AI bubble talk will ‘not stop the underlying shift in how work gets done’
Earlier in the summer, Telegraph columnist Russ Mould made waves, when he compared the current state of investments in the AI market, to the conditions which resulted in the Great Depression. But according to Adam Hofmann, Elixirr’s partner in AI and people transformation, critics may be pointing to “the wrong bubble” when it comes to the technology’s future.
Russ Mould is the Telegraph’s ‘Questor’ lead columnist for investment, giving an insight into the inner thoughts and workings of the City. In a piece in July, the equity analyst and investment director at AJ Bell became hinted one of the global market’s most reliable long-term indicators “is flashing red”.
Amid excited investment in AI technologies, Mould noted in his piece that the overvaluation of US stocks has passed the level that brought the stock market to its knees to kick off the Great Depression. With US stocks on the S&P 500 reportedly priced 41 times their average earnings over the last decade, Mould cited a measure called the Shiller CAPE ratio, which is now at its widest for US equities since the peak of the technology, media and telecoms (TMT) bubble of 2000 – and also exceeds the valuations reached at the highs of 1929 and 1901, both of which preceded seismic crashes.
He argues, “Optimists will assert that productivity gains thanks to AI more than justify the assumption of an era of premium growth. But the 40-year Cagr for S&P 500 earnings per share is just 6.7pc. That longer period encompasses not just the TMT profits boom of 1998 to 2000 but also the long-term economic benefits of the broadband, wireless telecoms and internet build-out that have exceeded even the wildest dreams of investors and analysts at the turn of the century.”
So, is the situation really as grave as it appears? According to Elixirr expert Adam Hofmann, the situation is not necessarily so black-and-white.
“I think there is a bubble here,” began the AI and people transformation partner. “But it isn’t the one everyone is pointing at, and the difference matters enormously. Indicators like CAPE put the focus on the multiple, and that’s the wrong place to look. A multiple only looks excessive if the demand behind it doesn’t show up. Get that right, and today’s pricing is far more defensible, and prices adjust quickly regardless.”
A matter of demand
The real question, Hofmann believes, was never what people are paying, but “whether the demand assumption underneath holds.”
“The actual bubble sits in how the demand is financed,” he continued. “Follow the money and it loops back on itself: chipmakers sell to cloud providers, cloud providers sell capacity to AI labs, and those labs are often funded by investment arms tied to the same cloud providers. It’s revenue and capital circulating within a closed group, not flowing in from outside. Markets read that internal momentum as genuine market pull, and it’s why the comparisons to 1999 keep coming. But the lesson of the dotcom crash wasn’t that the internet failed. The fibre laid back then is what your streaming runs over today. The economics just never supported the pace of the build.”
In 2026, unlike 1999 or 1929, Hofmann believes “the companies driving this build are profitable, rather than debt-financed”, meaning “if it corrects, it looks more like a repricing of good businesses than a collapse of bad ones”.
He concluded, “That’s the risk now: not that AI fails, but that the financing gets exposed before the demand fully lands. Every major technology follows the same arc: we overestimate what it does in two years and badly underestimate what it does in 10. What you’re seeing in the spending numbers now is just the opening act. The part that actually matters, where businesses get rebuilt around the technology, plays out over the following decade or more. Conflating the two is the most expensive analytical mistake being made right now. A correction would hurt investors and reset some valuations. It would not stop the underlying shift in how work gets done.”
