For much of the past two years, global markets have operated on a simple assumption: spending on artificial intelligence infrastructure was worth almost any price. Data centres, semiconductor capacity and cloud computing were treated as essential investments rather than optional extras, and companies willing to commit capital aggressively to the AI build-out were rewarded accordingly. Last week, that assumption came under its most serious test yet.

Investors start asking about returns, not just growth

The so-called hyperscalers — Microsoft, Amazon, Google and their peers, whose vast networks of data centres underpin the AI boom — have been burning cash at a striking rate in pursuit of AI dominance. That spending has begun to sharpen the focus on corporate earnings and, more specifically, on the path to actual profitability from AI rather than the promise of it.

Semiconductor shares had already softened heading into results season, as investors questioned whether AI-related investment had become too concentrated in too few names. What stood out last week was a shift in what the market chose to reward. Microsoft’s shares advanced after it signalled a more disciplined approach to capital expenditure, while Meta fell sharply despite management’s confidence in the long-term AI opportunity, as investors reacted against continued heavy spending commitments. For perhaps the first time in this cycle, balance sheet restraint impressed investors more than ambition did.

This matters because of where it sits in the typical arc of an investment cycle. In the early stages, markets tend to reward growth almost regardless of cost. It is only later that investors begin asking harder questions about the return on that spending. Last week looked like one of the clearest signs yet that parts of the market are moving into that second, more discerning phase.

None of this means the AI trade has run its course. Despite pulling back close to 10% from its peak, the Nasdaq remains near record levels, underlying semiconductor demand is still robust, and the broader AI ecosystem continues to lift sectors well beyond technology itself — metals, power infrastructure, industrial automation and selected international markets are all still seeing meaningful tailwinds from the build-out.

Smaller stocks start to close the gap

One consequence of this shift in sentiment has been a broadening of market leadership. Having surged sharply between April and June, the Nasdaq-100 has since retreated to around 110 on an indexed basis, down from a peak near 120. The S&P 500 Equal-Weight Index, by contrast, has climbed steadily to roughly 112, overtaking the Nasdaq-100 in recent weeks. In practice, this means strength is no longer confined to the handful of mega-cap technology names that have driven most of the rally — the average stock, rather than just the giants, is now doing some of the work.

Geopolitics and a defiant Fed add to the uncertainty

Away from AI, geopolitical risk remains a live concern. The fragile diplomatic situation between the US and Iran took a further turn via Yemen, where Houthi rebels are reportedly weighing transit charges for vessels using the Bab el-Mandeb Strait — a route through which an estimated 10–12% of global trade passes. Traffic through the strait has increased this year as tensions around the Strait of Hormuz diverted some shipping and eased pressure on global energy flows, but any serious disruption would force vessels onto the far longer route around the Cape of Good Hope, adding cost, delay and inflationary pressure across supply chains.

Central banks are watching that risk closely, and the Federal Reserve gave investors little comfort of its own last week. Rates were left unchanged, though three committee members voted for an immediate hike — a clear signal that inflation concerns have not gone away. Chairman Kevin Warsh’s press conference did little to settle nerves, and the accompanying FOMC statement painted the picture of a central bank determined to bring inflation back to its 2% target even if that means holding rates higher for longer. Bond markets responded accordingly: real Treasury yields climbed to their highest level since the Global Financial Crisis, and volatility rose across asset classes. Only a few months ago, investors were confidently pricing in rate cuts; today, the conversation has shifted toward the prospect of rates staying elevated well into the future.

What this means for portfolios

Taken together, last week felt less like a rejection of the AI story and more like the start of its next chapter. Investors are no longer debating whether artificial intelligence will reshape the economy — that question has largely been settled. The debate now is about who ultimately captures an attractive return on the enormous sums being invested to get there.

At the same time, the familiar risks of 2026 have not gone anywhere. Supply chains remain exposed to geopolitical shocks, inflation has yet to be tamed, and central bankers remain unwilling to declare victory. For internationally mobile clients navigating this environment, the case for diversification beyond concentrated technology exposure continues to build. A portfolio positioned across sectors — with meaningful exposure to areas benefiting from the same structural trends as AI, including industrial metals, infrastructure and Japan — is better placed to weather a market that is starting to ask harder questions about value, not just growth.

At NEBA Private Clients, we help internationally mobile clients build portfolios that can withstand exactly this kind of shift in market sentiment, combining diversified, evidence-based investment strategies with the cross-border planning expertise our clients need as they navigate an increasingly complex global landscape.

This article is based on insights and analysis provided by Francesca Le Feuvre of TEAM.