On the Radar: AI is driving up chip prices, power bills, and bond yields. Are mortgage rates next?

  • First National Financial LP

Quick Takes:

  1. U.S. AI-related investment is running at a roughly $1.5 trillion annual pace, with Microsoft, Meta, Alphabet, and Amazon alone spending on the order of $700 billion a year, and some estimates credit the buildout with about one third of recent U.S. growth.
  2. AI costs are reaching consumers, with memory chips up as much as 400% since 2024, U.S. electricity prices up 5.9% in a year, and Canadian computer equipment prices rising for the first time since 2020.
  3. Canada is building its own boom, with Alberta fielding 16 gigawatts of data centre applications against a 12-gigawatt peak load, wholesale power projected to more than double by 2028, and Ottawa and Telus adding three sovereign AI data centres in B.C.
  4. The Fed held at 3.50% to 3.75% on July 29 with three dissents in favour of a hike, Macklem warned stretched AI valuations could bring a “painful correction,” and markets price a 65% chance the Bank of Canada holds on September 2 with a 35% chance of a hike and none of a cut.

Oil finally loosened its grip this week, and the rate conversation barely noticed. On Monday, President Trump paused strikes on Iran and Brent crude fell about 5 percent into the low $80s, the kind of news that once would have revived rate cut hopes. Cut bets stayed scarce, because a second inflation force has been growing underneath the oil story all year.

That force is artificial intelligence. AI spending in the United States is now running at a pace of roughly $1.5 trillion a year, central bankers on both sides of the border have begun naming it in speeches and statements, and the costs are starting to show up in the prices Canadians pay. So this issue asks the question directly: is AI having an effect on Canadian mortgage rates?

The Boom in Numbers

Start with the scale. U.S. business investment in software, data centres, and computer and communications equipment is running at roughly $1.5 trillion a year, up from about $1 trillion two years ago, while some estimates credit the buildout with roughly one third of recent U.S. economic growth. Numbers that size stop being a tech story and become a macro story.

Last week’s earnings confirmed the pace is still rising. Microsoft, Meta, Alphabet, and Amazon are collectively spending on the order of $700 billion a year on AI infrastructure, with Alphabet raising its 2026 guidance to as much as $205 billion. Because the Bank of Canada pegged top U.S. tech firms’ AI investment at about $200 billion in 2024 and $400 billion in 2025, the trajectory looks like a doubling every year.

Monday’s construction data showed how lopsided the boom has become. Data centre construction reached a $68.3 billion annual rate in June, up $21.5 billion from a year earlier, while every other category of private construction combined fell by $101.6 billion. Houses, shopping centres, and hospitals are giving way to server halls.

Even a slowing economy leans on the boom now. The U.S. grew at a modest 1.5 percent annual rate in the second quarter, and AI-related investment was one of the few engines still pulling. Imports of computer equipment and chips from Taiwan tell the supply side, hitting roughly $90 billion in the first five months of 2026 against about $20 billion two years earlier.

How AI Raises the Prices You Pay

The most direct channel is hardware. Memory chip costs have surged by as much as 400 percent since 2024 as data centres soak up supply, and device makers have passed the increases along to laptops, tablets, and game consoles this year. Canada imports that hardware, so the hikes land directly in Canadian CPI.

Statistics Canada has already recorded the arrival. Computer equipment and software prices rose 3.9 percent in the May CPI report, the first year-over-year increase since 2020, driven by the same data centre demand the Federal Reserve has cited.

Electricity is the bigger long-run channel. U.S. power prices are up 5.9 percent over the past year with data centre demand named as a main driver, and analysts expect roughly 6 percent annual increases through 2027 as the grid strains to keep up. When power gets more expensive, everything made and stored with it follows.

Canada Is Building Its Own Boom

Alberta is the clearest example. The province has received data centre applications totalling more than 16 gigawatts of power demand, even though its peak electricity load is only about 12 gigawatts. The queue includes a Meta campus that could scale to 1.8 gigawatts, and power producers project wholesale electricity prices could more than double by 2028.

British Columbia joined this spring. Ottawa and Telus announced three sovereign AI data centres, one in Kamloops and two in Vancouver, with roughly $9 billion in projected economic activity attached. Sovereign computing is the point, because Canadian governments want AI workloads running on Canadian soil.

Adoption is climbing at the same time. The Bank of Canada says the share of Canadian businesses using AI has roughly quadrupled since 2022, reaching about 12 percent by 2025, while most adopters report no change in staffing. So the Canadian AI story is no longer hypothetical, and its costs and benefits are both arriving.

What This Means for Canadian Mortgage Rates

The Federal Reserve’s July 29 decision showed how AI has crept into the rate debate. Officials held the 3.50 to 3.75 percent range, but three of them dissented in favour of a hike, and the statement called productivity and capital investment robust. New York Fed President John Williams has warned that AI investment demand could prove persistent enough to require a policy response.

Fixed mortgage rates feel that pressure first. Canadian lenders price fixed mortgages off GoC bond yields, which take their floor from U.S. Treasuries, and a Fed facing AI-driven investment demand is in no hurry to cut. Even with oil easing, that keeps the floor under Canadian fixed rates in place.

Variable rates answer to the Bank of Canada, and the Bank is watching AI with two minds. Deputy Governor Michelle Alexopoulos argues AI could eventually lift productivity and ease price pressures, while Governor Macklem warned in Paris that capital flooding into U.S. AI is stretching valuations and “setting the stage for a painful correction.” Rate markets still lean toward a September 2 hold, but they price a 35 percent chance of a hike and no chance of a cut.

What Could Change the Picture

An AI-driven correction is the sharpest tail risk. If stretched valuations snap back, Treasury yields and, by extension, Canadian fixed mortgage rates would likely fall, but for unwelcome reasons. As Bank of Canada Governor Tiff Macklem put it, "we could be sideswiped if financial stability risks crystallize."

The benign path runs through productivity. If AI adoption delivers the efficiency gains the Bank of Canada describes, with most Canadian users already saving an hour or more a day, inflation pressure eases and rate cuts get easier to justify. That payoff arrives in years, however, not quarters.

Oil has not left the stage either. The ceasefire talk that knocked Brent into the low $80s on Monday remains unsigned, with Iran insisting its only active talks run through Oman, and Friday morning brings jobs reports on both sides of the border. Any of those could move yields before this article is a week old.

Bottom Line

For most of 2026, the answer to why Canadian mortgage rates stayed high was one word: oil. The honest answer now has two words, because the AI boom is big enough to move growth, inflation, and bond yields on its own. So yes, AI is having an effect on Canadian mortgage rates, and the effect is to keep them higher for longer.

Borrowers should read it this way: fixed rates carry a new structural floor, while variable rates wait on a calmer world. Both of the paths to lower rates, an AI correction or a productivity payoff, run on timelines nobody controls. Neither is on the calendar for this fall.