It is nowadays impossible to write about the stock markets without mentioning AI.
After Netscape, 30 years ago, opened the doors of the internet to anyone with a computer, ChatGPT brought Generative AI to the world in November 2022.
Since then, the Nasdaq is up 50 per cent, with an overwhelming contribution from the AI-related ‘Magnificent 7’ (Nvidia, Meta, Amazon, Microsoft, Alphabet, Apple, Tesla), gaining 100 per cent, including an acceleration last week when Nvidia released stellar quarterly results and outlook.
In a single day, the company’s market capitalization increased $275 billion, which is, by comparison, more than the total value of Coca Cola – or the annual GDP of Finland. It’s not alone.
Together with Meta, Microsoft, and Amazon, these four companies account for 90 per cent of the $1,500 billion increase in Nasdaq value in 2024. Still this year alone, Arm holdings, another chipmaker, or Super Micro Computer, provider of AI infrastructure, are up respectively 80 per cent and 200 per cent.
No doubt, it’s hot, and worryingly familiar for those who, like me, have grey hair. So, is there an AI stock bubble, and what should we do?
A bubble is the exuberant escalation of an asset price, which massively disconnects from its intrinsic value. History is full of them: from the Dutch tulips and UK’s South Sea Company centuries ago, to the most recent ‘.com’ and US housing at the turn of the millennium. They all share these three characteristics: rapid price increase, euphoria, and disconnection.
Then, they burst.
AI goes ballistic
No doubt, AI stocks show a vertical surge, and signs of exuberance, including a prominent Wall Street institution recently calling Nvidia ‘the most important stock on planet Earth’ (which reminded me of a report from 2001 calling Enron ‘the best of the best’). What about the third, crucial element, which is the disconnection from fundamental value? This is where it gets nuanced.
First, the surge in Nvidia (and others) responds to the supersonic growth of their earnings. This is definitely much sounder than other typical bubble drivers: pure multiple expansion, speculation about hypothetical future success, emotional group thinking, or even cynical hope that someone will pay more (the ‘greater fool theory’).
To illustrate, Nvidia Q4 revenues were up 265 per cent year-over-year, and profits close to 500 per cent. It’s not small numbers, $22 billion and $12 billion, respectively. Market performance, and its concentration, is not totally disconnected from earnings performance and their concentration.
Profit growth as well as forecast revisions are incomparably better for tech and AI than for the rest of the market. This is the first, and the major, reassuring indication.
Still, the valuation multiples that the market applies to these earnings are very high. This implies, and thus requires, future growth to be sustainably superior. Again, AI is still in early stages, with considerable investments in infrastructures, while widespread adoption is arguably only getting started.
It is not unreasonable to discount elevated future growth, which would rapidly bring the valuation multiple back to more sensible levels.
Bottom-line, the hype and the market performance are present, but there is enough fundamental grounding to hope that it is not just another market frenzy poised to end in tears. One of the typical issues with bubbles, however, is precisely that we usually are not sure that it is one until it bursts, with a lot of emotional pressure in the meantime.
What should investors do?
Discipline is the remedy to emotions to minimise bad decisions. It starts with three golden investment principles: diversification, selectivity, and a long-enough horizon to take risks.
Volatility is a given: stock markets always overestimate the short-term and often underestimate the long-term. When there is exuberance, this time distortion is even worse, and it adds a lethal feature which is ‘undifferentiation’.
What Amazon’s stock price of 1999 tells us
Let me illustrate with the ‘.com bubble’. Pretty much all hyped companies lost 90 per cent from their peak when it burst. Most had poor business models and disappeared. But there was also Amazon that IPO-ed in 1997. When the dotcom frenzy took markets by storm in late 1999, you could buy an Amazon share for the today equivalent (adjusting for splits) of around $5.
But it was scary: it had already doubled in the previous four months, despite being loss-making for years. Indeed, from $5 in December 1999, the price fell to $0.5 in early 2001, and it took 8 years to see $5 again.
Buying in 1999 at $5 would have been very painful, a terrible decision. Except for the fact that it is now worth $170.
Looking in the rear window is always easy, but my point is that investing during a bubble is not necessarily a recipe for disaster, under two conditions. First, pick the best possible business model, exposed to the right segments of the megatrend, with massive competitive advantages and barriers to entry and make sure it persists (think Google vs Yahoo, Apple vs Nokia).
Second, give yourself the power to stomach volatility: a long investment horizon, a diversified portfolio, and a reasonable calibration.
We do not believe that AI is currently an outright bubble, even if we expect volatility. Elevated valuations make sense for some of the high-quality companies involved. Nvidia and a few others are part of our recommended securities, for years and still today.
But we would always advise against any oversized position. For a 5-year investment horizon, we recommend an equity weight of around 40 per cent, of which US tech is between 5- and 10 per cent, with selectivity and diversification.
