I know. You looked at that title and said “Well! It will be interesting to see how he ties all that together.” I’m sorry to disappoint. It didn’t really tie together very well. It’s turned out to be something more of a rambling chain-of-thought exercise.

It all started when my friend Malcolm approached me in the hockey dressing room and asked me if I was aware of just how much energy was going to be required to drive AI. Well, the truth is that I had a vague awareness that AI computer centres were energy hogs, but I had no accurate idea of the magnitude. But Google found lots of articles for me, which sent me scurrying down a number of rabbit-holes.

First of all, yes, most forecasters agree that AI is going to drive a significant demand for electrical energy. Just to give you a feeling for how big it might be:

• Torys LLP, a Canadian law firm, expects global AI market to grow from ~ $200B US to $2 Trillion US between 2023 and 2030.
• According to Forbes Magazine, AI power demand is projected to grow from 8 TWh in 2024 to 652 TWh by 2030.
• The Electric Power Research Institute forecasts that data centers may see their electricity consumption more than double by 2030, reaching 9% of total electricity demand in the US.
• Data centres used to have power requirements of 100-200 megawatts (MW), but big ones now require 300-500 MW. (That means that a single data centre might consume almost the entire output of a Pickering nuclear unit.)

It’s interesting that one of my most trusted information sources, the Economist, disagrees with those robust projections, or at least with the timing of them. The Economist observes that the financial markets are pumping money into tech firms because they’ve bought into the AI explosion predictions. But the Economist notes that – despite the hype – outside the tech industry there is not yet a great deal of movement to incorporate AI into businesses. The Economist tells us that:

• As of Nov 2024, “AI has had almost no impact on America’s economy, with unemployment still very low and productivity growth weak. Expect this puzzle to continue in 2025.”
• America’s Census Bureau asks businesses whether they use AI to produce goods and services. Only 5-6% answer in the affirmative.
• In Canada, 6% of all firms used AI to produce goods or delivering services in the past year, according to official data.
• Capital spending across the rich world remains fairly weak, suggesting that businesses are not investing in the tools that would allow AI to give them a big productivity boost.

So right now, the tech industry is investing heavily in AI, but almost nobody is buying what big tech is selling.

Nevertheless, I think we should take it as a given that AI is going to place extraordinary demands on the electrical generation and distribution industries in the longer term. AI is like the Field of Dreams – if you build it, they will come. So, what impacts will that demand growth place on us?

Right now, the adoption of AI is resulting in an increase in electricity use rather than a decrease. But various factors might affect the magnitude of that increase. And the computer industry is adapting to deal with the changes, so the impacts might actually be lowered.

• The whole reason for the projected increase in the use of AI is that the tech industry is betting that it will make us faster, smarter, and more efficient. It is reasonable to believe that those efficiencies will result in energy savings in the production of commodities. As noted above, few industries have jumped on the AI bandwagon so we don’t know yet whether those energy savings will be realized.
• Older data centres are mostly air-cooled. As computer processing energy demands increase, data centres are beginning to use water cooling to reduce the energy spent on cooling – and increase the demands on water supplies.
• Next generation computer chips are consuming more energy than ever, but the information processing ability is growing faster than the energy demand, so chips and computers are becoming more energy efficient.
• At the same time however, as computers get more energy efficient, the growing use of Large Language Models (AI systems) places greater demand on computers. On average, a ChatGPT query needs nearly 10 times as much electricity to process as a Google search. The net result is that although computers are getting better, they aren’t improving fast enough to meet the information demand without a significant energy increment.
• The energy demands, and cooling requirements of new data centres are forcing a relocation of such centres away from metropolitan areas. In the United States, per Barclays, an investment bank, “Data centre developers are prioritising land with access to untapped power sources, water, workforces, and favourable regulation.”
• Data Centre developers are also looking at “behind the meter” generation – siting their data centre where they can buy most or all of the output of generating station. “Energy generation and storage facilities can be co-located with data centres to provide reliable long-term sustainable power while also lessening the need to expand transmission capacity on the grid.”
• Large, energy-intensive data centres cooled with water are opening up the possibility of another industry, which is to use data centre waste heat for community heating systems. The Canadian Energy Regulator tells us that “in Levis, Quebec, nearly 100 MW of energy from waste heat is expected to be redirected to households by the end of 2024. Liquid cooling systems capture heat generated by data centers and can raise water temperatures from 30-35°C to 70-80°C, which can then be used in district heating networks.

