Written by David Rogers and published in The Australian.
For three years, the stockmarket has had one obsession: artificial intelligence.
The chip makers, cloud operators and data-centre builders feeding them accounted for almost every percentage point of gains – and soaked up most of the spare capital – leaving everyone else on the sidelines.
Last week, that obsession wobbled. The Nasdaq Composite, the Philadelphia Semiconductor index and the chips-heavy South Korean sharemarket dived to three-month lows.
Credit default risk on AI infrastructure names such as CoreWeave soared as investors began to ask the question they had mostly avoided: Will all this spending on data centres actually pay off?
Closer to home, a run of planned floats of data centre companies such as Firmus and AirTrunk in Singapore have been pushed back or pulled altogether. NextDC, the local champion of the
AI infrastructure trade, lost 40 per cent of its value in a single session last week as community backlash delayed its data centre expansion.
For a market that has had little direct AI exposure to begin with, even that modest slice is now under pressure. Australia’s indirect AI exposure – resources – has been punished.
For fund managers who have spent years underweight the AI darlings and paying for it with underperformance, the wobble feels like vindication, though most are too careful to celebrate loudly.
LGT Wealth Management head of public markets Todd Hoare believes the next phase of returns will come from a different group of companies altogether: the ones using AI rather than building it.
“We’ve already seen these AI beneficiaries expand in terms of the access points you can get at”
The money has moved from Nvidia to the hyperscalers, then to the chip foundries and tool makers, and more recently into the bottlenecks of copper and electricity.
His view is that the next wave belongs to “AI adopters” – old-economy companies that use the technology to lift margins rather than build the infrastructure beneath it.
A software company lifting its margin from 30 per cent to 31 per cent barely moves the needle.
But a low-margin retailer, logistics operator or miner lifting margins from 5 per cent to 7 per cent effectively boosts profitability by 40 per cent. And because these businesses trade on low valuation multiples to begin with, any improvement tends to be rewarded disproportionately by the market.
In the railway boom of the late 1800s, it wasn’t the railroads or steelmakers that captured the lasting rewards. A quarter of US railroad mileage went bankrupt by the mid-1890s.
It was companies such as Rockefeller’s Standard Oil which used cheap freight to build an empire.
Cisco told a similar story in the dotcom era: it built the plumbing of the internet but it was Amazon, Google and later Netflix that captured the value.
For Australian investors, Hoare believes this cuts both ways. The local market has little direct AI exposure beyond NextDC and parts of Goodman Group, which makes it something of a natural hedge when the AI trade wobbles, as it did last week.
But it is also stacked with asset-heavy, labour-intensive businesses in retail, resources, logistics and banking which is exactly the kind of companies that stand to gain most from applying AI to existing operations rather than building new infrastructure.
He is less convinced about the banks, and says that competitive pressure may force them to pass AI-driven cost savings on to customers rather than keep them for shareholders.
Other investment managers see a broader opportunity opening up beyond the AI winners altogether.
Datt Capital chief investment officer Emanuel Datt says the extreme concentration of returns in a handful of large AI-exposed names over the past year has pushed small caps to historically cheap valuations relative to large caps – a set up that has typically preceded strong multi-year returns for the sector.
“The current environment for small caps is extremely attractive”
Datt says, noting that private equity buyers have been circling cheap, cash-generative ASX-listed small caps.
He expects any wobble in confidence around AI’s economics to send investors back towards parts of the market that are not correlated with the theme.
Eric Marais, head of investment specialists at Orbis Investments, is more cautious about calling a turning point. Orbis ran a stress test of 10,000 random 50-stock portfolios since ChatGPT’s launch in 2023 and found the index beat 71 per cent of them. The most differentiated portfolios lagged the most. Narrow markets like this one tend to reward looking like the index, not straying from it.
“The answer to a narrow market is not to look more like it”
Marais says, arguing genuine diversification comes from portfolios built differently, not from timing a rotation.
None of this means the AI trade is finished. But after years of one-way traffic, last week’s wobble is a reminder that capex cycles eventually meet reality.
History suggests the lasting winners from any infrastructure boom are rarely the builders themselves.
Receive ongoing access to LGT Wealth Management’s insights and observations - a curated stream of thought leadership, market perspectives, and strategic updates designed to inform sophisticated investors.
