What history tells us about the AI trend

Investors typically look for patterns – something in the past that may give them a sense of what will happen in the future. And with Artificial Intelligence all the rage, investors are looking for guidance as to what the impact might be for stocks and for the broader economy.

The typical playbook for new technology goes like this. New technologies emerge with the potential to transform society. Everyone gets very excited. Investments flood into the sector, allowing the development of the opportunity. Then everything takes a bit longer than people thought. Too much capital chases the returns, leading to excess capacity. Expectations for growth are pushed out and profitability is delayed. There’s a shake-out, with winners and losers. In the end, the new technology has a significant impact on the economy, and maybe even exceeds initial expectations, but not all of the early investors benefit. The railroad boom of the late 1800s and the telecom boom of the late 1990s/early 2000s are the most commonly cited examples. 

As we turn to Artificial Intelligence (AI), this raises some important questions. Is this how AI investments will play out? If so, where are we in that process? Who will be the winners and losers? Why might it be different?

At first glance, the AI experience seems to be following the historical pattern. There’s a huge amount of interest – if newspaper headlines are any guide. Everyone is talking about how Artificial Intelligence can transform businesses, for better or worse. 

There’s also been a huge amount of investment, and it looks set to increase. The chart below, from the Financial Times, shows capital spending from some of the big so-called hyperscalers like Alphabet, Amazon, Microsoft and Meta – large tech companies that are investing, mostly in data centres, to power the AI revolution.

This isn’t just a future vision. The chart below shows private sector manufacturing construction in the US. Policies in the previous US administration helped drive a significant increase in manufacturing construction.

Turning to the equity market, the experience of US equities since 2022 can, with a bit of gymnastics, be seen as similar to the period 1996-2001, the height of the dot-com boom. The chart below compares the performance of the S&P 500 between 1996 and 2001 (light line), and the period October 2022 to July 2025 (dark line). It might not repeat, but it does seem to rhyme. 

So, there’s been a lot of optimism. Have we seen any cracks in the story? A few headlines suggest so. The latest version of ChatGPT has been viewed as disappointing in some quarters, not showing the big improvement that users had hoped for. A study from Massachusetts Institute of Technology (MIT) indicated that 95% of AI pilot projects didn’t have an impact on the profit and loss of the companies that implemented them.

Most recently, the latest earnings report from Nvidia, the chip designer, showed extremely strong revenue and earnings growth, but perhaps not quite as strong as the most optimistic forecasts. At first glance, these might not change the long-term story for AI, but stories like these can shift expectations on the speed of adoption.

What about some other points of comparison with the past? There’s an interesting one in terms of where the investment is coming from. Equity markets had an important role to play in the dot.com boom, with an explosion in the number of IPOs (initial public offerings).

The chart below, from professor Jay Ritter, shows the number of IPOs on the Nasdaq and New York Stock Exchange over time. We can see the sharp increase in IPOs in the mid 1990s, and the relative paucity of IPOs in recent years. This means the equity market has not played a major role in financing the AI build-out – at least not directly.

Where is the money coming from? There are a few sources. First, the hyperscalers are spending aggressively, as we showed above. Those are listed businesses, by and large. That cash might not directly come from equity investors as new capital, but it’s still their money and they expect to earn a return on that investment. 

There’s also a lot of capital coming from private credit funds, that have grown aggressively over the past few years, and traditional debt financing. They are helping to finance the build-out of large data centres across the US, and elsewhere. If the adoption of Artificial Intelligence isn’t quite as fast as everyone hopes, this is one area where we could see some stress.

Looking back to history, we can worry about businesses disappearing. Lots of companies that IPOed in the late 1990s disappeared fairly soon afterwards. That seems less likely this time around. Most of the spending on AI is coming from large, highly cash generative businesses. Their share prices might suffer with some AI disappointments, but the businesses should continue. 

Where does this get us? There’s a lot of enthusiasm around Artificial Intelligence and that is being translated into significant capital investments. Historically, the prospect of significant innovation has attracted a lot of capital, and some of that gets wasted. That can cause volatility in financial markets.

So far, a lot of the spending has come from businesses that can deploy their own cash, rather than rely on equity or debt markets. And those businesses are already seeing the benefit, most notably in the growth of cloud computing revenues. We think they’ll continue to invest.

And while the speed of progress on Artificial Intelligence might be slower than the optimists would like, we think that most businesses are still in the early stages of realising the potential value from AI projects.