We have long believed that generative AI had the potential not only to go mainstream, but to be the next general purpose technology that revolutionises how we live and work. Our long experience investing in the technology sector helped us to appreciate the potential for AI to be more than an incremental technology development, which led us to the launch of the Polar Capital Artificial Intelligence Fund in 2017.
Indeed, AI appears to be a discontinuous technology change, making an advance of such significant scale that, by some estimates, there have only been 10 such developments in human history; these are sometimes viewed as representing ‘100 years of progress in one step’. Ultimately these moments fundamentally alter the base for future technological progress, creating entirely new markets and generating significant economic value.
While it is still very early to evaluate its full potential, according to Goldman Sachs AI could add 1.5% to annual productivity growth and 7% to global GDP (c$7trn)1. The remarkable scale of such estimates comes from the breadth of potential applications of AI – 25% of current work tasks in the US and Europe could be automated, with the potential to impact as many as 300 million jobs globally, while half of all work could be accelerated through AI enablement according to studies by Cognizant. The implications for corporate operating expenditure are huge. Labour accounts for 40-80% of corporate expenditure, compared to 3-6% on IT. The budget for AI is drawing from the first corporate budget, rather than the much smaller IT budget that new technologies more usually vie for.
Despite the manifest risk of hype, however, we have not seen the same over-exuberance as with other early-stage technologies. Valuations of key AI stocks have largely been backed up by robust earnings growth and differentiated returns profiles.
Rapid adoption reaching impactful levels
Since the initial release of ChatGPT in late 2022, the technology has improved rapidly. Text and video creation models quickly followed, giving way to fully multimodal models that can operate across all media. Model accuracy has improved, with hallucination rates declining sharply, and models are now more relevant and drawing on more recent and real-time data. A further class of model is emerging with the release of OpenAI’s o1 that aims to better replicate human thinking with the ability to reason and self-correct as it works through answers.
A recent paper from the St Louis Fed shows that AI adoption is running at about twice the rate of adoption of both the PC and the internet
Importantly, the cost of inference (drawing conclusions from new data) has collapsed. By some estimates it has fallen by 200x since the launch of ChatGPT, facilitating widespread adoption. A recent paper from the St Louis Fed2 shows that AI adoption is running at about twice the rate of adoption of both the PC and the internet. AI adoption stands at nearly 40% after two years, while the PC and internet reached 20% penetration after three years. As a result, AI has rapidly approached a critical level; historically, it is around the level of 50% adoption that a new technology drives an inflection in labour productivity3.
Corporate adoption of AI is, on the whole, still in the early stages but we are continuously seeing more data emerge from pilot studies and the first applications of AI. These are indicating that the hoped-for productivity gains can materialise; in these early applications a productivity uplift of 10-40% is often seen, depending on the application, industry and specificity of the task being accelerated, with an average of 25% according to Goldman Sachs studies4.
The immediate impact of this will be augmentation of existing work given the gains available without significant overhaul to current working dynamics. Changing workflows and processes is always an impediment to technological adoption, although the value being demonstrated from early AI applications should act as a strong incentive for corporates. However, we have always been of the view that the longer term, and much more significant, opportunity exists in the creation of entirely new markets and methods of work. Historically, significant technological revolution has triggered the creation of massive markets that have not previously existed. We have highlighted before how the development of the iPhone did much more than capture growth in the handset market. It facilitated the creation of the mobile internet and the app economy which were ultimately far larger markets. When Apple launched the first iPhone in 2007 the handset market was worth around $100bn; today the app economy alone is worth more than $6trn.
The longer term, and much more significant, opportunity exists in the creation of entirely new markets and methods of work
In our view, generative AI is an opportunity on at least this scale. We believe it will lead to the creation of enormous new markets, many of which might not yet be visible. With such potential, there is further upside for investors, in our view. We are not saying the path will be smooth – as we have seen recently, there will be periods of volatility along the way – but the pace of innovation is such that we expect to see a bright future for AI as well as the opportunity for second-order themes in due course.
Investing in AI
While the ultimate potential of AI is still emerging, we are convinced that the winners and losers will not be confined to the technology sector. The opportunity to invest alongside these winners across all sectors is a core proposition of the Polar Capital Artificial Intelligence Fund that we launched seven years ago.
The aim is to bring our technology expertise to a global equity fund, understanding that AI disruption will bring technology-like dynamics to non-technology sectors. Identifying the winners and losers from this new paradigm will require understanding that opportunities will exist within the technology sector, in particular the ‘AI enablers’ that contribute to the development and infrastructure of AI.
AI disruption will bring technology-like dynamics to non-technology sectors
However, we believe the key differentiation is our focus on the largely non-technology ‘AI beneficiaries’ in the wider economy. These companies can benefit from faster revenue growth, higher margins or more resilient earnings outlooks – or most likely, a combination of all three. Companies may now be able to realise AI productivity gains in such a way that they can dramatically rearchitect their labour force and hence margin structures. Labour typically equates to 40-80% of companies’ spending and is a much greater budget for AI to impact than previous technologies.
Beneficiaries may also own proprietary data that is now more valuable with the improved analytical capabilities AI brings, or use AI tools to generate incremental revenue that is not captured in current expectations. We believe these components also bring the potential for multiples to rerate higher as this underappreciated earnings power emerges. It is this approach that we believe sets us apart from other AI funds that concentrate more on technology companies and the AI functionality itself. The ability to invest in the disruptive AI winners across the economy as well as avoid the many companies that will be negatively impacted will be key to investment returns.
We have previously described how various companies are exploring the potential of AI to transform their businesses, including London Stock Exchange, Publicis and John Deere.
Beyond this, we want to identify those companies that will use AI to create the entirely new markets and applications that could completely upend existing profit pools. While the early stages of innovation involve significant infrastructure capex, we believe the longer-term opportunity will be much larger in the AI applications and beneficiaries. We are hugely excited about AI potential in the industrial, robotics, healthcare and information services industries, all of which are already experimenting with and deploying AI tools.
In our experience, returns will be concentrated, so the lion’s share will go to the winners. Many incumbents will lose and lose quickly – they will not be bailed out by mean reversion. We believe that investing in the face of such rapid disruption is best done with a technology-minded framework to identify and avoid threats to incumbency, and that our expertise will help us to capture what we believe will prove a huge opportunity for investors.
1. Goldman Sachs, 26 March 2023
3. Goldman Sachs, 26 March 2023
4. AI is showing "very positive" signs of eventually boosting GDP and productivity | Goldman Sachs
