Insights

This article was originally published in Investor Daily on Tuesday 25 August, 2026. 

The real risk to portfolios may not be an AI bubble waiting to pop, but investors sitting underweight while a capex-driven rally keeps running without them. The real risk to portfolios may not be an AI bubble waiting to pop, but investors sitting underweight while a capex-driven rally keeps running without them.

Concerns that global equity markets are overvalued, and that the current four-year bull market is set to suddenly end, have been on investors’ minds since the run began. Today, many investors, and commentators, see an AI bubble ready to pop.

However, I believe it is worth pausing to consider whether the opposite is true. Despite stretched valuations, geopolitical uncertainty, and market volatility, how are portfolios positioned should the AI-led bull market continue?

Consider this, if an investor avoided AI and semiconductor names over the past 12 months, they would have missed a 115 per cent gain – and that accounts for the share price falls we saw among some of the major players in recent times.

I remain quite bullish that this market rally can continue in the short to medium term.

If the US earning season showed us anything it is that the major players are doubling down on their AI future, and for the likes of Microsoft, this investment is starting to pay off with noticeable uptick in revenue.

In this context, my contrarian view is that one of the biggest concerns facing markets today is not a market correction but actually right tail risk, that investors sit underweight equities while the market keeps rising without them.

The last down market in equities was in 2022. Every year since, the same call has repeated: the rally has gone too far, it’s about to end. It hasn’t. US earnings growth ran at close to 25 per cent last quarter, strong enough to keep funding the rally, and the latest US earnings season showed the same pattern again: capex revised up, not down.

That’s a market grinding higher against a wall of worry on the back of real earnings, not a bubble waiting for a pop. As the bubble narrative gets repeated, more investors take profits and move underweight, making any potential catchup trade ever more painful.

This is a particular challenge for Australian institutional investors and wealth managers who are very much benchmark aware – the benchmark has never been more concentrated, or more binary.

The top ten stocks in the S&P500 now make up roughly 40 per cent of the Index. A long-only manager needs close to 8 per cent in Nvidia alone just to match benchmark, and 10 per cent or more to run a genuine overweight position. Sit on the sidelines waiting for a correction that keeps not arriving, and the benchmark leaves you behind. Chase the most expensive names blindly, and a single earnings miss becomes a portfolio event, like we saw when Meta’s disappointing earnings result sent the share price down 8 per cent in a single session.

Neither is a strategy. Investors need to separate durable winners from the businesses AI is quietly replacing, because this cycle is producing both winners and losers in equal measure.

Cheaper AI accelerates the trade

Last month, the release of Moonshot AI’s new low-cost AI model sparked bears to argue this would trigger the end of the AI investment bubble. However, again, I would argue this is a misreading of past evidence and that the opposite is true.

When DeepSeek showed the market that inference costs could fall by roughly 90 per cent through techniques such as mixture-of-experts architecture, the initial reaction was to sell the infrastructure winners. Nvidia lost the best part of $600 billion in a session.

However, NVIDIA, and its semiconductor peers, all recovered from the short-term blip and surged even higher. Why? Because cheaper AI isn’t a replacement, it just increases total consumption, driving more overall demand, albeit at a lower price to the customer; an economic phenomenom known as Jevon’s Paradox. Every 90 per cent cut in AI inference costs, so far, has been followed by consumption rising by a greater multiple.

Moonshot AI’s cheaper, more accessible model only widens the base of AI use cases. The biggest beneficiaries of AI – memory producers, foundry capacity, and the physical data centres – all benefit from that increased usage.

Sorting the winners from the losers

The more useful question for Australian portfolios: which parts of the economy will benefit from this increased use of AI, and which are being hollowed out by it.

On the winning side are cyclicals that benefit from a global AI capex cycle. Think energy, materials, industrials, and semiconductors. These are sectors Australian portfolios have been able to safely underweight for the last decade. That era is ending.

The US reporting season showed capital expenditure keeps being revised upward, not down. The mega cap market leaders driving it – Google, Microsoft, Amazon – have gone from recycling excess capital through share buybacks to issuing equity and debt to fund further investment.

On the losing side is a list that goes well beyond the obvious software-as-a-service names. It extends into staples, utilities, healthcare, and defensive sectors that lack exposure to the growth areas where investment is tracking, and who suffer from increasing price pressure driven by AI investment driven shortages. In addition, a stable cash flow is worth less in a rising-rate world – a structurally different environment to the low-rate, low growth years many portfolios are built for.

What this means for diversification

The instinct in Australia has been to look for uncorrelated return streams outside listed equities. Private markets absorbed a large share of that flow over the past decade. Some of that allocation will keep earning its place. But much of the private credit and private equity opportunity set is concentrated in the services-heavy, software-adjacent businesses now facing the most direct AI disruption. And they are materially underweight energy, materials, industrials, semiconductors. This has benefitted them in the past decade but will now be a source of headwind.

Real diversification here means building exposure to macro and thematic drivers that can move independently of both listed and unlisted equity risk.

That is the opportunity in front of Australian investors right now. Not just in the next year, but over the next decade. Markets will be defined by high capex spend, higher real rates, increased energy demand, and technology innovation as artificial intelligence increasingly powers our economies, just as electricity and the internet did before it. Investors who can build portfolios around where the capital is actually flowing, rather than where the headlines are pointing, will be best placed to seize the opportunities that come next.

By Fawaz Chaudhry, head of equities at Fulcrum Asset Management