Show notes
With A.I. rapidly reaching the mass market, investors are pondering the risks and upsides to A.I. diffusion. History may provide some answers.----- Transcript -----Welcome to Thoughts on the Market. I'm Ed Stanley, Morgan Stanley's Head of Thematic Research in Europe. And along with my colleagues, bringing you a variety of perspectives, today I'll be discussing ten key lessons from the last hundred years of tech diffusion. It's Tuesday, the 27th of June at 3 p.m. in London. Tech diffusion is one of the three big themes we at Morgan Stanley Research are following in 2023. The other two being the multipolar world and decarbonization. And when we say ‘tech diffusion,’ which has become a term of art, we mean the process by which any transformational technology is adopted widely by consumers and industries. Think of the light bulb, the first power plant, the internet, and now, A.I.. Our recent analysis of the last hundred years of tech diffusion helps to shed light on ten critical questions around how, when and where stocks will be impacted from the development of A.I.. One of the most important issues to consider is how fast A.I. diffusion is happening and whether regulation can restrain this. Since its pivotal moment when it was released in November, the leading generative A.I. tools are on pace to do in one year what the internet took seven years to achieve in spilling over to the mass market, and electricity took around 20 years to do the same thing. The next critical question to consider is whether we tend to see upside or downside happen first for industries being impacted. In examining 80 structural positive and negative adoption curves over the last 50 years, we find that downside disruption often occurs sooner and twice as quickly as upside disruption. So how does the downside play out for stocks perceived to be by investors more at risk from these types of technology disruption? The market typically de-rates and waits. So valuations fall somewhere between 50 to 60% in the years 1 to 3 post-a-disruptive-event with consensus sales and profit downgrades taking anywhere around 5 to 7 years to materialize. This process is shorter for business to consumer, B2B and longer for business to business contracts, B2B. And what about the ways that upside plays out? For perceived winners, upgrades need to arrive within 6 to 12 months post the initial re-rating. However, we find that missing the first year of upside tends to have little impact on long term compound returns for investors. Investors also wonder to what degree A.I. might be a bubble. And this is a fair question considering the market excitement and froth in A.I. at the moment, but we're watching Internet search trends to answer this question. And if you look at image generation tools for A.I., we're already about 50% lower than peak search volumes. So it's a trend we're going to have to continue to watch pretty closely. Given all this, at what point do we expect killer apps to emerge that are built on top of these technologies? Well, our analysis of the last 50 examples of these killer apps emerging suggests that they tend to take a year and a half to emerge. This is why it's often very challenging to find domain specific winners in the public markets because they are still likely to be in venture backed scale up stage at the moment. But when the killer apps do emerge, the next question becomes how much value will accrue to the incumbents versus the disruptors. And on this point, history suggests that diffusion of technologies that are transformational like this have tended to lead to changes in stock market leadership over the last hundred years, with ultimately 2.3% of all companies generating all $75 trillion of net shareholder returns since 1990. In this context, are pure play or diversified stocks the best ways to play these themes? Over the long run, we believe that pure play stocks exposed to themes such as A.I., can be expected to be valued at approximately 25% premium to non pure play stocks on average. And the final two questions we get from investors take a more macro tilt. First, how much and when can we expect to see productivity gains? We are already seeing these productivity gains. The question is, what range? And we've seen anywhere between 20 to 55% for software developers, we've seen 14% for call center workers, and healthcare is also a large focus of academic research in terms of A.I. productivity and efficiency gains. Finally, there is the question of deflation. When and how much can we expect from this kind of technology? This remains the most challenging question to answer. Technology of all kinds has proven consistently deflationary, and we think this is no different. But we do suggest that investors familiarize themselves with the emerging debates on virtual assistance, which could accelerate these deflationary spillover effects. We'll continue to track all these developments around the ten key lessons and questions from history, and we'll provide you timely updates accordingly. Thanks for listening. If you enjoy the show, please leave us a review on Apple Podcasts and share Thoughts on the Market with a friend or colleague today.