AI – should I care?
Like many new and transformative technologies, many of us feel somewhat ‘at sea’ as to what they actually are. With so much talk of bubbles and market exuberance barely three years since ChatGPT landed, it would be easy to dismiss AI as just another fad.
That would be a mistake.
In March 2023 Bill Gates said, “In my lifetime, I’ve seen two demonstrations of tech that struck me as revolutionary: the GUI (Graphical User Interface) and ChatGPT”.
Jeff Bezos (Amazon) added, “I don’t see how anybody can be discouraged about AI right now-the benefits will be gigantic”
Investment at unprecedented scale
The sheer size of the investment in AI by tech leaders suggests they wholly believe AI is the next frontier. In Q2 2024 Google CEO Sundar Pichai said “The risk of under-investing is significantly greater than the risk of over-investing” while in Q3 Meta’s Mark Zuckerberg commented, “The very worst case would be that we have just pre-built for a couple of years”.
Robeco reported in November, that the combined capital expenditures of US hyper-scalers, neo-clouds, and Chinese hyper-scalers are now expected to rise 60% year on year in 2025 to USD488.5 billion, which is an upward revision of USD 67 billion from estimates published in August. Estimates for 2026 cloud spending also jumped to USD 645 billion.
Leveraging, finance and market risk
While the large hyperscalers have strong margins and balance sheets, the AI roll out has led to a wave of debt financing, and Morgan Stanley estimate that over half of the USD 2.9 trillion in AI investments expected between 2025 and 2028 will be funded by credit and debt sources. Tight credit spreads would suggest the market is unconcerned at this point, but as more complex structures with special purpose vehicles and off balance sheet joint ventures are construed, the risk increases as we have observed in previous periods of exuberance.
Understanding what AI really is
At the moment, AI looks like new software, but unlike traditional software, AI uses a probabilistic framework as opposed to traditional software’s ‘deterministic’ models.
Traditional software is deterministic: it follows explicitly defined rules and logic written by developers. Given the same inputs, it will always produce the same, predictable outputs.
AI software works differently. Instead of fixed rules, it relies on models trained on data to identify patterns and make inferences. As a result, its outputs are probabilistic—based on likelihood rather than certainty—and may vary even when given similar inputs .
AI is not new, but it’s different now
While AI has captured public attention, forms of AI have been embedded in the economy for more than a decade, in the form of machine learning, supporting applications such as routing, fraud detection, recommendations and optimisation across supply chains.
What has changed is capability, scale and accessibility. Advances in model design, exponential growth in data, and surging computing power have pushed AI systems to levels of performance that now rival or exceed average human outcomes across a range of tasks. This shift has important implications for long-term economic growth and industry structures.
What happens for example, if you don’t need 1,000’s of people to do certain tasks, like the steam engine in the UK, which gave Britain the equivalent labour of 5x its population by 1900.
Productivity, growth and economic change
Already we are seeing AI reach human level capabilities in new functions. Invesco in their August 2025 report “Exploring Artificial Intelligence”, provide an illustration which tracks the accelerating progress of AI models.
Estimates cited by Invesco suggest that generative AI could deliver a productivity uplift comparable in magnitude to the introduction of the internet and personal computing, though realised over many years.
Many predict AI to be a solution to many of the problems western economies face, such as aging workforces and populations, rising debt levels and a lack of productivity.
When big ideas meet market reality
However, as it stands today, PIMCO point out in their recent publication, “The economics of AI scale”, that the current investment in US AI infrastructure is hard to comprehend when you do the maths. At the current rate, assuming $20 per month for every MAU (monthly active user) it will service 4.9b MAUs. For context, the US population is 350m and the global population ~8b so either half the people in the world will use AI hosted on US -based servers, or that non-human users (bots, agents, enterprise software bundling) will constitute a large share of MAU demand. PIMCO believe optimism appears evident in AI-related net present value.
What this means for investors
Beyond simple subscription models, much of AI’s long-term value is likely to come from its ability to generate new ideas, products and technologies yet unimagined, much like Uber and Air BNB did with the internet.
While AI appears to be genuinely ‘game changing’ technology, history shows that market leadership in the early stages of innovation does not always translate into durable investment returns.
Periods of rapid technological change often attract significant capital, rising leverage and elevated valuations, increasing the importance of selectivity and discipline.
At Saxe Coburg, we do not seek to invest by chasing themes or short-term market enthusiasm. Instead, we access structural opportunities such as AI through experienced managers with deep insight into market dynamics, valuation risk and capital cycles. This approach allows clients to participate in long-term innovation while remaining focused on protecting and growing capital across market conditions.
Get in touch with Mark or Sam if you’d like to discuss how themes such as AI fit within your portfolio, we’d be happy to arrange a time to discuss further.