The scale of investment underway is exceptional. Across five leading US hyperscalers – Amazon, Microsoft, Alphabet, Meta and Oracle – annual capital expenditure climbed from US$162 billion ($283b) in 2022 to US$448b in 2025. It is expected to exceed US$700b in 2026. That is equivalent to more than 2% of total US GDP, or more than double the entire inflation-adjusted cost of the Apollo Moon programme, spread across all its years, in one single year. Importantly, this investment is being made by profitable businesses responding to real demand, not by speculative start-ups alone. But its influence extends well beyond their balance sheets: the 10 largest companies now comprise about 40% of the S&P 500, with many closely tied to the AI build-out. Their fortunes therefore carry increasing weight in the returns of the wider market.
Genuine growth in AI does not guarantee shares are fairly priced. Photo / Getty Images
For all AI’s novelty, the market response to it has historical precedents. The closest recent comparison is the internet boom of the late 1990s.
The internet unquestionably transformed commerce and communication, but the technology-heavy Nasdaq Composite rose nearly 300% from the beginning of 1996 to the end of 1999, even though many internet companies had little or no realised earnings. Railways offer an earlier version: the infrastructure transformed transport and trade, but investor enthusiasm also funded speculative projects and pushed share prices beyond what their economics could support. Japan’s economic rise in the 1980s was similarly real, but confidence in its permanence became embedded in asset prices; after the market peaked in 1989, Japan entered a prolonged period of asset deflation and economic stagnation.
Those examples do not prove the AI theme is overvalued. But they do show that genuine transformation and excessive expectations can coexist.
Building on Templeton’s warning, two investment disciplines remain especially relevant: respecting market cycles and staying anchored to fundamentals.
Ignoring cycles can lead investors to mistake temporary conditions for permanent ones. A long run of rapid growth can quietly reset what people think a normal return looks like, while rising prices can lead investors to underestimate the risks they are taking to generate that return. Exceptional performance can persist longer than expected, but periods of return well above the long-term norm have often given way to more subdued stretches, or corrections, as prices and expectations adjust.
Fundamentals provide the second anchor. Earnings, revenue, cash flow and debt levels tie a valuation to how the underlying business is actually performing. They rarely point to a single “correct” price, but they ground any view of what an investment is really worth. Abandon that foundation, and the investment case comes to depend on market psychology, speculation and momentum instead.
AI may prove every bit as transformative as many expect, but being right about the technology is not the same as being right about the price.
Together, those principles matter most when market leadership narrows to a handful of names. In an index weighted by company size, the shares that climb fastest tend to gain greater weight. Investors can end up more exposed to those companies after they’ve had an exceptional run, leaving their results increasingly dependent on the same few names continuing to deliver.
As AI-linked firms become increasingly important to major equity indices, investors’ fortunes are more closely connected to the performance of a relatively small group of market leaders. Photo / Getty Images
Spreading investments more widely reduces that reliance. But it comes with a visible cost. A diversified portfolio or fund can lag while the market’s biggest companies keep climbing. That gap does not itself mean risk management has failed. It may instead reflect a deliberate choice not to let long-term results hinge on the same narrow group continuing to meet already elevated expectations.
The benefits of diversification often become clearest after market leadership changes. During the dot-com downturn, the S&P 500 Information Technology sector fell about 80%, compared with 47% for the broader index. Japan offers a longer-term example:
Investors diversified beyond its domestic market participated in decades of much stronger global equity returns while the Nikkei 225 took 34 years to regain its late-80s peak.
The challenge is that investors must make those diversification decisions before their benefits become obvious. That principle sits at the heart of a risk-managed active approach. Rather than simply accepting the composition of an index, we assess investments individually and in the context of the wider portfolio or fund, including concentration risk. Concentration can amplify gains while those investments are rising, but it can also magnify losses when conditions or expectations change.
Being active managers does not mean we are sceptics of innovation – we invest in transformational businesses and want clients to benefit from their growth. Active management gives us the flexibility to participate in the opportunities presented by the largest index companies while managing the associated risks.
Being right about AI does not mean being right about its price. Photo / 123rf
It means we can select individual investments, adjust exposures as conditions change, allocate across markets and asset classes, and construct portfolios around a fund’s particular objectives. The purpose is not merely to differ from an index, but to apply research, judgment and discipline in ways we believe add long-term value.
We therefore weigh three things together: an investment’s potential return, the price being paid for it, and the risk involved in pursuing that return. When the potential reward doesn’t justify the risk – including the additional concentration it would create – we’re prepared to limit our exposure or look elsewhere, even while a narrow part of the market continues to run hot.
The aim isn’t to lead every market over every period. It is to deliver strong outcomes consistent with each fund’s objectives, while avoiding excessive dependence on any one company, theme or market environment. That may feel unrewarding when the risks being managed have not yet materialised. But abandoning risk management because markets have kept rising is a little like cancelling insurance because there hasn’t been an accident for a while. Over the long run, the S&P 500 has delivered an annualised total return of about 10%. Returns over the past three years have been considerably stronger than that. This does not tell us when returns will slow or whether a correction is imminent, but it does caution against treating this exceptional period as permanent.
We can’t say when market leadership will change, or whether today’s enthusiasm for AI will prove justified. That’s the point. Sound investing doesn’t require predicting the future perfectly – it requires preparing for a range of ways it might unfold.
AI may reshape the world in ways we cannot yet anticipate. But it will not remove the need to understand market cycles, weigh fundamentals and manage risk. Those disciplines are not a rejection of change, but rather a way of investing through it.
Disclaimer: This article is intended to provide general information only. It does not take into account your investment needs or personal circumstances. It is not intended to be viewed as financial advice. You should not rely on any information in this communication in making financial decisions.
Catch up on the debates that dominated the week by signing up to our Opinion newsletter – a weekly round-up of our best commentary.