There is a simple rule I think investors should start paying a lot more attention to as AI develops:
“Do not own stuff that gets worse as intelligence gets better.”
It sounds obvious, but think about what it actually means.
AI isn’t just creating a new industry or giving a handful of technology companies an opportunity to make more money. It has the potential to fundamentally change the economics of thousands of existing businesses. And some of those businesses could find themselves in serious trouble.
The obvious casualties are companies whose business models depend on people doing work that AI can increasingly do faster, cheaper and, in some cases, better.
Think about business process outsourcing, basic software development, routine legal work, content production and first-line customer support. These businesses have traditionally made money by selling people’s time. The more hours they can bill, the more money they make.
But what happens when a customer can get much of that work done using AI at a fraction of the cost?
It doesn’t necessarily mean demand for these services disappears. In fact, demand could increase as prices fall. But if the price customers are prepared to pay falls faster than the volume of work increases, margins are going to get squeezed.
There is another group of businesses that worries me: software companies whose main selling point is little more than an interface sitting on top of an AI model. If the likes of OpenAI or Google can eventually provide the same functionality themselves, what exactly is the customer paying for?
Then consider traditional seat-based software businesses. Their revenues are often tied to the number of employees using their products. But if AI agents can do the work previously performed by several employees, why would a company continue paying for all those software licences?
And perhaps the biggest problem is facing businesses that sell expertise which is no longer particularly scarce.
Something that was once difficult, expensive and time-consuming to produce could become almost commonplace.
That’s a pretty uncomfortable position to be in as a business owner.
So Where Should Investors be Looking?
The other side of this argument is arguably much more interesting.
If AI makes intelligence cheaper, the things needed to support that intelligence could become increasingly valuable.
AI doesn’t run on fresh air. It needs enormous amounts of computing power, electricity, cooling, semiconductor manufacturing capacity and physical infrastructure.
The more powerful and widely used these models become, the greater the demands on that infrastructure are likely to be.
This is why I find the companies supplying the physical building blocks of AI particularly interesting.
A few obvious examples:
- NVIDIA (NVDA) – supplying the GPUs and computing power needed to train and run increasingly sophisticated AI models.
- TSMC (TSM) – manufacturing the advanced chips that make much of this possible.
- ASML (ASML) – supplying the highly specialised lithography equipment needed to manufacture the most advanced semiconductors.
- Vertiv (VRT) – providing the power management and cooling infrastructure required by increasingly demanding data centres.
- Equinix (EQIX) and Digital Realty (DLR) – owning and operating the data centre infrastructure where much of this computing takes place.
These companies are interesting for a fairly simple reason: AI might get better and better, but it still needs physical infrastructure to function.
There are other areas worth considering too.
Companies with genuinely proprietary data, valuable distribution networks or deeply embedded relationships with customers may have something that AI developers cannot easily replicate.
Likewise, businesses providing essential systems that organisations rely on to operate could remain very difficult to displace, particularly where they are paid for usage or outcomes rather than the number of employees using their software.
And don’t overlook businesses where trust, regulation, liability or the physical world matter more than intelligence itself. AI might be able to provide advice, but that doesn’t automatically mean it can take legal responsibility, build a power station or operate a complex physical business.
A Different Way of Looking at Investments
I think investors need to start looking at companies through a slightly different lens.
Rather than simply asking whether a company is using AI, ask whether AI makes its business more valuable or threatens the very thing it makes money from.
Does the company own something genuinely scarce? Or is it selling something that is becoming increasingly abundant?
Will AI improve its margins, or will it give customers an excuse to pay less?
And perhaps most importantly, if AI becomes ten times more capable over the next few years, does that make the company stronger or weaker?
That last question could prove particularly revealing.
Of course, none of this means that NVIDIA, ASML or the other infrastructure companies are automatically good investments at today’s prices. Valuation still matters. So do competition, capital expenditure, margins and the possibility that expectations have already become excessive.
Nor should we assume every software company is doomed. Some will use AI to reduce costs, improve their products and become considerably more profitable.
The point is to distinguish between companies that benefit from AI and those that are simply hoping it doesn’t destroy their existing business model.
The Bottom Line
I keep coming back to the same principle:
Don’t own stuff that gets worse as intelligence gets better.
If a company’s competitive advantage depends on employing lots of people to perform increasingly automatable tasks, I’d be asking some serious questions.
On the other hand, companies supplying the electricity, computing power, chips, cooling and infrastructure needed to make AI work could find themselves in an increasingly important position.
We are heading into a world where intelligence may become abundant and relatively cheap.
But electricity, land, advanced chip manufacturing, cooling capacity and physical infrastructure are another matter entirely.
And that, in my view, is where some of the more interesting long-term investment opportunities could lie.
