Is the AI boom a re-run of the dotcom bubble?

“This time it’s different” - this is a phrase that should immediately fill investors with scepticism and fear. But could there be a modicum of truth in it?

We looked back to the dotcom era and other big technology booms and busts to see what lessons could be learned and found some compelling arguments why this time could actually be different.

• Get our updates on markets, ISA funds, pension saving and much more

The casualties of infrastructure

On the one hand, the lesson from history seems clear: the companies that invest heavily in developing the infrastructure behind a new technology do not generally go on to benefit from it.

Back in the 1990s, the giants of the internet were not Google or Amazon, but the telecommunication titans. Companies like WorldCom and Global Crossing spent billions laying undersea cables and building internet backbone infrastructure. But the market was oversupplied with capacity, competition was brutal, and both collapsed in the early 2000s.

Their story of these ‘lost giants’ is no one-off. Other internet ‘builders’ that went on to collapse include Canadian telecom giant Nortel, which, at its height, accounted for more than a third of the total valuation of all companies listed on the Toronto Stock Exchange. “Early data centre and hosting pioneers such as Exodus Communications and PSINet also went bankrupt, with their assets bought by stronger operators,” says Matt Egerton, co-manager of the Fidelity Future Connectivity Fund.

A key backdrop was a massive overbuild of capacity in the late 1990s. Many companies invested heavily on the assumption that internet traffic and revenues would scale rapidly. But demand arrived later than expected, and, when it did, broadband margins became commoditised, squeezing margins. “The sector’s heavy leverage made that price reset fatal for companies with weak balance sheets,” Egerton adds.

Instead, the profits went to companies like Facebook, Google and Amazon, who developed new ways to monetise the technology.

“It’s the companies who come up with the use cases that win not the ones that carry the traffic. In every infrastructure boom, the carriers of the traffic ultimately become commoditised,” says Ben Rogoff, lead manager of Polar Capital Technology Trust plc.

“Before the internet, it happened with canals, railroads and electrification. We have cloud infrastructure we’ve spent 20 years building that today looks like canals did after the railways came along. AI workloads require a completely different infrastructure to the internet. But history demonstrates that most infrastructure builds end badly.”

By the end of the UK’s rail boom in the 1800s, most of the companies building the railroads had gone bust.

Will history repeat itself?

“Drawing on historical parallels, semiconductor companies today are the public-market ‘picks and shovels’ of the AI boom. Capital intensity remains high and valuations in parts of the technology sector have expanded on expectations that AI will drive multi-year growth,” says Terence Tsai, Egerton’s co-manager.

Other AI ‘builders’ include Alphabet (Google’s parent company), Microsoft and Meta, who are developing data centre infrastructure and/or investing in large language models, like ChatGPT.

As during the dotcom bubble, valuations for the AI ‘builders’ have soared, as investors set their hopes on a technological transformation.

But Rogoff is sanguine: “The companies that are embarking on the biggest capital spends of their lives are all aware of these historic parallels. The tech companies today want to be AI model makers but also to create the applications that people will use. Almost none anticipate just providing the model that someone else will monetise.”

Just look at Microsoft: it once provided operating systems for PCs but now it provides a whole host of applications that people use and pay for - whether on their PC or another device.

Rogoff adds that investors need to think laterally about which companies will be best at finding new uses for AI. “We have a dedicated AI fund [the Polar Capital Artificial Intelligence Fund] that we launched eight years ago - around 60% of that fund is non-tech companies. Walmart, for example, has said it doesn’t think it will need to hire anyone for the next three years, in part thanks to AI gains. It’s one of the big companies in the portfolio.”

Foundations are looking more stable

Tsai also believes this time is different. “Unlike the speculative environment of the dotcom bubble, today’s leading technology firms are fundamentally strong: they operate with proven business models, resilient earnings, and healthier balance sheets,” he says. “Where telecom operators faced intense competition, the leading semiconductor companies of today benefit from enormous competitive advantages, not just around chips themselves, but across entire ecosystems, and supply chains.”

He currently likes companies such as Meta, which he believes has a strong AI monetisation opportunity, and application software players like Salesforce and Intuit.

Part of the problem in previous infrastructure buildouts was that companies were simply too early, as the timeline to monetisation was much longer than many people realised. 

“A lot of the applications of the internet needed broadband and the smartphone. It took another 10 years for smartphones to come about, which made the audience for the internet profoundly bigger,” Rogoff says. Similarly, it took decades for the UK to create a dense national network on railways.

“When it takes that long, it makes sense that the companies failed to monetise. If it takes three to five years to build the AI infrastructure, there’s a much better chance the builders will be able to monetise.”

He adds: “History rhymes but it doesn’t repeat. Some people not in AI are leaning heavily into the internet example as a reason not to trade. But a lot of what AI required was technology that had already been developed.”

After all, it took 11 years for Google to go from launch to a billion searches per day, whereas it took ChatGPT two years to hit one billion prompts a day.

Of course, there have been compelling reasons during every bubble as to why, this time, it is not one. It is up to individual investors to make a judgement on whether they are comfortable with their AI exposure or whether it is time for a portfolio shake-up.

• Read: Are the good times about to end?
• Read: Top 10 best-selling ISA and SIPP funds this year
• Read: How does Scottish Mortgage stack up against its rivals?

Important information - investors should note that the views expressed may no longer be current and may have already been acted upon. Reference to specific securities should not be construed as a recommendation to buy or sell these securities and is included for the purposes of illustration only. Select 50 is not a personal recommendation to buy or sell a fund. Overseas investments will be affected by movements in currency exchange rates. Before investing into a fund, please read the relevant key information document which contains important information about the fund. This information is not a personal recommendation for any particular investment. If you are unsure about the suitability of an investment you should speak to one of Fidelity’s advisers or an authorised financial adviser of your choice.