Where’s the revolution? AI and the long-term portfolio

Despite the recent hype around artificial intelligence (AI), it is estimated that the technology will only result in a one-off increase in annual GDP of 7%. GAM Investments’ Julian Howard looks back at previous technological revolutions to evaluate how AI might impact a long-term portfolio.

20 July 2023

• Long-term investors need to think about AI on two levels
• First is whether the economic growth outlook is likely to change
• Second, whether stock market valuations today are sustainable

Pets.com was a new internet company which, as the name suggests, sold pet food and supplies into the US retail market. It came into being in 1998 but by November 2000 had ceased operations. Its rise was stellar, with the firm promoted during the 2000 Super Bowl and its advertising mascot even being interviewed on Good Morning America. Remarkably, Pets.com was never profitable and it became totemic of the technology boom and bust of the turn of the century. It could be said that while everything changed, nothing changed. Yes, information became instantly available, but this was not as paradigm-shifting as many would have hoped. Referencing the advent of the SMS message PayPal co-founder Peter Thiel famously lamented in 2013: “We wanted flying cars, instead we got 140 characters.” As for the impact on long-term growth and productivity, the dot.com boom appears to have had little lasting legacy. By the early 2000s the concept of secular stagnation had taken hold in academic circles as a way of explaining the sluggish growth and low interest rates which had become characteristic of the economic landscape. The dot.com boom therefore sounds a note of caution on the nature of technological change. It is with this in mind that investors should approach the current artificial intelligence (AI) mania. Any change to long-term portfolio positioning will need to establish whether the promised change will profoundly alter long-term growth and interest rate prospects, since these are vital for determining potential returns. Then of course there is the practical issue of whether anything should be done in the very near term as the tearaway performance in US large cap technology stocks threatens a repeat of previous episodes of technology overvaluation.

A full survey of AI’s impact deserves more space than available here, but it would be fair to say that the more extreme views of the future are unsurprisingly currently getting the most airtime. None other than OpenAI’s founders themselves declared in late May that “Within the next ten years, AI systems will exceed expert skill level in most domains and carry out as much productive activity as one of today’s largest corporations.” Similarly, US investment bank Goldman Sachs believes that 300 million jobs could be lost, but that the reward for the associated increase in productivity will be a one-off increase in annual GDP of 7%. These predictions, while grabbing headlines, could be overstated given that technology’s impact on society and the economy is often poorly understood or anticipated. The dot.com boom described above serves as a reminder that however awe-inspiring a given technology is, its long-term impact often feels muted. An even more vivid example was the introduction of the railway in the US in the 19th century. Robert Fogel’s Nobel Prize-winning work in the 1960s argued persuasively that the positive productivity impact was in fact far less than assumed because canals were already in existence and could not be significantly improved upon. Similarly, today several key economic sectors such as construction and agriculture are notoriously hard to automate or apply AI to and may not experience any revolution at all.

For a ‘live’ view of future growth expectations today, we need look no further than the bond markets. If the nominal yield on government bonds is adjusted for inflation we are effectively left with what bond investors perceive to be future growth expectations. The results are a bleak view of future growth across the US, Japan, Germany and the UK and, importantly, one which has not meaningfully lifted as a result of the AI excitement. A reasonable conclusion might therefore be that bondholders are sceptical given the patchy record of previous technological ‘breakthroughs’ and their limited impact on growth and productivity. This is important because it suggests that the most compelling organising thesis of future growth – secular stagnation – sadly remains intact. If AI is to be no different to previous technological revolutions in this regard, the existing challenges of worsening demographics, inequality and climate change will continue to weigh heavily on the global economy’s future prospects.

Figure 1: Where’s the revolution? Bond-implied growth prospects unchanged by AI (From 31 Dec 1997 to May 2023)

This then leaves the issue of equity valuations, which appear to be pointing the other way. The US stock market has been on a tear since 30 November 2022 when OpenAI introduced the ChatGPT large language AI model to the world. From that date to 14 July this year, the S&P 500 Index is up 11.6% and the Nasdaq 100 Index of technology stocks is up a stunning 30.1%. Valuations are unsurprisingly stretched as a result. The forward price-to-earnings yield on the S&P 500 Index is now trading at 21x while the Nasdaq 100 Index of technology stocks is trading at 30x. Much of this advance has come from the so-called ‘Magnificent Seven’ stocks – Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia and Tesla. While many are rightly concerned at this lack of breadth, which has historically tended to predict underperformance, this does not make the market ‘wrong’ per se. Even if the so-called AI revolution does not profoundly change growth or productivity patterns, it seems likely – just as with the internet and smartphones – that adoption of AI will be universally widespread. Furthermore, firms which enable the AI revolution may yet see significant earnings growth from, for example, renting computing processing power or distributing new AI-driven software applications to diverse corporations. Framing large cap technology firms as the ‘house’ rather than the ‘gambler’ in this way suggests that technology stocks might continue to benefit as businesses of all kinds seek to adopt AI into their working practices. Even beyond the current wave of enthusiasm, the technology sector’s prospects remain undimmed as a structural play that hedges against the dismal low growth future outlined earlier. Continuous innovation by well-resourced technology firms over time can create sizeable, steadily growing earnings streams which should transcend the broader macroeconomic backdrop of subdued growth.

Figure 2: Just the beginning? The Nasdaq 100 still may not have fully priced in corporate AI adoption (from 31 Jan 1996 to 14 July 2023)

AI poses a unique challenge to investors in that it rightly forces an assessment of whether the latest iterations of the technology really are going to change the world. Daunting as such a question may seem for any investor, social and economic history is in fact rich with technological precedents that proved less than revolutionary, from the railways of the 19th century to the dot.com boom over 100 years later. As such, there seems little reason today to suddenly throw out well researched theories of future stagnation. This matters because a sensible long-term investment strategy up to this point would have focused on areas of the global equity markets that can provide secular earnings growth in a low growth world. There is no obvious reason now to revise this on the current evidence. Indeed, one of said sources of secular earnings growth – large cap technology stocks – happen to be a key beneficiary of AI and may well continue to be so for some time. For long-term investors, therefore, the AI story should just be one of many stepping stones along the journey of technological progress, coolly captured over the decades via a structural allocation to the asset class. Whether AI represents a true revolution or, more likely, a broad-based evolution, should not really matter.