I think all of the above tells us that data centre energy demand will grow significantly, but I think you can put fairly large error bands around any forecast you might read.

It was about here that I got distracted and dived down the first rabbit hole. I noticed that although AI is projected to be a rapidly growing component of computer data centre needs, it isn’t alone. I was astounded to find IEA graphs that showed that in 2022, about 20% of global data centre energy demand, approximately 100TWh, was expended in mining cryptocurrencies. Those graphs show that cryptocurrency mining will grow from 100 TWh to 150 TWh by 2026 and will still be taking more energy than AI at that time.

I don’t know cryptocurrency from horseshit, so I decided to try to improve my understanding by reading a few articles. One of them was described as a tutorial for beginners and I read some of that and I still know diddley-squat about cryptocurrency.

I was really trying to discover how cryptocurrency mining provides value. What justifies the consumption of 150 TWh of electricity? For perspective, 150 TWh is close to 25% of Canada’s electrical energy consumption in 2022. That’s an awful lot of energy, so it has to be producing something of value. Doesn’t it?

The answer seems to be that computers are solving difficult and complex problems that somehow improve security of cryptocurrency transactions. I do not have a good answer about how it does that, nor do I understand how the complex problems are originated or by whom. And I wound up asking myself whether cryptocurrency miners understand who they’re really working for when they set their computers to solving problems. Would miners be working for the good guys, or the bad guys?

That led me to inquiring whether cryptocurrency was the favoured vehicle for illegal transactions on the web. The UK’s Finance department estimates that in 2023, money laundering using cryptocurrencies to make money disappear and re-appear somewhere else amounted to $22B. Crypto is also being used to get around government-imposed sanctions. A current estimate by a site called Chain Analysis shows a total of $24.2 B of cryptocurrencies going to “illicit addresses” in 2023, at least half of which are to “sanctioned entities”. Hello Mr. Putin. How’s business?

The fact that cryptocurrencies are a big factor in illegal commerce didn’t surprise me at all, but it was interesting to get an idea of the scale of these enterprises.

What did surprise me, since I’m just scurrying down a rabbit hole anyway, was an article which suggested that devices connected to the internet can be captured and turned into slaved mining devices. Mostly you wouldn’t notice. Your iPhone might be running slow and your electrical bills might be a bit high but that’s all. A CBC article suggests even kitchen ranges or fridges, (if you connect them to the grid because you want a smart home, might become mining devices). In 2024 a cybersecurity firm called SonicWall reported that they had detected 332 million “cryptojacking” attacks in the first half of 2023. Who knew? Certainly not I. Nor do I care – I have a stupid home and I plan to keep it that way.

I have no conclusion to offer, and no advice to give. Don’t worry, there is no check-out. This is, after all, just a rabbit hole.

Let’s return to more serious issues about how we incorporate AI into Canada’s energy future. My first thought was to understand the scope and assumptions built into Canadian energy demand forecasts, because AI is by no means the only upward pressure on the Canadian grid.

Reports from the International Energy Agency and the Canadian Climate Institute in 2022 found Canada would need to double or triple its electrical grid capacity by 2050 to reach its net-zero goal. The Canadian government has committed to aligning Canada’s electricity system with the country’s climate goals. To that end, they have implemented Clean Energy Regulations which are backed by a commitment of more than $40 billion over the next decade to support Canada’s clean electricity sector through tax measures, public financing and grant contributions.

