Blog

Recently there’s been a lot of commentary on a potential bubble in AI. Some suggest challenges lie ahead for those exposed to US equity, where elevated valuations are concentrated in AI-related stocks. Others say that investing is worthwhile to fuel innovation in a revolutionary technology. While discussions around a bubble are important, one thing that’s missing is a reminder about investing basics when it comes to assessing US exposure. Several factors suggest being underweight in the US is sensible: the nature of current valuations, underlying economic conditions and other metrics that indicate the AI narrative is largely driven by speculation.

It’s easy to get excited by market trends, as investing always requires risk taking. Nevertheless, investors must be aware of the long-term implications and tangible risks associated with the narrative fuelling valuations.

The AI narrative: The road to superintelligence

AI’s massive capital expenditure is being channelled into the race for ‘superintelligence’ – the aim of creating AI that matches or surpasses human intelligence. However, despite impressive progress and major breakthroughs in neural networks and machine learning since 2022, there’s no guarantee superintelligence will be achieved. Or that it’ll be as revolutionary or instantaneous, and capable of generating vast profits from the get-go, as promised.

Spending on AI is already unprecedented and still increasing, driven by the costs of data centres and the energy needed to power them (see my article from June for more info). A key concern is much of the expenditure is reinvested among a small number of tech giants, with little sign of profitability beyond the promise of superintelligence. The sector has remained largely debt free so far, but Meta’s recent $25bn corporate bond sale suggests that could change.[1]

AI’s narrative is built on spending with the expectation that returns will come. But outside that, there’s little indication of how valuations will keep up if confidence weakens.

The nature of the valuations

Inflated valuations are driven by expectation and momentum. This isn’t a problem in isolation, but when the economic returns for investment appear abstract, it represents a significant layer of leveraged risk.

AI progress is intrinsically linked to the holy grail of superintelligence, but there’s a limited amount of cash available, especially since many of those companies are still seeing negative revenues.[2] Some winners could be robust enough to survive price corrections, but many won’t – a recent MIT study revealed  95% of generative AI enterprises fail.[3]  In spite of this, many early-stage AI start-ups continue to command very high valuations despite limited profitability.[4]

The Economist reported that investors seeking exposure to the S&P 500 are paying 41 times the underlying cyclical earnings, which they point out is a level only exceeded during the dotcom bubble.[5] It’s hard to see how people can remain exposed without putting significant amounts of capital at risk.

The pitfalls

AI has managed to withstand these wobbles so far, but confidence could falter. If one of the giants underdelivers, given the recycling of capital between the major players, any corrections could become contagious.

Another prominent threat comes from China. Earlier this year, reports of the Chinese Large Language Model (LLMs), DeepSeek, reaching equivalent performances to American LLMs, at significantly lower cost and without access to the US’s most advanced chips, sparked considerable market volatility.[6] Although reports were likely overstated, China still leads on innovation and has a more balanced economy. If it can build on its successes more cheaply and with better resilience, it could pose a headwind for the US dominance and have implications for related share price valuations.

Risk vs reward

Valuations dipping severely in one of the world’s premier stock exchanges would certainly inflict pain, but the full extent is uncertain. Regardless, we can safely assume that exposure to the US is becoming expensive and riskier, with the tech giants developing AI driving up valuations to what certainly looks like a bubble. The question for investors is, can my portfolio withstand a significant hit if it bursts?

Outside the US, carefully considered regional exposure has delivered solid returns this year. We also see room for further growth. This is supported by strong underlying narratives, rather than purely speculation or short-term trends.

We’re seeing opportunities in Japan thanks to corporate reforms; in Europe where stimulus, slimmer regulation and low interest rates are attractive; emerging markets boosted by the weaker dollar, which may continue under current US policies; and the UK, which has benefited from rising defence spending and strong financial returns.

There are headwinds, but these markets are generally cheaper than the US and the risks are more quantifiable, making them easier to mitigate and hedge against.

With bubbles, it can be difficult to know when corrections will happen. Many would’ve thought AI valuations had peaked long before now. But investor Michael Burry, made famous for shorting the American housing market in the lead up to the 2008 financial crisis, has recently bet against Palantir and Nvidia – he reckons they’ve peaked.[7] It may seem like missing out by not following the herd, but the benefits of steadier growth and more considered allocation are likely to put capital to work with lower risk.

This article is intended for regulated financial advisers and investment professionals only. Copia does not provide financial advice. This information is not intended as financial advice and should not be interpreted as such.