AI: Hype or genuine investment opportunity?

January 28, 2025
Join my first live webinar: "What's next for 2025?". Reserve your spot here.

The transformative impact on the world that many people believe will follow the widespread application and use of AI has been described as the fourth industrial revolution. The previous three, mechanisation in the 18th century, electrification in the 19th and the digital revolution in the 20th century, all introduced fundamentally new ways of living and working and profoundly changed human society. But in each case, it’s important to remember that these changes took place over a long period following the first discoveries of steam power, electricity and the construction of Alan Turing’s first computer. That is not to say that the application and adoption of AI will take decades. Nevertheless, it is important to understand that, despite the vast sums being invested in the technology across the world, its current applications are still relatively limited.

Experts in the field have different views about its future impact; some are fearful, and others are very optimistic; others have a much more prosaic view of its current potential, but some are very excited indeed, like Google’s CEO, who believes it will be more transformational than fire.

I have to say that I am excited by its potential but also fearful (for obvious reasons) of AGI (artificial general intelligence). Although a long way off, this type of AI will match or surpass human cognitive capabilities across a wide range of tasks. However, this post is not about my own views about where the technology may end up; instead, I hope it will be an informative guide to those interested in the technology and who may be looking for a way to invest in it.

Many investors naturally look to the largest technology stocks in the US as the most obvious way to gain exposure to AI. For example, the performance of Nvidia, which tripled in 2024 to a $3 trillion valuation, is entirely the product of its dominance of the market for chips, originally designed for gaming applications, that now power the AI revolution. It specialises in GPUs (graphics processing units), which are able to perform calculations concurrently in a way that the CPU (central processing unit–the chip required to run applications and the operating system) cannot, making them more energy efficient and better able to deliver the scale of computing power their customers need for their AI models and data centres.

Whilst Nvidia is a clear and obvious beneficiary of the AI revolution, not least in its financial results, many of the US market’s leading social media, hardware, and software stocks that have driven substantial market gains over the last few years have yet to show a return on their AI investments. Despite this, leading US technology stocks are ramping up their investment even further. According to the FT, after massive increases in 2024 (to over $200 billion), Amazon, Microsoft, Meta and Google will spend significantly more in 2025. Microsoft announced in December that its investment in this area will exceed $80bn in 2025. Quite apart from the huge transition in their business models represented by this scale of investment, it will also come with considerable execution risk, much greater capital intensity, and higher depreciation and operating costs. It’s also worth bearing in mind that this investment is not just about building and equipping data centres, it also includes significant expenditure on building clean electricity generation infrastructure to power the centres, including nuclear, which is notoriously expensive and very challenging to deliver.

Currently, it is not at all clear what the returns on this investment look like, nor have these businesses been particularly transparent about what their investors should expect in terms of future returns either. Maybe this is not surprising given the relative youth of this opportunity, but given the scale of what is happening, this has rapidly become a key risk factor confronting these businesses.

Investors should also be concerned about the costs of delivering the capability that is, for example, now available with ChatGPT Pro. Some analysts have estimated that the price per query for an AI chatbot can be up to ten times the cost of a Google search. Customers are also demanding more of their AI providers, and for the big tech companies to justify their investment in the technology, it needs to be capable of replacing expensive human capital. This may be possible in the future, but given that in a recent test, the latest OpenAI model used more than $1,000 of computing power per task to solve a problem that a human completed for $5, there is a long way to go.

Nvidia would argue that the answer to these challenges is better chips, which are more powerful and energy efficient (which it hopes to supply). However, this must also be combined with cheaper energy, too, not an insignificant problem to solve.

The AI industry has been in the news this week for a very different and potentially very disruptive reason. It has been announced that a Chinese AI start-up called DeepSeek has been able to develop an advanced model (which, in comparison tests, has matched the performance of the industry’s leading models) in a fraction of the time and apparently, although this is yet to be verified, at a fraction of the cost of the established US players such as OpenAI.

Analysts and investors around the world have been trying to understand the implications of what this relatively unknown company has been able to achieve. In summary, it has apparently developed a sophisticated model with a tiny budget of only $5 million whilst only having access to a limited supply of older Nvidia and AMD GPUs. This announcement has received a huge amount of attention for understandable reasons, and it has had a significant impact on the share prices of a broad range of technology stocks. In essence, what DeepSeek has announced may challenge the widespread belief that larger, more expensive models and datasets are always superior by demonstrating high performance with significantly fewer, albeit very smartly deployed, resources.

