Market sentiment has shifted from “we’re in an AI bubble” to “I really wish we were in an AI bubble”.

Chris Trottier

What people actually miss is the pre-AI world. Markets felt slower. More predictable. Valuations moved in familiar ranges. You could pretend you understood the cycle. That stability was mostly an illusion, but it felt real enough that people now want it back.

Vancouver real estate is the cleanest example I have, and I live here, so I watched it in real time.

Through the 2010s and early 2020s, Vancouver became one of the most expensive housing markets in Canada. In March 2024, the average home price in Vancouver was around $1.18M, the highest in the country.

By October 2025, the benchmark home price in Metro Vancouver was about $1.13M. That is only about 10% below the all-time high of roughly $1.26M. That is a softening, not a crash.

The whole time, people kept calling it a bubble. For more than a decade, locals said “this has to crash.” They wanted a reset to some earlier, more “normal” price level. Talk of a bubble did not pop anything.

What actually changed the market was policy. In 2016, British Columbia added a 15% property transfer tax on foreign buyers of residential property in Metro Vancouver.

The city then layered on an Empty Homes Tax. It started at 1% of assessed value in 2017 and is now 3% from 2021 onward.

On top of that, the province added a Speculation and Vacancy Tax starting in 2018, aimed at under-used homes in key urban markets.

Academic work on the foreign buyers tax found that targeted neighbourhoods dropped about 6% more in price than less-exposed areas after the tax. That is a material hit, but it is not a collapse.

Put differently. You had one of the most expensive housing markets in the country. You had years of people yelling “bubble.” You had multiple layers of taxes aimed directly at speculation and under-used properties. After all of that, prices dipped, then plateaued near record territory. They did not revert to any comforting “old normal.”

That is the part people miss about sentiment. Constant caution does not necessarily kill a bubble. It can help it grind on.

When everyone expects a crash, they hedge. Sellers hold. Buyers wait for deals. Governments step in to “cool” things without triggering a full-on panic. You end up with a long, expensive plateau instead of a dramatic blow-off top followed by a clean reset. The anxiety stretches the cycle instead of ending it.

Now apply that thinking to AI.

Right now, the big AI-levered names are expensive, but the data does not look like peak dot-com. During the 2000 tech bubble, the four leading tech stocks of that era traded near 70 times 2-year forward earnings at the top.

Today, the major AI datacenter spenders like Microsoft, Alphabet, Amazon and Meta trade at an average 2-year forward P/E around 26. That is rich, but it is nowhere near 70.

If you zoom in on the so-called Magnificent 7, you see the same pattern. Recent data for 2025 forward P/E has names like Alphabet around 20, Meta around 28, Apple around 30, Nvidia around 24 and Microsoft around 38. Tesla is the outlier at roughly 180.

So yes, valuations are stretched. The broader S&P 500 sits above its long-run forward P/E average, and tech is even pricier. Recent work from RBC, for example, puts the S&P 500 forward P/E in the low 20s versus a long-term norm in the high teens.

But this is not an environment where the whole AI complex is trading at triple-digit forward multiples as a rule. It is a mix of high multiples, very high capex, and real earnings growth, wrapped in a big argument about how durable those earnings will be.

On top of that, sentiment is noisy and often skeptical. You have major investors and central banks warning about AI over-investment. You have pieces arguing that AI valuations are “not cheap but far from dot-com extremes.”

You also see direct evidence of hesitation on the demand side. For example, recent reporting on Microsoft highlighted customer resistance to some AI add-ons and weaker than expected uptake in certain enterprise products.

That is not the mood of a classic mania. That is a market that is nervous even as it spends.

But you see, that is the trap.

Speculation runs on sentiment. To get a true bubble, you need more than high prices. You need people to believe. Robert Shiller’s work on “irrational exuberance” and narrative economics makes this point over and over. Bubbles form when optimistic stories about a new era spread faster than doubt, and when those stories feed back into price rises, which then reinforce the story.

Right now, the AI story is powerful, but it is not pure. The optimistic narrative is constantly interrupted by warnings, regulation talk and “this is a bubble” discourse. That caution tempers the multiple. It keeps valuations high, but below the full escape velocity you saw in 1999 or in the hottest phases of some housing markets.

If you ever get to a world where AI-exposed names routinely trade at forward P/E above 100 across the sector, and where the dominant story is “this can only go up,” that is when you are in obvious bubble territory. We are not there yet. The numbers and the mood both say so.

Until then, what you have is exactly what it feels like. Ambition with a bit of helium. Not a full-blown dot-com rerun.

Wandering Adventure Party

Market sentiment has shifted from “we’re in an AI bubble” to “I really wish we were in an AI bubble”.