhoods
- Emissions reduction remains a common goal, but adding a lot more electrical supply quickly will inevitably increase power-generation emissions
How will this play out? We can forecast (guess), but we will end up where people’s preferences, market forces, and innumerable random events take us. My takes, for what they are worth:
-
- Energy security – affordable, available and abundant – trumps all and will continue to trump all. Data centres will not be built where people are sufficiently worried about their electricity bill.
- Emissions from electricity generation will rise more quickly, as gas-fired and even coal-fired generation will be stepped up to meet accelerating demand. Timely availability of new gas turbines will play a role.
- Organizations promoting climate-crisis thinking will ramp up the volume as emissions growth accelerates.
- Innovation will accelerate to address some of the problems. AI computing will become more efficient. Capture and application of waste heat will reduce the overall energy footprint of data centres. Local solutions will evolve, such as locating data centres in cold places or adjacent to carbon storage capacity.
- Eventually, we will see dynamic balances between all of these factors that will determine how much AI capacity actually gets built, where, and at what cost. Lots of change? Yes. A “crisis”? Unlikely.
Finally, I think John Rapley’s opinion piece in the Globe and Mail is worth a read (Before AI can kill us the bubble will burst). He does not think we will see the rate of AI growth envisioned by many because “the funding model of the AI boom now resembles an elaborate Ponzi scheme, and it could already be approaching its limits.” In other words, there is just not enough money and other resources to build and power massive, globe-spanning AI. That business/finance point of view ties directly to the energy-demand story.
I will close with another Rapley quote that is worth careful consideration, again thinking about energy costs:
“In the best case, AI runs out of steam; in the worst, it crashes. Neither scenario will end its advance, but after the slowdown AI will enter a new, less alarmist phase, as firms pivot away from pursuing ASI (artificial superintelligence) to monetizing their existing technology – producing practical applications and cheap, efficient models.”
In other words, a new, dynamic balance.
(Brad Hayes – BIG Media Ltd., 2026)