He says most New Zealand KiwiSaver members and global investors get their international exposure through cap-weighted indices – where individual stocks are weighted proportionally to their total market value.
The top 10 stocks in the S&P 500 now account for around 40% of the index’s capitalisation. Gardyne says that is double their share in 1990 and the highest level of concentration since the early 1970s.
Eight of the top 10 stocks in the index are exposed to artificial intelligence (AI) as chipmakers, hyperscale cloud providers (data centres) or AI-platform owners. They are Nvidia, Apple, Microsoft, Amazon, Alphabet Group (owner of Google and YouTube), Broadcom, Meta Platforms (Facebook, WhatsApp, Instagram) and Tesla.
The other top 10 stocks are Berkshire Hathaway, and Micron Technology or Eli Lilly or Walmart depending on day-to-day trading and share price fluctuations.
Chipmaker Nvidia, the largest company in the world on market capitalisation with a valuation of more than US$5.2 trillion, makes up nearly 8% of the S&P 500 Index (at the time of writing).
AI powerhouse Nvidia posted a profit of US$26.4 billion on record revenue of US$46.7b in the recently ended quarter. Photo / Getty Images
Gardyne says “a single stock representing close to one-12th of the entire US market is something we haven’t seen since IBM in the 1970s – and we know how that ended.”
Over the past 12 months, the S&P 500 has risen 27% and in year-to-date more than 9%. In the eight weeks to the end of May, alone, the S&P 500 had risen 17.5%.
Gardyne says about 80% of the index’s gain this year has been driven by AI-linked stocks: “Semiconductors and semiconductor equipment stocks – just one industry group – have generated around 50% of the index’s year-to-date return, from only 15% of its weight.
“Only about one in five stocks in the S&P 500 has outperformed the index this year. That is an extraordinarily narrow market. Most stocks are not, in fact, going up.”
Gardyne says the US stock concentration story is also elsewhere. The South Korean KOSPI has risen more than 80% this year, with two memory chip companies Samsung and SK Hynix making up 42% of the entire index.
Taiwan Semiconductor Manufacturing Co now represents more than 14% of the MSCI Emerging Markets Index – the largest single-stock weight the index has carried in 30 years.’
Gardyne says the MSCI World Index which most New Zealand investors hold through their KiwiSaver fund’s global allocation, has Nvidia as its largest position.
“Wherever you look, the same handful of stocks, the same (AI) theme and the same supply chain keep appearing.
“Investors buying a low-cost global index fund believing they are getting safe, broad exposure across thousands of businesses are, in reality, taking a very large and specific bet,” he says.
“They are betting on AI infrastructure spending continuing to accelerate, on a small group of customers – principally four US hyperscalers and a handful of AI labs – and on a supply chain that runs through Taiwan and Korea.
“But it is an active bet sitting inside a passive product, and most investors do not know they are making it.”
Gardyne says now is a good time for investors to check what is in their portfolios, and how much of the top 10 is in the hands of technology beneficiaries and whether they are happy with that.
Investors buying a low-cost global index fund believing they are getting safe, broad exposure across thousands of businesses are, in reality, taking a very large and specific bet.
Ashley Gardyne
“Active managers don’t think the market is providing sensible diversification, and it’s important for an investor to go in with eyes wide open.”
None of this is a call to abandon exchange traded funds or index investing, he says. The benefits – costs, simplicity, broad market exposure – remain real and trying to time these cycles precisely is a fool’s errand.
But there are some sensible things to think about:
Understand what you actually own. The label on the fund matters less than the underlying holdings. A global fund in 2026 may be a much narrower bet than its name suggests.Consider whether the equal-weighted version of the major indices, which give every stock the same weight rather than rewarding the biggest, might form part of a diversified portfolio.Think more broadly about geographic and sector diversification – emerging markets such as Asia, and healthcare, financial, industrial and quality value companies have all been left behind by the AI trade and may offer better risk-adjusted opportunities.
Gardyne says historically, after periods of high concentration, equal weighted indices have outperformed their cap-weighted equivalents.
Research by Goldman Sachs, looking back over a century of US market data, shows that periods of extreme concentration have consistently been followed by below-average forward returns.
Gardyne says Goldman Sachs’ work currently implies a real return for the S&P 500 of close to zero over the next decade – not because the underlying businesses are bad but because concentration this extreme tends to revert back to their historical average levels.
The dot.com peak in 2000, the Nifty Fifty in the early 1970s and the financials concentration of 2007 all unwound the same way, he says. The cap-weighted index spent years underperforming the equal-weighted equivalent as the stock leadership broadened.
In the 2020 Covid sell-off, stocks with higher exchange-traded fund ownership experienced greater volatility on the way down.
“That risk has not really been tested at the current level of concentration but the structure is now there for it to play out,” says Gardyne.
“The part I find most uncomfortable is what is happening underneath the headline numbers. A growing share of recent gains have come from the more commoditised parts of the semiconductor market – particularly memory chips made by Micron, Samsung and SK Hynix.
“Memory has historically been one of the most cyclical industries in the world. Prices rise, every producer ramps up capacity and within 18 months supply catches up and prices fall 50-80%.
“Right now, we have record-high prices and record-high capital spending on new semiconductor fabrication facilities that come online from 2027. The pattern is not new.”
Gardyne says on the demand side, hyperscaler (Google, Amazon, Microsoft, Meta) capital expenditure on AI and data centres is genuinely enormous – US$600-$700 billion this year alone and US$4 trillion over five years.
“That is real, contracted spend and over the long run it will likely prove worthwhile. But it is being underwritten by a small number of AI model providers and the financing arrangements have become increasingly interlinked.
“Chipmakers investing in their own customers, customers committing to buying chips with money raised from bond markets, and equipment vendors taking equity in fabs.
“If any one part of this slows – a bottleneck in power supply, a cash flow problem at one of the AI model providers, a single hyperscaler trimming spend – the impact can ripple quickly through the whole chain.
Model providers OpenAI and Anthropic, which developed the AI assistant Claude, are planning to list with mega initial public offerings.
“Their valuations will be trillions of dollars and they add another large portion to the index,” says Gardyne. “The result is even more concentration in the AI thematic – the tech stocks may account for 45% of the S&P 500.”
He says history offers a useful warning. At the peak of Japan’s asset bubble in 1989, Japanese equities made up close to 45% of the MSCI World Index.
A global index investor was, without realising it, almost half-invested in a single overvalued market. The Nikkei Index fell more than 80% over the following decade and took 31 years to reclaim its 1989 high.
Gardyne says today Japan has around 5% of the MSCI World Index; the United States has around 70%.
“The lesson from Japan was not that Japan was a bad market – it was that buying more of the most expensive stock, at the moment of peak enthusiasm simply because it is the biggest weight in the index, has historically been a poor strategy.
“Now is the time to be really selective,” he says. “If you look at an average US company, excluding the AI businesses, earnings are still growing, something like 8% a year – now that’s a pretty decent return.”
Fisher Funds is a sponsor of the Herald’s Capital Markets and Investment report.