The year that’s coming to an end might well be the last in which the promise of AI, both in the wider economy and as a driver of stock market gains, has been relatively unambiguous.
Both 2024 and 2025 were breakthrough years for the technology, on many levels. AI-related stock valuations have boomed, and adoption in the economy has risen rapidly. Also, AI-related subjects were awarded Nobel Prizes in both years, indicating an ongoing wider societal and technological impact.
An adjustment, a coming down to earth of stock prices and a more informed and time-tested estimate of real-world impact, is likely within the next few years. We’ll also be getting a clearer picture of what AI will do to the wider economy.
Rising unemployment in the US and the UK at the end of 2025 has raised questions of whether we’re starting to see the impact of AI-related job losses.
Even if other factors are still dominant, previous warnings about a hiring crisis in entry-level positions and the beginning of AI-induced redundancies are now bound to be taken more seriously.
Many firms are now investing heavily in AI. But how they will realise growth if AI adoption also affects consumer demand through rising unemployment and lower incomes in the short to medium term is anybody’s guess.
AI boosterism and bubble warnings
In the meantime, AI boosterism, promulgated by some of the leading CEOs in the field - though not all – is taking on an ever more Alice in Wonderland quality, urging everybody to “run twice as fast” if they want to get anywhere.
Anyone who believes in the now oft-repeated adage about data centres of ‘build it and they will come’ only has to look at how that worked out for the Chinese economy, where millions of apartments stand empty, and infrastructure is seemingly permanently under-utilised.
While not all data centre construction and AI infrastructure outlays are equal, the whole sector could be put at risk by a remarkably high level of circular investment. OpenAI’s current talks with Amazon, seeking a $10 billion investment, is but the latest example.
An adjustment might come in the form of a bust after the boom, though that need not necessarily be the case. A slower and less comprehensive deflation of the balloon might also happen if the fundamentals of both the wider economy and the tech sector remain strong.
Saying that there is no bubble because everybody is now talking about a bubble recalls the whole known unknowns jumble
Still, the arguments that are now being deployed to wave away the bubble warnings are almost Rumsfeldian in their hubris, and that’s never a good sign.
Saying that there is no bubble because everybody is now talking about a bubble, hence forewarned, recalls the whole known unknowns jumble.
It is also depressingly and familiarly unaware of historical precedent. Bubbles are often foreseen. More than 300 years ago, Daniel Defoe, the author of Robinson Crusoe for those without any historical awareness whatsoever, described the infamous South Sea bubble before it popped:
“Tis plain, the Novelty of things at this time has its beginnings in the new fashioned frenzy of men's minds, I mean in their hunting after money which is done with such rage in their avarice that suffers no restraint and that knows no bounds.”
Although the frenzy is anything but ‘new fashioned’ by now, not much else has changed.
Not the South Sea, not the dot-com
That’s not to say that AI buoyance is completely the same as the South Sea bubble and indeed the much more recent dot-com bubble that burst in the early 2000s. One key difference that many point at is the solvency and relative maturity of many of the main players, such as Google, Amazon and Meta.
Clearly, there is a real and widespread use case for AI, as opposed to, let’s say, Mark Zuckerberg’s now mostly abandoned $70 billion bet on a metaverse.
Both adoption by businesses and the return on investment (ROI) show that AI growth is not about to disappear. Although a McKinsey report last month warned that only 6% of businesses saw a really significant ROI and most struggled to scale it up.
2025 stood out for the rapid adoption of ‘agentic AI’, which distinguishes itself from the only slightly older ‘reactive’ models
It’s a young technology that’s still developing, thus, the impact on companies, employment and society might not become clear for years.
In this respect 2025 stood out for the rapid adoption of ‘agentic AI’, which distinguishes itself from the only slightly older ‘reactive’ models. Agentic AI is able to anticipate and operate independently of human prompts once given an overall goal.
The impact is, if anything, even bigger than the original bursting onto the scene of the technology in 2022.
When AI does the hiring
Take the highly sensitive area of employment, though not on the redundancy but the hiring side. Estimates are that now just some 20 to 25% of job application résumés are seen by human eyes. This had already started with earlier algorithmic tools but has increased hugely with agentic AI.
Retailers leave their inventory management now to agentic AI, financial institutions their raw data analysis, and journalism increasingly the writing of articles
Retailers leave their inventory management now to agentic AI, financial institutions their raw data analysis, and journalism increasingly the writing of articles – though not this one yet (I use it for research).
Almost one in ten pieces, particularly in sports, weather and financial coverage, and especially in smaller local papers, is now produced by AI.
That’s still nothing compared to the over half of all online articles that are now at least partly AI-generated.
As with the description of business adoption and ROI above, there are significant regional differences. North America and some Asian countries are leading, while Europe is more cautious.
It’s a safe bet to predict that AI will be more widely adopted in the coming years. This is not just because of competitive pressures but also because of the ageing populations and aversion to immigration in many of the advanced economies of the world.
Someone will have to do the job, and that someone could very well be an AI. As the technology evolves, that could happen in as yet unforeseen ways.
As long as the field keeps growing and the big AI beasts can see their way to eventual profitability, it could sustain the stock markets. That is, if we find a way to keep up consumer spending at the same time.
We’d better hope so, because otherwise AI in the coming years might eat up both our jobs and our savings and pensions at the same time.