So, it seems that we are starting to plan for conversion of our energy requirements to a clean grid. But, I wasn’t able to find a good summary of the assumptions that have gone into those predictions.

To what extent will electrification (to a clean grid – otherwise it’s pointless) change the home heating profile of Canada? We’re told that converting Toronto’s 650,000 natural gas-heated homes to best-in-class heat pumps would double the city’s electricity demand – an increase of about 5,000 megawatts. Do the grid projections assume that most home heating will convert to heat pumps? If the GTA is about 10% of Canada’s population, we’d need something like 50,000 incremental MW to convert all those homes to heat pumps.

If we’re making assumptions about electrifying home heating, are we assuming great uptake in government programs to dramatically improve insulation values in most buildings? By how much do we assume we can reduce total heating and cooling demand?

A 2020 study commissioned by Natural Resources Canada found that based on federal electric vehicle targets, EVs will consume 156.5 terawatt-hours of electricity per year by 2050 — equivalent to 22.6 per cent of electricity consumed in the country in 2020.

Do projections for Canada’s energy needs in 2050 include a big demand for AI (and crypto) or did they only consider EV’s?

While the demand for electricity is going to be a problem it also presents some interesting opportunities for Canada. One opportunity is to attract high tech firms to Canada to preferentially locate data centres here. In October, a Canada Energy Regulator report has identified that “Multiple factors make Canada an attractive destination for data centers:

1. Relatively low electricity prices in some regions
2. Significant renewable and clean electricity resources
3. A relatively cool climate”
   That CER report drew from an Energy Monitor report which ranked Canada as the world’s best choice for Data Centre location.

We’ve already mentioned the opportunity to use “waste” heat from data centre cooling to provide community heating, so attracting data centres looks like an intriguing solution.

Another opportunity is to export electricity. Canada has traditionally been a net exporter of electricity to the United States, to the tune of almost $3B per year. Recently that position has been significantly reduced owing to two factors. First, drought conditions in 2023 reduced the availability of excess hydro-electric generation to be marketed, and second the growth in US fracking-generated natural gas has reduced the cost of natural gas generated electricity. Those conditions need not be considered irreversible. Let’s face it, if Canada is going to need grid expansion and more generation, the US will need it even more. So let’s generate lots and sell it!

Canada’s largest electricity grids are connected to each other and to US grids, with more than 35 Canada-US transmission line interties, across all provinces bordering the US. Most of the Canada/US trade in electricity is driven more by the need to stabilize the interconnected grids than by pure price competition. Nevertheless, if we invest heavily in the increased generation that we need anyway, and in a more inter-connected and secure national electricity grid, there is the prospect of being able to sell electricity to a power-hungry neighbour. (No, that wasn’t a reference to Trump).

While I was musing about whether, and how well, we’re planning for the electrification of Canada, I came across “Powering Ontario’s Growth”, which is the Ontario Government’s 91 page energy plan. I have to say, I was impressed by it. It is, I think, both comprehensive and pragmatic. It seems more sensitive to the climate change issue than I would have expected from the Doug Ford government, while at the same time being a pragmatic approach that tries to use everything of value. It is also fairly bold and assertive about really tackling the problem. The plan includes:

