If the film series “Terminator” is your base case view for AI’s future impact on humanity, you may want to own put options pre-“Judgement Day.” Thereafter, financial assets will be of little help.

Our own read of history is that every technological breakthrough has resulted in a larger economy. While technology inevitably brings disruptions to industries, companies and workers’ lives, the resulting productivity gains have given rise to higher standards of living and higher profits over time. (Please see our Outlook for more)

For the moment, we assume that this historical precedent will continue, with the caveat that the scope for change Artificial Intelligence is delivering is truly unprecedented, so “this time could be different” in the level of disruption ahead of us, between now and that more profitable future.

Over the past few weeks, we have dwelled on the sudden, severe selloff in software shares that occurred after Anthropic’s release of Claude Opus 4.6, which is capable of augmenting and replacing knowledge work. Broad-based, highly correlated sell-offs are typical of “shock periods” where investors liquidate shares due to general fear. (Figure 1).

Markets believe AI agents will displace both software and people. Among the firms most impacted by worker headcount, shares of office REIT Vornado Realty Trut have fallen 13.0% this year. In contrast, broader US REITS – including data centers - have risen 8.4% YTD.

Yet, the across the board sell-off in software is indiscriminate. Cyber-security provider Palo Alto Networks has fallen 33% from its peak. Firms that provide CRM and HR management (i.e. Salesforce and Workday) have fallen 30%-35% year-to-date. Shares of Microsoft have dropped 27% from last year’s peak. Microsoft is far more than a software company, with a massive presence in cloud computing, home to AI. Microsoft also has its own AI product releases and annuity-like income from both households and business accounts.

Looking broadly, it is clear that markets do not assume “Judgement Day” is here.  With the boom continuing in hyper-scaler capex (estimated +54% in 2026 for the large public companies), memory chip makers have seen shares rise about 45% this year (ex. Samsung, Micron).  US semiconductor shares have rallied 17% in the year-to-date, even after booming in 2025.  Investors, collectively, are not taking the view that a broad-based decline in profits of most industries is close at hand. 

Looking at share price action across narrow industry groups (Figure 2), investors are repricing business categories where demand will fall or where core activities may become obsolete (think office space and IT consulting).  In contrast, investors see AI infrastructure components and the energy to power AI as likely beneficiaries.  
Is this sustainable? Tech hardware has shown itself far more subject to cyclical busts and less  adaptive to rapid change than software.  But as noted above, the “shock” of AI may be game changing for many.

Our CIO Group Investment Portfolios invest widely across regions and asset classes, far more than tech.  Yet the fact that AI is still apparently boosting the fortunes of some industries tells us about the “market regime.”  Investors, collectively, are not taking the view that a broad-based decline in profits of most industries is close at hand. This is despite much doomsaying of late.

As the US economy has evolved over recent decades, business cycles have become less about routine, relatively benign phenomenon such as inventory building and shedding. The shift from manufacturing to services has seen production and employment rise and fall periodically, without severe financial consequences.

The larger “booms” and consequent busts are different. We think a major transition point is ahead of us. Booms stand out distinctly as their ends are far from benign.

The US recession of 2001 was historically mild, yet the US equity market fell about 50% at its low point. Tech shares were down 82% at their bottom, reflecting the wild optimism prior to the early-2000 peak (Figure 3). The recession of 2008/2009 was historically severe, but it also was preceded by an irrational explosion of optimism in the housing sector (Figure 4). One can remember unfinished housing units as far as the eye could see across the deserts of Nevada. They went unsold for most of a decade.

The AI boom has boosted spending on information processing equipment and data centers to a $729 billion pace at the end of 2025. Its broader impact will be comparable to other periods of extraordinary investment in US capital such as the internet buildout. It’s impact is happening on a much faster timeline.

The arrival of AI Agents marks the start of a second major wave of AI impact. There are 4.4 million software engineers employed in the US and 47 million globally. Executive assistant agents greet you by name when you call airlines, credit card companies and banks. They have already reduced headcount. (There are 5.4 million executive assistants and secretaries in the US in 2025).

Even as we are experiencing a third year of datacenter investment spending growth of roughly 50% in 2026, we are doubtful that rapid AI spending gains will be needed beyond 2027 (Figure 5).

AI leadership will see big winners and losers. The failure of some firms to compete with those able to adapt rapidly to shifts in demand for their services suggests a meaningful drop in AI capital spending at some point. Beyond this, resources constraints such as power and water alone will be an issue. For a time, fewer new semiconductors and memory chips will be needed.

That’s why we believe that this boom will end like all other booms. From whatever high they ultimately reach, AI-infrastructure related share prices will fall. Software firms whose business models don’t adapt will also see diminished growth rates, poor retention levels and ultimately lose profitability. Fear and despondence will likely drag down broader markets for a time.