When we think about the advent of electricity as a transformative technology, it’s not the electricity itself that warrants consideration, but rather what electricity has made possible. And that distinction may prove important as we think about Artificial Intelligence (AI). 

Tucked away in Lower Manhattan, New York, is a street called Pearl Street. Most people walk past it without a second thought. Yet at 255 Pearl Street stands one of the most important sites in modern economic history.

It was here, in 1882, that Thomas Edison built the world’s first commercial power station.

Only three years earlier, Edison had successfully demonstrated the incandescent light bulb. The invention was revolutionary. Yet the light bulb itself was not the real breakthrough. The breakthrough was the system that sat behind it: the generators, cables, meters and distribution network that allowed electricity to move from a laboratory into homes and businesses.

By 1900, investors knew electricity would change the world. They were obviously right, however:

1. What they did not know was which companies would win,
2. They did not know how long it would take, and
3. They certainly did not know where the profits would ultimately accumulate.

The fascinating thing about electricity is that the economic winners were not always obvious ones. Generating electricity proved essential to modern life, yet electricity utilities rarely became the most profitable businesses in the economy. The greatest economic value is often accumulated elsewhere in factories powered by electricity, in new industries enabled by electricity, and in entirely new business models that could not have existed without it. History tells us that while the future was relatively easier to predict, the winners were much harder. This is relevant today. We know AI will change the world. On that, there is little disagreement.

Similar to electricity, what we do not know is:

1. Which companies will ultimately win,
2. How long it will take, and
3. Most importantly for investors - where the profits will accumulate.

We know where many of today’s profits are accumulating and who the market darlings are. Semiconductor designers, equipment manufacturers and infrastructure providers are earning extraordinary returns from the AI buildout. But it would be a mistake to assume that tomorrow’s profit pools will look the same as today’s. After all, investing is ultimately an exercise in valuing future cash flows. While parts of the AI value chain are already highly profitable, many of the companies at the centre of the AI narrative remain only modestly profitable or loss-making. Their valuations therefore depend less on what they earn today and more on assumptions about the profit pools they may capture many years into the future.

History suggests that this is where investing becomes difficult. The current winners are visible and increasingly priced for that visibility; the future winners are not. The technology may be obvious, but the economics and where the enduring profit pools will ultimately emerge remain far less certain.

Every era has its transformational technology.

Every era has its transformational technology, and today, it is AI. In the 1920s, it was radio. One of my December reads was Andrew Ross Sorkin’s book titled 1929. I expected a story about leverage, speculation and the Wall Street Crash. Instead, I found myself captivated by the technology of the era. Radio was the internet of the 1920s.

For the first time, a single voice could reach millions of people simultaneously. Sporting events became national events. Presidential speeches reached audiences far beyond town halls. Celebrities emerged with unprecedented reach. News was no longer confined to local newspapers. Information travelled at a speed and scale previously unimaginable.

Adoption was explosive. The number of radio stations in the United States increased from just five in 1921 to more than 500 by 1923. Sales of radio equipment grew from $60 million in 1922 (around $1 billion today) to more than $840 million by 1929 (around $15 billion today). Radio had become one of the defining technologies of its age.

At the centre of this revolution was RCA, and investors could not get enough of it. Between 1921 and its 1929 peak, RCA’s share price rose almost 100-fold. It was, in many respects, the Nvidia of its generation. Investors were not buying RCA because of what it earned. They were buying it because they believed radio would change the world.

They were right, radio did change the world. However, the investment lesson came later. By 1932, despite radio continuing to spread rapidly throughout society, RCA’s share price had fallen by almost 98%. The technology succeeded, but the investment was far more complicated.

Nearly a century later, investors once again find themselves trying to identify the companies closest to a transformational technology. OpenAI, Anthropic and SpaceX are expected to command extraordinary valuations despite generating little or no profit today. Investors are paying not for today’s cash flows, but for the possibility of tomorrow’s. This is not an argument against AI.

Nor is it an argument against these companies. It is simply a reminder that technological success and investment success are not always the same thing. The future is often easier to predict than where the economics will ultimately settle.

Yes, AI is here, but something unusual is happening 

The world’s greatest software companies have started behaving less like software companies and more like utilities.

Microsoft, Amazon, Alphabet and Meta are now engaged in one of the largest private infrastructure buildouts in modern history. Their spending is going into data centres, semiconductors, servers, networking equipment, land, power and cooling. These Hyperscalers are projected to spend around $700 - $800 billion on physical infrastructure. 

The numbers are striking. Microsoft’s net property and equipment rose from $135.6 billion in 2024 to $205.0 billion in 2025. Its additions to property and equipment increased to $64.6 billion, and the company disclosed a further $32.1 billion of commitments for new buildings and improvements, primarily related to data centres.

For years, these businesses generated extraordinary free cash flow. Today, more of that cash flow is being recycled back into physical infrastructure. In some cases, the phrase “free cash flow” itself becomes more complicated.

Balance sheets, once dominated by intangible assets, are increasingly filled with tangible ones: data centres, servers, networking equipment, power infrastructure and land. The language remains digital, but the economics has become industrial.

