The Growing Concentration Risk in AI
The artificial intelligence buildout has reshaped equity markets, and with that comes a specific and growing risk that investors need to reckon with: concentration. We routinely describe the NASDAQ as tech-heavy, but the reality now is that even the S&P 500 has become tech-heavy, because the so-called "Magnificent Seven" names loom so large within it. This concentration is not just an index-level curiosity — it is directly reflected in the composition of ordinary people's portfolios.
The central concern is that investors may be far more exposed to a single theme than they realize, and therefore highly vulnerable to any shift in sentiment. We have already seen evidence of this fragility: a recent pullback in the semiconductor trade demonstrated how quickly weakness in one theme can drag down an entire portfolio when that theme is represented too heavily. The AI theme is especially dominant in two areas — US large-cap stocks and emerging markets — which makes broad-based portfolios far less diversified than they appear on the surface.
For that reason, thoughtful diversification is a central theme of the current mid-year outlook. The argument is not to abandon the winners, but to recognize that parts of the market have been left behind by the recent rally, and that these overlooked areas now represent attractive opportunities worth examining.
Unbundling the Hardware Trade
Over the last three and a half years, the areas that benefited most were in tech hardware. When you unbundle that category, it splits into chips and, more recently, memory. Memory in particular has performed exceptionally well. A striking illustration of the concentration problem: this year, roughly ten stocks contributed about 60% of global market returns, and a large share of those were memory names — the "Microns of the world" and similar companies.
Given how far this segment has run, the case is to diversify away from an exclusive focus on memory and hardware. The current positioning is underweight the hardware segment broadly. Instead, more value is seen in the hyperscalers — companies like Alphabet, Meta, and Amazon — which have actually not performed as well in recent months, leaving them relatively more attractive.
Do you have to be in semiconductors, memory, and infrastructure? Probably yes, at least in some form — but you do not need to be too directly exposed. There are ways to play semiconductors and the broader hardware story without concentrated direct exposure, and indeed some hardware exposure is held in these portfolios. The point is about degree and method, not wholesale avoidance.
Selectivity and the "Perceived AI Losers"
After several strong years for equities, selectivity has become more important than ever, and this applies not just to tech but across the whole market. Three areas stand out as places worth adding to.
The first is the group of businesses that have been labeled or perceived as "AI losers" — chiefly software companies. Adobe is an example of this kind of opportunity: a software business that the market has punished on fears of AI disruption, but which may in fact be well positioned.
The second is financial services, which has received relatively little attention despite significant developments. Within financials, the appealing sub-areas include analytics businesses, insurance brokers, and the payment space. In each case the thesis is that the market overreacted to fears of AI-driven disruption.
Why the Fear of Disruption Is Overdone in Financials
What is actually driving the opportunity in financial services? In the first quarter, a wave of new AI model releases (the Anthropic releases were cited) fueled significant fears about disruption in the software segment — a fear that was widely highlighted. But that anxiety spilled over into other high-quality businesses as well. Dominant credit-card companies like Visa and Mastercard came under pressure at the same time.
The counter-argument is that these are high-quality, deeply entrenched businesses embedded within global payment networks, and the sell-off was an overreaction. The same logic applies to exchanges that operate associated data businesses. There is an open question about whether AI will make data collection dramatically easier and thereby erode these companies' advantages — but the view here is that these firms possess a genuine moat, and the market has probably overreacted to the perceived threat.
Broadening Beyond Equities: Healthcare, Alts, and Non-US Bonds
Under the same banner of thoughtful diversification, the strategy extends into liquid alternatives, healthcare, and diversified fixed income. The core worry, again, is that large-cap US stocks are dominated by a single theme.
Healthcare is one answer, because it offers structural growth drivers while being less tied to the AI cycle and less tied to the broader economy. That makes it a useful source of diversification that does not simply replicate the dominant AI exposure.
For a multi-asset investor, the role of diversification itself must be reconsidered. A key question is: to what extent do bonds still provide diversification in today's environment? With inflation still running at elevated levels, traditional bond diversification is less reliable than it once was. The response is twofold: using alternatives (liquid alts) as a diversifier, and looking at non-US high-grade bonds to build additional diversification into portfolios. The willingness to go outside the United States for high-grade fixed income is a notable part of this approach.
The Next Wave of AI Monetization
Looking at the bigger picture, AI-driven disruption has itself created the attractive opportunities described above — in high-quality software, payments, analytics, and insurance. But a natural question follows: with AI spending still climbing, how will AI monetization actually play out, and where can investors make money on it?
The expectation is that the payoff will be broader than the hardware infrastructure play that has dominated so far — which is precisely why diversifying outside the semiconductors matters. Ultimately, the productivity gains from AI are expected to be impactful. In some cases, the very software businesses that were weak in the first quarter have the ability to take AI, integrate it into their existing platforms, and produce value-added services. In other words, the next wave of value creation will go beyond the hardware play that has been so dominant in recent months and years.
A Possible Speed Limit on Spending
On the spending side, the analysis extends beyond just the headline amount being spent by hyperscalers to include how that spending is being financed. There is a growing reliance on debt issuance and secondary issuance, which is increasingly driving the AI buildout. That financing dynamic could impose speed limits on any further increase in spending from this segment — an important caveat for anyone assuming the current pace of AI capital expenditure will continue unabated.
The Core Takeaway
The through-line across every point is the same: the AI theme has become so dominant that ordinary portfolios are far more concentrated and correlated than they look. The prudent response is not to reject AI exposure but to hold it more selectively and to deliberately broaden out — into overlooked software and financial-services names bruised by disruption fears, into healthcare with its economy-independent growth, and into alternatives and non-US high-grade bonds. The productivity gains from AI are real and will spread well beyond the hardware layer, but the financing behind the buildout, and the extreme concentration of returns in a handful of names, are reasons to diversify thoughtfully rather than chase the crowd.
