Join our experts as they explore planning, structuring, and investing for a changing world.
Article last updated 14 September 2026.
In this webinar, Sanjiv Tumkur, Head of Equities, Ben Derber, Equities Desk Lead, Simon Lapthorne, Senior Research Analyst, and Claire Titmarsh, Equity Research Analyst, explore whether the current excitement around artificial intelligence represents a speculative bubble or the early stages of a lasting structural shift in the global economy and what it all means for investors.
From chess computers to generative AI
To understand where we are, it helps to understand how we got here. AI has evolved significantly, from early chess-playing computers to today's generative AI systems that can write, code, design, and reason. The modern AI supply chain stretches from power generation and semiconductors, through vast data centres, up to the "hyperscalers" - Microsoft, Amazon, Google (Alphabet), and Meta, and on to the AI models we interact with directly, such as OpenAI's ChatGPT, Anthropic's Claude, and Google's Gemini.
Each layer of this chain represents both a challenge and an opportunity for investors.
Is this the dot-com bubble all over again?
The dot-com era of the late 1990s saw extraordinary enthusiasm for technology, followed by a painful crash. So are we heading the same way?
Today's situation is meaningfully different, and the data supports this. The companies at the heart of the AI revolution - Microsoft, Alphabet, Meta, Amazon, are not speculative start-ups. They have enormous revenues, strong profits, and robust balance sheets. By contrast, many dot-com era firms had little more than a website and a business plan.
Valuations tell a similar story. The technology sector is currently trading at around 22 times earnings. At the peak of the dot-com bubble, that figure was closer to 50 times. There is froth in parts of the market, but we are not in the same territory.
Crucially, demand is currently outstripping supply. This is the opposite of the dot-com era's "build it and they will come" mentality. Hyperscalers will spend $1 trillion in 2027 on AI infrastructure, with a further $4-5 trillion expected over the next four to five years. That is not speculative, it is a response to real and growing demand.
The long view: Amara's Law
Simon Lapthorne highlights a useful lens for thinking about AI's trajectory: Amara's Law. Named after the American scientist Roy Amara, it holds that we tend to overestimate the impact of technology in the short run, and underestimate it in the long run.
There may well be short-term turbulence, valuations can overshoot, sentiment can shift, and not every AI company will succeed. But the long-term potential of this technology is, if anything, likely to be underappreciated today. Patience, as ever, is a virtue for investors.
The infrastructure bottleneck and the opportunities it creates
One of the most compelling themes discussed was the sheer scale of infrastructure required to power AI. The largest AI data centres consume nearly 2 gigawatts of electricity per year, enough to power around 2 million households. Along with power, AI is also constrained by memory supply (specifically high bandwidth memory), networking, and water for cooling.
These bottlenecks create real investment opportunities, not just in the obvious technology names, but in the "old economy" businesses that make AI possible. Think grid infrastructure, nuclear energy (including small modular reactors), gas turbines, and water infrastructure.
Some of the most interesting opportunities may be in companies you wouldn't immediately associate with AI. Caterpillar, for example, supplies power generators for data centres. Copper miners provide the cables that connect them. These businesses stand to benefit from the AI buildout without carrying the same valuation risk as the headline technology names.
Which sectors face disruption and which are more resilient?
Not every industry will experience AI in the same way. Our team identified areas of genuine vulnerability: graphic design, parts of advertising and media, and roles that involve routine information processing. Where AI can replicate a task quickly and cheaply, disruption is likely.
But some sectors are more resilient than people might expect. Finance and accounting software, for instance, requires a high degree of accuracy, and the consequences of errors are significant. This creates a natural brake on AI adoption in certain areas, at least in the near term, and favours incumbents with deep regulatory expertise.
How we're approaching AI in our portfolios
Our investment approach to AI is grounded in the same principles that guide everything we do: valuation discipline, quality businesses, and strong balance sheets. We are not trying to pick the "winner" of the AI model race, that is an uncertain bet. Instead, we favour a "picks and shovels" approach: investing in the companies that enable AI, regardless of which model ultimately prevails.
Areas of interest for our team include chip designers such as Nvidia, semiconductor equipment makers such as ASML, Taiwan Semiconductor Manufacturing Company (TSMC), the major hyperscalers, and software companies with high switching costs and proprietary data, including Microsoft and London Stock Exchange Group.
The key is that investors don't need to predict the future of AI to invest well through it. What matters is having a framework grounded in fundamentals, one that remains alert to risk and changing circumstances, and is patient enough to let quality companies do their work over time.
How Rathbones can help
AI is reshaping the investment landscape and it raises important questions for anyone thinking about their long-term financial future. Whether you're an existing client or considering working with us for the first time, our investment managers are here to help you think through what this means for your portfolio.
If you're already a Rathbones client, please speak to your usual contact. If you're new to Rathbones, you can get in touch via the enquiry form below to start the conversation.