That observation may prove to be one of the defining developments of the AI era. For decades, the winners of the digital economy became more capital light as they grew. Today, the leaders of the AI economy appear to be becoming more capital intensive.

The question for investors is profound: what happens to returns on capital, profitability and valuation multiples when software businesses start to resemble infrastructure businesses?

Capital intensive businesses can be enormously valuable.

Railroads, utilities and telecommunications networks were all valuable. But they rarely enjoyed the economics associated with the world’s greatest software companies. Artificial intelligence is frequently described as the next software revolution.

However, the scale and nature of the spending suggest a different possibility: AI may ultimately prove less like the internet and more like electrification. That distinction could be enormously important.

Electricity was not simply another product. It became a foundational input into almost every industry in the world. Factories did not become electricity companies; they became factories powered by electricity. In the same way, banks will not become AI companies. They will remain banks powered by AI. The same will be true of retailers, asset managers, manufacturers and many other businesses.

This changes the investment question. If AI is primarily another software product, the economics are likely to accrue mainly to those building the applications. But if AI becomes a foundational input into the global economy, as electricity ultimately did, then the largest profit pools may emerge in industries and business models that are not yet the centre of attention.

History suggests that when a technology becomes infrastructure, the biggest beneficiaries are often not those producing it. They are those who learn how to use it to reshape industries.

There is a quote often attributed to Henry Ford:

“If I had asked people what they wanted, they would have said faster horses.”

Whether Ford actually said it is almost beside the point. The idea is what matters.

People tend to imagine the future as an improved version of the present: a faster horse, a better newspaper, a more efficient taxi, a better hotel, or a better television channel.

Yet the greatest beneficiaries of technological change are often not those who simply improve an existing model. They are those who use new technology to challenge the assumptions on which an industry is built.

Uber did not create a better taxi company. Airbnb did not build a better hotel chain. Netflix did not create a better video-rental store. YouTube did not emerge from a traditional broadcasting network. Amazon did not build a better shopping mall.

They did not simply improve the prevailing business model; they challenged its underlying assumptions and reinvented how the industry worked. In doing so, they changed both how value was created and where the profit pool accrued. 

As an asset manager, we are already experiencing the benefits of AI. Our research is faster. We can process more information, analyse more companies and investigate opportunities across more geographies than would previously have been possible with the same number of analysts. These productivity gains are real, valuable and already changing how we work.

But in many respects, this is still the “faster horse” stage of AI: using a new technology to perform existing tasks more efficiently.

The largest investment opportunities may emerge when businesses move beyond efficiency and begin to ask a different question altogether:

1. Who is challenging the assumptions on which an industry operates?

2. Who is changing how value is created?

3. And, most importantly, who is shifting where the profit pool ultimately accrues?

History teaches us that profit pools do not simply grow; they move

The internet produced an extraordinary range of new businesses that changed the world by challenging established industries, overturning prevailing business models and redirecting value away from incumbents. It did not merely expand existing profit pools. In many cases, it disrupted them, reshaped them or created entirely new ones.

Our task as investors is therefore to remain vigilant about where AI may cause profit pools to be disrupted, enhanced or repriced. At this stage, we do not know with confidence where the greatest value will ultimately accrue. But we do know that, if AI lives up to its promise, the economic structure of many industries will change materially.

That leaves us with a fascinating possibility. Perhaps the greatest beneficiaries will not be today’s AI leaders, the model developers, semiconductor companies or cloud providers. Perhaps some of the most important winners have not yet been founded.

Somewhere today, in a university dormitory, a garage or a small start-up office, entrepreneurs may be building a business that becomes possible only because AI exists. History offers many examples in which the technology provider earned attractive returns, but the greatest economic value accrued to those who used the technology to reinvent an industry.

The truth is that we do not yet know where those profit pools will settle, and that may be the most important admission an investor can make. Artificial intelligence may become as foundational as electricity. If it does, the largest winners may not be the companies generating intelligence, but those that learn how to use it to rewrite the rules of the game.

AI may be inevitable. Investment returns are not.

If you visit Pearl Street in New York today, there is very little to suggest that you are standing near one of the most important economic developments in human history. Most people walk past it without noticing. Yet from that small power station Thomas built in 1882 emerged a technology that transformed almost every aspect of modern life. Electricity changed how we work, communicate, travel, manufacture, consume, etc. Investors who recognised its importance were right. What they could not know was where the economics would ultimately settle. Some profits accrued to utilities. Many accrued elsewhere. Entire industries emerged that would have been impossible to imagine from the vantage point of Pearl Street.

Artificial intelligence may prove to be a similar story. The technology appears transformative, and the adoption appears inevitable. The investment opportunities appear enormous. However, humility is encouraged in instances like this because nobody knows who tomorrow’s market darlings will be. As investors, our task is not to predict whether AI will change the world. It almost certainly will.

Our task is to identify where the profit pools migrate, who captures them and whether the price we pay today adequately reflects those future possibilities. History rarely provides answers. It often provides better questions. Perhaps the most important question today is not who is building AI. It is they who will ultimately use it to change the rules of the game. Because technology changes the world. Economics determines the winners.