The implications of this announcement will take time to sink in. Although not completely verified in all respects, it is being seen as very disruptive to the established view. Many commentators are already concluding that it will result in a significant re-evaluation of current AI spending plans by the hyperscalers and a significant reduction in the future demand for Nvidia chips, for example, and data-centre building. This may be true, but it is by no means certain. AI industry leaders like OpenAI, Microsoft, Amazon, Google and Meta will still want to maintain a lead in this industry, and I am confident that, as a result, they will still want to have the best data centres, the best models, the largest datasets and the most computing power.  I don’t see any of these businesses being seduced by the opportunity to develop this technology, for example, on cheaper, older chips or diminished datasets. I still see the AI industrial revolution as a very demanding and expensive race that the industry leaders will not be in a hurry to opt out of.

Having said that, what does worry me is the route to a return on the massive investment that the industry is devoting to this technology. If it does anything, this announcement from DeepSeek reminds me of the precarious path that the industry’s leaders are embarking on, which, as I have said, they will be very reluctant to leave but which, for the time being, is likely only to stress the returns on investment these businesses have been so familiar with for a long time.

Consequently, I still believe that for those investors looking for ways to play the AI industrial revolution, the most obvious and crowded trades should probably be avoided. Right now, what is becoming clear is the scale of these businesses’ ambitions in this area, but what is far from clear is the impact this investment will have on their financial results in the future. Rather like the adage attributed to Mark Twain–when everyone is digging for gold, it’s good to be in the pick and shovel business–the best way to invest in the AI industrial revolution may well be to look to the companies manufacturing and supplying some of the key components that will enable the efficient and cost-effective processing of vast amounts of data required either to train and improve AI models or in real-world commercial applications. An added advantage of this strategy is that many businesses doing this are trading on “normal” and sometimes very low valuations, which provides a significant and additional attraction.

Some of these businesses are not exclusively focused on AI chip demand, and so it pays to be aware of how these other activities are performing. However, it is very interesting that markets appear to have ignored, or at least not focused on, these businesses’ ability to grow quickly over the years ahead as a result of the same demand trends that have elevated stocks like Nvidia.

A good example would be a range of memory chip companies that have in the past and still are relatively lowly rated cyclical businesses reflecting the industry's capital intensity, the volatile supply-demand dynamics of the markets the businesses address, and their corresponding inventory and pricing cycles. However, for some businesses in this sector, strong demand growth for very advanced high bandwidth memory (HBM) chips, which are required in all demanding AI applications, is providing a new and powerful growth dynamic that markets have yet to properly recognise. A couple of examples of businesses in this sector that I think are poised to benefit from these trends are SK Hynix in Korea and Micron Technology in the US. Both of these companies suffered recently from the downturn in traditional memory chip prices and volumes which impacted their financial results in 2023 in particular. Both trade on very low ratings and yet both are also forecast (looking at consensus expectations) to deliver very strong growth over the next few years at least. Both may also be affected by the investor turmoil that this DeepSeek announcement has triggered, but my guess is that they are very attractive at these levels.

Another good example might also be the relatively small number of global businesses that make highly complex semiconductor manufacturing equipment. The demand for this is forecast to grow significantly as the AI revolution takes hold in the years ahead. The best of these is the Dutch business ASML, which dominates the market for highly advanced extreme ultraviolet lithography (EUV) machines. It, too, has had a few difficult years recently and has also been affected by recent events, but it is poised to grow strongly in the future. Although more highly rated than SK Hynix and Micron, it is starting to look very attractive. Very recently, for example, its largest customer, TSMC, talked about increasing its capex in 2025 by 34% and a further 20% per annum growth through to the end of the decade.  

In summary, my conclusions are that AI has the potential to do what industrial revolutions of the past did, but that in the immediate future, it is far from clear how AI can be profitably commercialised other than, possibly, in the automotive sector, where increasing vehicle autonomy is something that consumers appear willing to embrace and pay for, but where regulators are still treading very carefully. Ultimately, I do believe that those pioneering in this revolution will find profitable ways to deploy and exploit this incredible technology, but this is many years away. In the meantime, I would caution against investing in the crowded and highly valued trades that many have highlighted as the key players in this new industry. Instead, I would focus on the suppliers of the “picks and shovels” to this industrial revolution and seek out those businesses likely to see significant demand growth but whose share prices do not yet reflect this key role that they will increasingly play.

Disclaimer: These articles are provided for informational purposes only and should not be construed as financial advice, a recommendation, or an offer to buy or sell any securities or adopt any particular investment strategy. They are not intended to be a personal recommendation and are not based on your specific knowledge or circumstances. Readers should seek professional financial advice tailored to their individual situations before making any investment decisions. All investments involve risk, and past performance is not a reliable indicator of future results. The value of your investments and the income derived from them may go down as well as up, and you may not get back the money you invest.

Related posts

No items found.

Subscribe to receive Woodford Views in your Inbox

Subscribe for insightful analysis that breaks free from mainstream narratives.