AI Bubble Talk is Premature

I’m pleased to see there is a decent amount of bubble talk around the AI theme. Bearishness and worry ‘that everything’s gone up too far too fast’ are the bricks in the wall of worry that propel bull markets. It’s true that AI stocks have performed well, and it’s also true that the theme has become central to the performance of US and global markets, but commentators who predict that it’s all going to end badly and soon are likely to be disappointed.

The natural comparator period when assessing how far into this theme we’ve come is the 1990s dotcom boom. This period is similar in that it was a prolonged build-out of a new, transformative technology. The dotcom era started in 1995 with the IPO of Netscape and peaked in 2000, a 5-year lifespan. The AI theme came to the public’s attention in January 2023 when Microsoft announced their $10bn investment in Open AI, so we are just under three years in. But it’s not the time we care about so much as the scale of the investment required.

The hyperscalers, Microsoft, Meta, Google, Amazon and Oracle are the ones building and paying for the infrastructure and they have committed to more than $330bn of investment in 2025, growing to $450bn in 2027. These are clearly huge numbers.

One of the factors that brought the dotcom frenzy to an end was the capex overbuild. Ultimately, too much fibre was laid down, which became known as dark fibre, as there was no demand for it and it sat idle. That was until YouTube and social media came along, and used it up, but that came much later. Demand for AI capacity, data centres, chips and compute power is running way ahead of the capacity to deliver it and so we can be comfortable that there is no overbuild today. Moreover, the need for power generation is bringing previously mothballed nuclear plants, like Three Mile Island, back from the dead.

The need for more power is a key gating factor in the AI race and there is a race to build it as quickly as possible. All forms of electricity generation will be used from solar to natural gas and we are seeing another bull market in the Utilities sector as these companies benefit from a big increase in demand. The bull market is also evident in the Industrials sector, as new power plants are being built and improvements to the grid are required to deliver the electricity. Even the fuel cell industry is making a comeback.

It is clear to most that this technology will change how things are done. In the same way the Industrial Revolution of the nineteenth century transformed the global economy, AI-led productivity gains are going to be game-changing. And because AI is being launched into a much more tech-savvy world than the internet was back in 1995, the uptake has been the fastest in history. It took Chat GPT just two months to reach 100m users, whilst it took 7 years for the Internet to reach the same number. And because it is free to use, and CoPilot is even installed by Microsoft onto PCs worldwide, everyone can enjoy the productivity gains in their daily lives today.

There will be a time when the AI build-out will peak as an investment theme and we are not complacent in any way about that. In fact, we watch for it as part of our process; getting out being as important as getting in. I started in the industry in September 1999 and so remember well the importance of participating in the boom but having a strategy for selling when stocks start to sour.

AI stocks have done well, but we are nowhere near the extreme moves seen at the peak of the dotcom years. Qualcomm, one of the darlings of that boom rose 26-fold in 1999 alone, a reminder of what real blow-off tops can deliver. Compare this to Nvidia, the darling this time around, is up around 4-fold since its breakout quarter in June 2023.

We believe there is still much further to go with the AI-related theme and are invested in Tech, Utilities and Industrial companies to take advantage of it. The next leg of the AI story will be the new applications and new industries borne out of AI, which will no doubt bring some as yet undiscovered companies to the fore.