Have the Latest ‘AI Trades’ Jumped the Shark?

Here is something everyone forgets about bear markets: They start gradually, little noticed, sliding down a slope of hope as headlines preach “buy the dip.” Their worst declines come late, when panic selling reigns. And they generally end when panic jumps the shark, creating outlandishly bad expectations.

With all this in mind, let us talk about Software.

If you have tuned into financial news at all, you have probably seen headlines about AI tanking Software stocks after a newfangled chatbot appeared capable of automating various tasks some Software companies built whole business models around. What you might not have seen: Software is actually down since late last October, with gradual early declines snowballing into the recent, steeper drop. At the low to date, last Thursday, the MSCI World Software & Services Industry Group was down -28.4% from its October 28, 2025 high.^[i] If you want to call that an industry bear market we can see your logic, given bear markets generally exceed -20%. Though given we are also talking about a decline lasting less than four months thus far and largely confined to specific industries, the timing seems more correction-like, as does the sentiment-fueled nature.

But this isn’t worth getting pedantic over. A downturn is a downturn, and for those with Software exposure, we know it isn’t fun seeing the red arrows concentrated in that industry. At the same time, the industry’s woes haven’t knocked the broader global market, considering the MSCI World Index hit its latest record high on Monday. The market will always have pockets of strength and weakness among sectors, industries and countries. Hence, the first lesson: Always be diversified so that no one category makes or breaks your returns. Concentration might seem great when it works in your favor, but it can be devastating when it doesn’t.

Also, investors can’t look backward. What matters is what happens from here, and judging from the latest headlines, wherever Software sentiment was last year, it has clearly turned more dour—arguably, crazily so. You can see it in the fresh batch of AI bubble warnings, including a particularly colorful, metaphor-laden one from former Greek Finance Minister Yanis Varoufakis.^[ii] This week, AI disruption fears hit commercial real estate brokers, which apparently have no purpose once everything is automated and no one needs offices. They also hit trucking and logistics firms, as analysts fear only negative fallout on margins from price transparency and fail to fathom the possible cost benefits from new AI tools. European asset managers got whacked by fears of AI automating personal tax prep earlier this week. This mass hunt for the next domino is classic late-downturn behavior. It doesn’t matter than virtually none of these AI tools are likely to jar the profit or sales outlook for the industries for years. As illogical selling gets ramped up, it gets more illogical as it goes.

It is also just really, really weird in context of broader AI bubble fears—a classic case of investors trying to have it both ways, which is a sign those bubble fears were overstretched. For over two years, headlines have warned the giant Tech stocks investing in AI’s infrastructure (data centers, graphics chips and the like) are far out over their skis. We are told this is a bubble, that investors are throwing good money after bad, and that it will never pay off because businesses aren’t adopting AI tools anywhere near to the degree to justify this mass buildout. One company that shares its name with a very large river got hammered last week when it announced a big data center investment. So we ask: How can all these core AI infrastructure companies be overstretched if AI is going to be adopted to the degree that a vast array of services businesses cease to exist as we know it? Or if all knowledge workers are displaced, put on the dole and only electricians, cobblers and home health aides find work? Or if it is indeed a tsunami set to be a “general substitute for cognitive work?”^[iii]

Make it make sense.

Both theses can’t be true, which is a good indication the truth is probably boring and in the middle. Advances in machine learning are creating useful tools. Those tools need heavy backend infrastructure to run. Creative people will find ways to use these tools to make their businesses run better, and it will be people using those tools. Automation may make some jobs obsolete while creating new ones, just as it has since the flying shuttle and spinning jenny. Anyone saying “it is different this time” is ignoring centuries of history.

So no, we don’t think this is the end of traditional Software, Software as a Service or other industries employing or competing with AI models. These last couple weeks seem like a classic sentiment freak-out. Impossible to predict and time with precision at both ends, so we strongly suggest not trying. Instead, grit your teeth, remember to be and remain diversified, take a deep breath, focus on your goals and the broad market’s ability to get you there, and try to have a laugh at the madness. We find it very therapeutic.