• the largest expansion of non-emitting nuclear energy on the continent.
  o Starting the approvals process for locating up to 4,800 MW of new nuclear generation on the Bruce site.
  o four SMRs at the existing Darlington nuclear (1200 MW).
  o supporting Ontario Power Generation’s plan for refurbishing the Pickering Nuclear Generation Station.
  o leveraging Ontario’s supply chains and expertise to assist Alberta, Saskatchewan and New Brunswick to develop SMRs.
• Procurement of nearly 3,000 MW of battery energy storage, as well as natural gas and clean on-farm biogas generation capacity,
• continued use of natural gas generators as required to provide electrical peaking power. (OK, non-ideal from the greenhouse gas perspective, but pragmatic and probably the right call from the perspective of managing power variations on the grid.)
  o Over the long-term, an economically viable natural gas network can also support the integration of clean fuels to reduce emissions, including renewable natural gas (RNG) and low-carbon hydrogen.
• funding for energy-efficiency programs (total funding more than $1 billion over the current 2021–2024 framework.
• Plan to export clean energy and technology to the world.
• Ontario’s plan will never include a carbon tax. (You bonehead, Ford!) Instead, Ontario will meet its emission reduction targets with affordable and clean power that supports families and businesses to move away from higher emitting sources of energy.
• Transmission projects between London, Windsor and Sarnia represent an investment of more than $1 billion and are proposed to be developed in phases through 2030.
• Three transmission projects that will support the production of clean steel at Algoma Steel in Sault Ste. Marie.

The big challenge for Canada and the US and the world, is to generate clean electricity so that huge increases in energy demand don’t drive the climate issues totally out of control. Goldman Sachs estimates that the rise of data center carbon dioxide emissions will represent a “social cost” of $125-140 billion (at present value).

One answer is for data centres to find non-emitting sources such as more nuclear power. Data centres in the US are planning to restart a nuclear power station at Three Mile Island in Pennsylvania to power several Microsoft data centres, and Google has ordered nuclear reactors from Kairos Power, a startup. Industry experts expect substantial investments by tech firms to underwrite new renewables and commercialize emerging nuclear generation capabilities.

But these non-emitting sources will be slow to deliver. Analysts expect incremental data center power consumption in the US will drive around 3.3 billion cubic feet per day of new natural gas demand by 2030, which will require new pipeline capacity to be built. One research team expects AI to be “a long tailwind for natural gas demand”. Access to natural gas pipelines is now a major preference for data centre development.

And that brings me to the next rabbit-hole.

At the end of all this digging into the electrification of North America (and Europe and the rest of the world as development requires) I began to wonder what we’re really going to do with all that energy we project to be pouring into AI? How deeply will Artificial Intelligence be integrated into industrial productivity? How much will the average citizen rely on it in his day to day life? How much will key processes like banking, air traffic control, electrical grid stability control, and National Defence depend on AI and data centres? And finally, what happens 20 years from now if we lose the electrical grid and/or the internet?

I’m mostly in favour of electrification with a clean grid (more nukes!) as the way to beat the climate change problem. However, the problem with getting rid of the good old-fashioned, dependable, petroleum driven internal combustion engine is the “dependable” part. Large-scale electrification seems to imply a loss of diversity and that implies, to me at least, an increased vulnerability.

It’s easy to imagine a loss of electrical grid bringing down the internet and all of the processes driven by internet and AI. It’s also easy to imagine an internet failure bringing down the electrical grid. And there are lots of people out there attacking the internet.

Statista, a database site, says that there were over 3000 successful cyber attacks in the United States last year, affecting more than 300 million people. Cybercrime costs in Canada in 2023 are estimated at $3.8 Billion (US dollars). And the Canadian Centre for Cyber Security asserts that in addition to the cyber crime threats, “State-sponsored cyber operations against Canada and our allies almost certainly extend beyond espionage…We assess that our adversaries very likely consider civilian critical infrastructure to be a legitimate target for cyber sabotage in the event of a military conflict.”

So, there it is. We are investing, or expecting to invest, a great deal of money and energy in expanded data centres and greatly increased use of artificial intelligence. Concurrent with that, we’re going to commit nationally to a greater dependence on electricity. And in so doing, we’re going to lay ourselves open to sophisticated computer attacks on our information systems. At the same time, we’ll be getting more vulnerable to conventional destructive physical attacks on an electrical grid upon which we’ve come to rely heavily. I have no solutions to offer – consider this just a little something for you to worry about. You’re welcome. Have a nice day.

Sorry Malcolm. I started to write an article about AI, but it turned into an article on electrification and climate change. It turns out Mick Jagger was right – you can’t always get what you want.