Will the AI boom ever end?

The US stockmarket is starting to show signs of fatigue with Artificial Intelligence (AI). But the underpinnings of this revolution remain fundamentally sound. Investors’ main focus should be long-term equity returns and effective diversification.

25 November 2025

Investors have been giving mixed signals in response to the US Technology sector’s Q3 earnings updates, with Meta down -11%* on 30 October alone and volatility affecting all the technology megacaps in the wake of their latest updates. While reported earnings for the so-called AI superscalers were generally strong in themselves, investor consensus on the outlook may be starting to fracture.

The problem was neatly encapsulated by Microsoft’s Chief Financial Officer, Amy Hood, who declared that “When you see these kinds of demand signals, and we know we’re behind, we do need to spend.”¹ Big tech can’t build data centres and cloud services fast enough but Meta again provided evidence of further doubt that market patience on this matter was infinite. By early November its bond yields were trading nearly 1% higher* than equivalent risk-free US Treasuries in a sign that lenders wanted to be compensated more for the perceived financial risks of extending credit to the Facebook owners. A tech firm that never used to borrow money is now subject to the same bond vigilante action as an indebted G7 economy or wayward emerging market one. But the issue is not specific to just this name. The debate comes at a difficult time in terms of market valuation for large-cap technology firms generally. The tech-heavy Nasdaq-100 Index now trades at a nosebleed-inducing 31x forward earnings. It is up 20% year-to-date to 11 November compared with ‘just’ 15.6% for the broader-based (but still one third tech) S&P 500 and 8.7%* for a sector equal-weighted version of the same index. To say that technology is influencing the US stockmarket - and by inference the global one given its 70%* weighting of the MSCI AC World - would be an understatement. It is dominating it. Not for nothing is there increasing chatter about an ‘AI Bubble’. According to Bloomberg, no fewer than 140 stories were written containing the words ‘AI Bubble’ on the day that AMD announced a new partnership with OpenAI to share technical expertise. As of 7 November, the story count stood at over 600. This matters because as Robert Shiller wrote in his seminal work Narrative Economics, evolving narratives can in themselves alter economic and market behaviour.

Talking ourselves into a sell-off?

From 30 Dec 2022 to 7 Nov 2025

Fundamentally underpinned

On this basis, caution around AI in particular and technology stocks in general, would seem warranted. But AI’s impact on the real economy is surely a sound indicator of its intrinsic worth and the evidence points to it starting to have an effect. A recent study by Microsoft Research identified copywriting, teaching, sales, translation and customer service all being transformed by AI. While potentially worrying (or empowering) for those directly involved, AI disruption should nonetheless lead to strong productivity gains at the corporate and macroeconomic level. So much so that Vanguard believes that by 2035, AI integration into the world of work could increase productivity by 20%, raising annual US GDP growth to 3% by the 2030s². Policymakers appear to be in little doubt that this will be the outcome and are starting to address real concerns about job elimination and inequality. Such is the perceived inevitability of AI’s transformation that the Berggruen Institute recently urged rapid action to address the inequality gap that will arise between those who own and manage AI technology and those who remain on the sidelines. ‘Pre-distribution’ has become the policy watchword, describing a democratising move to mass ownership in the capital that will drive the AI revolution, for example through Money Accounts for Growth and Advancement in America, Junior ISAs in the UK (although these are not government funded in the way the previous Child Trust Funds were) and potentially also a European Sovereignty Fund. In this way, the hope is that the spoils of AI don’t continue accruing to the few. In government circles therefore the socio-economic transformation associated with AI is becoming a vexed policy question, on top of an already active debate around the unevenly distributed ‘K-shaped’ economy in the wake of the pandemic. Few are seriously doubting that the revolution will transform society.

Ask the right question

The unknowable therefore is less whether AI is intrinsically going to change the world as how long and how far stockmarket momentum can keep going. On this point there are no easy answers. In a recent interview while on a visit to JPMorgan’s Bournemouth offices in the UK, CEO Jamie Dimon expressed concern about stockmarket valuations but simultaneously would not be drawn on the timing of any adjustment, instead indicating a timeframe of “In the next six months to two years”. In a sense, such information is useless for sensible investors because getting the full benefit from the stockmarket’s returns is only reliably achieved through long term buy-and-hold anyway. The observed rate of return from the stockmarket of around 7% (see Stocks for the Long Run by Jeremy Siegel and also The Rate of Return on Everything 1870-2015 by the San Francisco Fed) is only available for those who remain invested. After all, these long-run rates of return have seen it all – war, disease, bubbles. The better question to ask might be how to prepare for a correction without compromising participation in the market’s long-term trajectory. Multi-asset portfolios offer a compelling answer via diversification. Rather than agonising over whether now is the right time to underweight equities, the US and technology stocks versus a given index, investors should focus their intellectual bandwidth on those diversifying assets whose role it is to hold the line if and when markets stumble, as they periodically do. On the basis that if something can’t be explained easily it’s not worth holding, multi-asset portfolio diversifiers need only fill three criteria – transparency, independence (from equities) and reliability. Examples of such assets could include inter alia: gold, money markets, short-dated investment grade bonds, well-researched European financial debt, carefully selected climate bonds and well risk-controlled long/short strategies. To illustrate the simplicity argument with an extreme example, a simple comparison of a 60:40 (equities:bonds) portfolio versus the Lipper and ARC peer groups reveals how ‘less is more’ is not just a mantra but a real-world observation.

It's (not) complicated:

Performance from 30 Apr 2007 (inception) to 31 Oct 2025

All of this should be cause for constructive optimism rather than concern for the future. Sticking to a globally benchmarked equity allocation without needlessly over-emphasising US or technology stocks, combined with a simple, repeatable diversification allocation is probably the closest to having your cake and eating it anyone can get in investment terms. If ‘baked’ competently (see your professional advisor) it should ensure continuing participation in all the innovation and ingenuity corporations capture over time - including AI - while offering a degree of smoothing for the inevitable upsets and reversals on the way. No one knows when (or indeed if) the stockmarket will ever cease to reflect the transformational nature of the AI revolution, but by introducing some traditional multi-asset resilience into their portfolios, investors can avoid getting too hung up on the question in the first place.

Julian Howard is Chief Multi-Asset Investment Strategist at GAM Investments. This article represents the views of GAM’s Multi-Asset team.