AI changes everything – so why do you have the same portfolio?

Every investor seems to know and accept that AI is changing the world. But what many seem to be overlooking is the speed of change in AI itself. 

The release of ChatGPT towards the end of 2022 felt like a watershed moment, blurring the lines between science fact and fiction. Every day since then has seemingly brought news of yet another industry, company or business model at risk of disruption from AI.  

The irony is that the list now includes the AI companies themselves. ChatGPT’s revenues have recently fallen behind those of rival Anthropic, with data from Ramp suggesting the increase in paid subscriptions for the latter business are coming at the expense of those in the former. 

This raises an important question: if AI is changing the world as we know it, and AI itself is changing even quicker, why are investors’ portfolios the same? 

Valuations up, cashflows down 

At the end of January 2023, the top five constituents of the MSCI World, at 12.5% of the index, were Apple, Microsoft, Amazon, Alphabet and Nvidia¹.  

Fast-forward to today and the top five constituents of the MSCI World are Nvidia, Apple, Microsoft, Amazon and Alphabet – but in an even more concentrated weighting of 19.0%². 

If AI is changing the world as we know it, and AI itself is changing even quicker, why are investors’ portfolios the same?

You could argue I’m being disingenuous here as, with the exception of Nvidia, only a tiny fraction of these companies’ revenues are directly generated by AI. 

However, while AI is not responsible for how these companies make their money, it now overwhelmingly accounts for how they spend it.  

Will the ‘$1trn’ investment in AI pay off?  

Microsoft, Alphabet and Amazon – plus Facebook owner Meta, which sits just outside the top-five – plan to spend close to $725bn on AI infrastructure in 2026³, a figure we expect to hit $1trn by 2030. This is up from just $150bn in 2023⁴.  

As a result, not only do these companies’ free cashflow yields sit below where they were in that year, they also sit below where they were a decade ago. 

If such capital expenditure starts to generate healthy returns, or if a clear winner emerges among the so-called hyperscalers, this may turn out to be a shrewd use of investors’ capital. Until that time, we are happy to take an underweight position in this sub-sector of the market.  

We are not bearish on AI 

This does not mean we are bearish on AI. Although we have pivoted away from the companies collectively set to make a trillion-dollar investment in the technology, we have pivoted towards the loose body of industries and businesses set to benefit from it.  

A key tenet of our investment process involves seeking out stocks where the potential upside outweighs the potential downside by a factor of two to one.  

Among the beneficiaries of AI spending, this has frequently led us towards companies where demand for their product or service is outstripping supply. That $725bn is an enormous amount of money to push through the economy in a single year and unsurprisingly it has led to bottleneck after bottleneck. 

A couple of years ago, we would have probably named semiconductor designers (such as Nvidia) and manufacturers (such as Broadcom) as the only real beneficiaries of this trend. These companies have high incremental margins (meaning any increase in revenue results in an exponential increase in profitability) so their cashflows spiked just as those of the hyperscalers were beginning to collapse.  

Super-normal profits in the AI value chain 

But since then, we’ve learnt that AI data centres need to be wired up in a fundamentally different way, leading to a surge in demand for laser equipment. Storage has gone from a sickly industry that only made hard drives for PCs to suddenly becoming a limiting factor in AI content generation and cloud growth.  

$725bn is an enormous amount of money to push through the economy in a single year and unsurprisingly it has led to bottleneck after bottleneck

And it’s not just tech. At a more mundane level, labour provider Comfort Systems is seeing a net retirement of electricians, welders, plumbers and pipe fitters, meaning there are more tradesmen leaving the industry than are coming in. If you're Meta or Alphabet, you’re desperately trying to pay these guys as much as you can to make them carry on working.  

This supply/demand imbalance has allowed many companies in the AI value chain to generate super-normal profits. We see nothing to suggest this trade will come under threat any time soon – especially if the hyperscalers’ spending commitments remain in place. 

Opportunities amid AI disruption 

Of course, every investment has its price and with these bottleneck beneficiaries having already re-rated, there may come a time when another sector or theme offers a better trade-off between risk and reward.  

Storage has gone from a sickly industry that only made hard drives for PCs to suddenly becoming a limiting factor in AI content generation and cloud growth

Counterintuitively, this may eventually lead us to the sectors currently deemed most at risk of AI disruption. Fears of obsolescence have pushed down valuations in areas such as software and consultancy, even though earnings growth has in many cases outpaced that of the market.  

Should such fears turn out to be misplaced, investors will reappraise these as fast-growing companies generating healthy levels of free cashflow and a re-rating will likely follow.  

We are currently a long way from knowing which companies in these sectors will prove immune to the threat of AI. But we will continue to monitor them for clues that suggest their business models are more robust than previously thought. 

The point is, things change. And so should your portfolio.