MILAN CACIC
October 02, 2026
Money Financial literacy Economy Commentary In the news News Trending Weekly update Weekly commentaryAI’S NEXT BOTTLENECK ISN’T CHIPS, IT’S POWER
For the past few years, the AI race has largely been about chips. Who could get enough GPUs, who could build the biggest data centres, and who could spend the most money. But, increasingly, the most important question is becoming much simpler: where are we going to get all the electricity?
The numbers are enormous. The international Energy Agency expects global energy consumption from data centres to roughly double between 2025 and 20301. In the US, data centres could consume around 12% of all electricity by the end of the decade. After years of relatively stagnant electricity demand, AI is helping turn power into a growth industry again.

Source: IEA (2025), Energy and AI, IEA, Paris https://www.iea.org/reports/energy-and-ai, Licence: CC BY 4.0
And we are already starting to see the bottleneck. This week, regulators highlighted a projected 6,800-megawatt power shortfall in the PJM electricity market, which serves more than 65 million Americans. Some data centre developers are now installing their own natural gas turbines rather than waiting years for a grid connection. In other words, having the chips and the money isn’t much use if you can’t plug the building in.
This changes how we think about the AI investment story. The obvious winners have been semiconductor companies, but the next beneficiaries may increasingly be companies involved in power generation, natural gas, nuclear energy, electrical equipment, transmission, cooling, and the grid itself. AI is beginning to look as much like an industrial and energy build-out
as it is a technology boom.
It also creates an interesting opportunity for Canada. We have abundant natural gas, uranium, hydroelectricity, and other energy resources. Exactly the things that an increasingly power-hungry world needs. The opportunity is obvious. As we have written before, our challenge is generally not figuring out what we have; it’s figuring out how to actually build
something with it.
The AI boom isn’t slowing because we’ve run out of ideas. Increasingly, the constraint is much more old-fashioned.
The world’s most advanced technology may ultimately depend on something remarkably simple: finding somewhere to plug it in.
I have also included a piece from our CIBC Economics Team entitled “Earnings and economic growth: Mind the gap”.
As always, if you have any questions, please feel free to give us a call at any time.
Have a great weekend.
Milan


