Are we watching an AI bubble inflate?

After hitting a new all-time high at the beginning of March, the S&P 500 might be on track for one of its best first quarters in decades. However, the gains are almost exclusively powered by a handful of large technology firms that are benefitting from an explosion of interest in artificial intelligence (AI). This begs the question: is AI in a bubble, reminiscent of that which encapsulated tech firms in the late nineties?

First, let’s clarify our definition of a stock market bubble. We consider this a significant run-up in stock prices without a corresponding increase in the value of the businesses they represent. A company's valuation should be determined by its business fundamentals - its profits, growth rate, and similar factors. In a bubble speculation and ultimately euphoria take over.

There are indeed some parallels between the dot-com boom and the ascendancy of AI.

Firstly, both concern transformative technologies. As was the case with the internet back in the nineties, hopes are high about AI and its capacity to alter the very fabric of society – how we work, interact, communicate and so on... According to Forbes, 64% of business expect AI to increase productivity. The risk is that expectations about AI’s potential become inflated, at least in the short-term. History has demonstrated that it is notoriously difficult to predict the amount of time it will take to capture value from technological innovations. The internet did usher in profound changes, but much more gradually than people expected, leading to disappointment along the way.

There is another similarity – a frothy investment landscape. During the dotcom bubble, investors poured capital into internet-based companies indiscriminately, resulting in a sector peak value of USD 2.95 trillion before the bubble burst. Some signs of exuberance are emerging in AI; one of the most glaring examples of this might be OpenAI CEO Sam Altman’s quest to raise up to USD 7 trillion for a new AI chip project. That's more than the entire US federal budget, roughly ninety times the size of Luxembourg’s GDP, and 13 times the value of global chip sales in 2023!

As investors rush to get onto the AI bandwagon, a handful of US tech stocks are seeing their prices rise substantially – especially those perceived as “early winners”, like those who manufacture the chips needed to build AI technology and cloud service providers with the computing infrastructure to commercialise it. Consequently, the concentration of performance on the S&P 500 is now at levels unseen since the dot-com bubble era, with only 28% of constituents outperforming the benchmark itself. In turn, that means 72% of firms on the index are currently detracting from its performance!

However, there is also a key differentiating point between now and the nineties: Valuations are not stretched like they were back then. Insofar as now, the multiples of some of the leading names in AI are accompanied by decent fundamentals. At the turn of the millennium, valuations were very much momentum-driven.

Source: Bloomberg Intelligence
(Four Horsemen: Microsoft, Intel, Cisco, Dell)

Even though valuations today are slightly expensive relative to earnings, they are still far off the levels evidenced during the dotcom bubble. When you exclude the technology sector, the price-to-earnings (P/E) ratio of the S&P 500 is more aligned with the historical average (19.4x), meaning the broader optimism that spilled over into equity valuations in the late 1990s is not [yet] prevalent.

Beyond this, on aggregate, the current market leaders have strong balance sheets; leverage doesn’t appear excessive, and their cash as a percentage of market capitalisation is double what companies had during the internet bubble. Their return on equity and average margins are also nearly double what was seen during the 1990s runup.

Another feature of the dotcom bubble that is not apparent today is a frenzy of IPO activity.

For these reasons, we can say the current AI boom is not perfectly analogous with the dotcom era. Nonetheless, there is still the possibility that this buoyant theme balloons into a bubble eventually.

How can investors protect their portfolios?
The key piece of advice we can give, is to be diversified and discerning. It is crucial not to assume that today’s the dominant players will retain their position in the future. AOL is a good example of this. In the late nineties, it reigned supreme as the dominant internet company of its time. By 2008, AOL's revenue was declining at a rate of 25% each year, and in 2015, it was acquired for USD 4.4 billion - 98% lower than its peak valuation.

New disrupters come along and disrupt the old disrupters and many companies created during hype periods fail. Those who sustain are those who best understand customer needs and adapt accordingly, not necessarily those who built the underlying technology. When it comes to AI, what will be key for firms is mastering distribution, the roll out of practical applications and using proprietary data to boost user experience and demand.

A theme for the long haul
Investors should bear in mind the old adage that people tend to overestimate the effect of innovation in the short run but underestimate it in the long run. We believe that AI will bring profound changes, just like the internet did 25 years ago, but this will happen gradually. As such, it should be treated as a long-term structural theme in an investment portfolio. There will be winners and losers, and as always, it is unlikely that there is a straight line to profitability. A correction is wholly possible if signs emerge that AI is not living up to the priced-in expectations.