Brian Belski's Case Against the AI Bubble Rests on One Word: Early
In the second quarter of 2026, Canadian business spending on computers and peripherals jumped 16.7%, and almost nobody was celebrating. Statistics Canada reported that overall machinery and equipment investment hit its highest level since the second quarter of 2024. That's companies buying real hardware with real money. It doesn't look like a market drunk on a story.
The common worry is that AI stocks have climbed too far, too fast, and that a crash like 2000 is waiting. That worry deserves respect. Prices are high, and a handful of companies carry a big share of the market. But BMO Capital Markets strategist Brian Belski argues that AI won't be the cause of a bubble, and his reasoning comes down to timing. We're early. If he's right, the better question for anyone with money in the market, or a down payment riding on it, is whether the profits behind the price are real.
Bubbles have a recipe, and some ingredients are missing
Look back at the dot-com peak. Many of the hottest companies had little or no revenue. Ordinary investors were piling in, and confidence felt unlimited. Prices floated free of any earnings that could hold them up.
Today's AI leaders look different. The so-called Magnificent Seven, the giant U.S. tech firms that drove much of the S&P 500's gains back in 2024 and into 2025, are among the most profitable companies ever built. Belski's view is that their weight in the market has come from earnings and cash flow, not just excitement. And plenty of investors remain openly skeptical. Constant talk of a bubble works against the euphoria a bubble needs.
The lesson: when you size up a hot sector, ask where the money comes from. If a company earns real cash from what it sells, high prices have something underneath them.
Early means the builders get paid first
Belski compares this stage to the internet's build-out in the mid-1990s, years before the 1999 peak. Right now most of the money goes into chips, servers and data centres. The businesses that will use AI every day are mostly still getting ready.
Think of a new LRT line in Edmonton. Nearby properties often rise before the first train runs, because buyers are paying for what the line will do later. AI stocks are doing something similar. The companies laying track profit now. The passengers, including factories and logistics firms, are expected to see productivity gains over time.
Edmonton sees this closer than most. The Alberta Machine Intelligence Institute (Amii) here is one of Canada's three national AI institutes, and the city has drawn real investment for its work in machine learning.
Expensive and early can both be true
Here the bubble crowd has a fair point. A market can avoid a bubble and still fall hard. Tech stocks are pricey, and a correction, meaning a temporary drop, can happen any time. Adoption may also move slower than markets hope. A local business might take years to actually save money with AI.
Concentration adds risk too. If your index fund leans heavily on five or six companies, one bad earnings season can hit your whole portfolio.
This matters if you plan to sell investments for a home. Many Alberta buyers fund down payments with portfolio gains. Treat those gains as yours only once you've sold. If you're buying in the next year, move what you need into something stable before you write an offer, rather than betting the deadline on a market that can dip.
A better test than "it went up a lot"
Selling everything because prices feel high is a blunt tool. A sharper approach is to keep watching free cash flow, the money a company has left after paying for its operations and investments. As long as AI companies keep earning real cash, the market rests on something solid. If spending keeps climbing while cash flow stalls, that's the signal worth acting on.
Use the same rule you'd use with property. You wouldn't own only one type of building in one neighbourhood. Hold AI the same way, as one part of a mix, so being early works in your favour and being wrong doesn't sink you.
In the second quarter of 2026, Canadian business spending on computers and peripherals jumped 16.7%, and almost nobody was celebrating. Statistics Canada reported that overall machinery and equipment investment hit its highest level since the second quarter of 2024. That's companies buying real hardware with real money. It doesn't look like a market drunk on a story.
The common worry is that AI stocks have climbed too far, too fast, and that a crash like 2000 is waiting. That worry deserves respect. Prices are high, and a handful of companies carry a big share of the market. But BMO Capital Markets strategist Brian Belski argues that AI won't be the cause of a bubble, and his reasoning comes down to timing. We're early. If he's right, the better question for anyone with money in the market, or a down payment riding on it, is whether the profits behind the price are real.
Bubbles have a recipe, and some ingredients are missing
Look back at the dot-com peak. Many of the hottest companies had little or no revenue. Ordinary investors were piling in, and confidence felt unlimited. Prices floated free of any earnings that could hold them up.
Today's AI leaders look different. The so-called Magnificent Seven, the giant U.S. tech firms that drove much of the S&P 500's gains back in 2024 and into 2025, are among the most profitable companies ever built. Belski's view is that their weight in the market has come from earnings and cash flow, not just excitement. And plenty of investors remain openly skeptical. Constant talk of a bubble works against the euphoria a bubble needs.
The lesson: when you size up a hot sector, ask where the money comes from. If a company earns real cash from what it sells, high prices have something underneath them.
Early means the builders get paid first
Belski compares this stage to the internet's build-out in the mid-1990s, years before the 1999 peak. Right now most of the money goes into chips, servers and data centres. The businesses that will use AI every day are mostly still getting ready.
Think of a new LRT line in Edmonton. Nearby properties often rise before the first train runs, because buyers are paying for what the line will do later. AI stocks are doing something similar. The companies laying track profit now. The passengers, including factories and logistics firms, are expected to see productivity gains over time.
Edmonton sees this closer than most. The Alberta Machine Intelligence Institute (Amii) here is one of Canada's three national AI institutes, and the city has drawn real investment for its work in machine learning.
Expensive and early can both be true
Here the bubble crowd has a fair point. A market can avoid a bubble and still fall hard. Tech stocks are pricey, and a correction, meaning a temporary drop, can happen any time. Adoption may also move slower than markets hope. A local business might take years to actually save money with AI.
Concentration adds risk too. If your index fund leans heavily on five or six companies, one bad earnings season can hit your whole portfolio.
This matters if you plan to sell investments for a home. Many Alberta buyers fund down payments with portfolio gains. Treat those gains as yours only once you've sold. If you're buying in the next year, move what you need into something stable before you write an offer, rather than betting the deadline on a market that can dip.
A better test than "it went up a lot"
Selling everything because prices feel high is a blunt tool. A sharper approach is to keep watching free cash flow, the money a company has left after paying for its operations and investments. As long as AI companies keep earning real cash, the market rests on something solid. If spending keeps climbing while cash flow stalls, that's the signal worth acting on.
Use the same rule you'd use with property. You wouldn't own only one type of building in one neighbourhood. Hold AI the same way, as one part of a mix, so being early works in your favour and being wrong doesn't sink you.
Sources
Read Next
Five Money Checks Canadian Snowbirds Should Finish Before Heading South
Canadian Bank Security Assumes a Human Is Logging In. AI Agents Break That
Before Storm Winds Arrive, Turn 5 Yard Hazards Into Protective Buffers
A 25 Basis Point Hike Doesn't Erase Variable's Long-Run Edge Over Fixed