Markets tread a narrower path with less room for error

FOUR years of double-digit equity returns can do strange things to investors. Risk starts to feel theoretical; every dip becomes a buying opportunity; and a 20 per cent return from Nvidia somehow looks disappointing.

That is usually when it pays to become slightly less brave.

After staying overweight on equities for a full year, we recently moved to neutral the day after the US Federal Reserve hiked interest rates.

This is not a call to hide under the bed. The global economy remains resilient, earnings are growing and the artificial intelligence investment boom still has plenty of fuel.

But that “special fuel” is becoming expensive. Oil prices and bond yields have risen, while the Fed has turned more hawkish. At the same time, Wall Street expects companies to knock it out of the park, yet again.

A home run is impressive. But a home run everyone has already pencilled in is quite a yawn.

We do not think the bull market is ending, but the easy part probably is.

Four years, several fortunes

It is worth remembering how far we have travelled.

ChatGPT was released on Nov 30, 2022. Since then, Nvidia has risen by more than 1,000 per cent. Micron, Taiwan Semiconductor Manufacturing Company, Advanced Micro Devices and Broadcom have also delivered returns that would ordinarily belong in consultants’ PowerPoint slides that nobody believes.

Even the law of large numbers is beginning to bite. Nvidia is up around 15 per cent at the time of writing. In most markets, that would call for champagne. After its gains since 2022, investors react as though the waiter brought sparkling water.

The broader market has been just as formidable. Including dividends, the S&P 500 returned 26.3 per cent in 2023, 25 per cent in 2024 and 17.9 per cent in 2025. As at Oct 16, it has gained another 11.9 per cent.

An investor who entered the market at the start of 2023 would have effectively doubled their money.

But as usual, success creates its own risk. After a 100 per cent gain, investors do not merely become richer. They also quietly reset what they consider normal.

If 2026 finishes with a double-digit return, this will be only the fourth time since 1928 that the S&P 500 has recorded at least four consecutive double-digit years. The previous runs were in 1942 to 1945, 1949 to 1952 and 1995 to 1999.

If that sounds like the sort of statistic designed to ruin your breakfast, it should not.

The five-year run ending in 1999 was followed by the dotcom bust, with the S&P 500 losing 9.1 per cent in 2000. Yet, the four-year runs ending in 1945 and 1952 were followed by declines of only 8.1 per cent and 1 per cent, respectively.

Markets do not fall because they have risen for too long. They fall when something breaks.

The streak is not the problem. Valuations, liquidity, policy and profits determine what happens next. History does not come with a timer that starts flashing red after four good years.

A shakeout is not a breakdown

Our concern is more mundane. Investors have become complacent, and complacent investors are easily startled.

A modest earnings disappointment could trigger profit-taking, systematic selling or an unwinding of leveraged positions. None of these requires a recession or an earnings collapse.

Markets sometimes fall simply because too many people are standing on the same side of the boat.

A 5 to 7 per cent correction would therefore not surprise us, and it may have little to do with fundamentals. After four strong years, it would be less a crisis than a badly needed cold shower.

We would buy such a decline if the conditions in our policy, positioning and earnings (PPE) framework hold: First, policy should not induce overly tight financial conditions that stall the economy. Second, investor positioning should not show complacency. Finally, earnings should stay solid and growing.

The distinction is important. A falling market is not automatically cheap. But a cheaper market with largely unchanged fundamentals is usually more interesting.

Bonds have something to say

Equity investors may be relaxed. Bond investors are sweating it this year.

US investment-grade corporate bonds were yielding close to 6 per cent in mid-October, while high-yield debt offered around 8 per cent. Those yields have risen sharply this year, even though investment-grade credit spreads remain relatively tight.

We continue to favour shorter-duration bonds because long-term yields could remain volatile. However, hold-to-maturity investors able to tolerate greater interest-rate sensitivity can selectively consider longer-dated bonds from high-quality technology companies.

Some long-dated debt from Microsoft, Alphabet, Amazon and Meta now offers yields of around 6 per cent. These remain investment-grade issuers, although we acknowledge that a very long-tenured bond can still provide a memorable lesson in duration risk.

Investors now have a curious choice. They can buy the equity of an AI winner and rely on ambitious growth forecasts. Or they can lend to the same company at yields once associated with far less respectable borrowers.

Who holds the chequebook?

Then there is the familiar worry about market concentration. The largest technology companies represent an outsized share of the S&P 500. If something goes wrong, surely they must be the first dominoes to fall.

Perhaps. But the opposite is also possible.

The best hyperscalers could become a flight to quality. They have strong balance sheets, recurring revenues and, crucially, control over their capital expenditure. If conditions deteriorate, they can reduce spending and preserve cash.

Their suppliers do not enjoy the same luxury.

This matters because the five US hyperscalers are expected to spend around US$1 trillion to US$1.2 trillion in 2027. The estimates are forecasts, not cheques already written, but the scale is extraordinary.

US$1 trillion is roughly Switzerland’s entire annual economic output. It is about 1.7 times Singapore’s 2025 gross domestic product and more than three Apollo programmes in today’s money.

Spread across one year, it works out to US$2.7 billion a day, US$114 million an hour or nearly US$2 million every minute.

Five companies may spend in one year what a wealthy country produces in 12 months. That is no longer just a technology investment cycle, but an industrial policy conducted by a few CEOs.

Thus, the companies writing the AI cheques may prove safer than those waiting to receive them.

If hyperscalers trim their budgets, their cash flows improve almost immediately. The chipmakers, equipment suppliers, data-centre developers and power companies counting on those orders face the opposite outcome.

That is the hidden concentration risk. It is not simply that seven large stocks dominate 32 per cent of an index. It is that an increasingly large ecosystem depends on the spending decisions of a very small club.

Our outlook remains constructive, but with less room for error. We are neutral rather than negative, cautious rather than frightened.

The road ahead is still open. It has simply become narrower, with rather more expensive guardrails.

The writer is head of investment strategy, UOB Private Bank