The US dictionary Merriam-Webster’s word of the year for 2025 was “slop”, which it defines as “digital content of low quality that is produced, usually in quantity, by means of artificial intelligence”. The choice underlined the fact that while AI is being widely embraced, not least by corporate bosses keen to cut payroll costs, its downsides are also becoming obvious. In 2026, a reckoning with reality for AI represents a growing economic risk.
Ed Zitron, the foul-mouthed figurehead of AI scepticism, argues pretty convincingly that, as things stand, the “unit economics” of the entire industry – the cost of servicing the requests of a single customer against the price companies are able to charge them – just don’t add up. In typically colourful language, he calls them “dogshit”.
Revenues from AI are rising rapidly as more paying clients sign up but so far not by enough to cover the wild levels of investment under way: $400bn (£297bn) in 2025, with much more forecast in the next 12 months.
Another vehement sceptic, Cory Doctorow, argues: “These companies are not profitable. They can’t be profitable. They keep the lights on by soaking up hundreds of billions of dollars in other people’s money and then lighting it on fire.”
It’s not new for frontier businesses to be loss-making, sometimes for years. But moving into profitability tends to happen as costs fall. Each iteration of large language models (LLMs) has so far tended to be more expensive, burning up more data, energy and highly paid tech experts’ time.
The vast datacentres required to train and run the models are so expensive to build and kit out that in many cases they are financed by debt, secured against future revenue.
Recent analysis by Bloomberg suggested there had been $178.5bn of these datacentre credit deals in 2025 alone, with inexperienced new operators joining Wall Street firms in a “gold rush”.
Yet the precious Nvidia chips with which the datacentres are equipped have a limited shelf life, potentially shorter than that of the loan agreements.
As well as leverage – borrowing – the boom increasingly involves another bubble indicator: financial engineering, including the kinds of complex, circular funding arrangements that carry ominous echoes of past corporate crashes.
Believing generative AI will eventually produce enough revenue to match the colossal sums invested, relies – as in all bubbles – on telling big, dramatic stories about the scale of the transformation under way.
So LLMs are not just brilliant tools for analysing and synthesising large amounts of information. They’re fast approaching “superintelligence”, as OpenAI’s chief executive, Sam Altman, has it; or about to replace human friendships, according to Mark Zuckerberg.
They certainly do seem to be replacing some unfortunate human employees in specific sectors. Brian Merchant, the author of Blood in the Machine, which compares the backlash against big tech to the Luddite rebellion of the 19th century, has assembled scores of first-hand testimonies from writers, coders and marketers laid off in favour of AI-generated outputs.
Yet many of them highlight the bland quality of the work being produced by their digital replacements, or worse, the risks at play when sensitive tasks are shifted outside human control.
Indeed, the dangers of charging headlong into replacing human workers wholesale have become increasingly apparent in recent months.
In the UK, the high court issued a warning about lawyers’ use of AI after two cases in which examples of completely fictitious case law were cited.
Police officers in Heber City, Utah, learned to manually check the work of a transcription tool they were using to draft write-ups from bodycam footage after it mistakenly claimed an officer had turned into a frog. Disney’s The Princess and the Frog was playing in the background.
Specific examples such as these fail to take into account the costs of what Merchant calls the “slop layer” of AI-generated content coursing through every online space, making it harder to identify what is real or true.
Doctorow argues: “AI isn’t the bow-wave of ‘impending superintelligence.’ Nor is it going to deliver ‘humanlike intelligence.’ It’s a grab-bag of useful (sometimes very useful) tools that can sometimes make workers’ lives better, when workers get to decide how and when they’re used.”
Thought of in this way, these technologies may still have significant productivity benefits, but perhaps not quite significant enough to justify today’s toppy valuations and the tsunami of investment under way.
Any rethink would cause chaos on financial markets. As the Bank for International Settlements (BIS) recently pointed out, the “Magnificent Seven” tech stocks now account for 35% of the S&P500, up from 20% three years ago.
A share price correction would have real-world consequences far beyond Silicon Valley, rippling out to hit retail investors on both sides of the Atlantic, Asian tech exporters and the lenders, including loosely-regulated private equity firms, that bankrolled the sector’s expansion.
In the UK, the Office for Budget Responsibility (OBR) estimated in its budget forecasts, that a “global correction” scenario, in which UK and world stock prices fell 35% in the coming year, would knock 0.6% off the country’s GDP and cause a £16bn deterioration in the public finances.
That would be relatively manageable compared with the 2008 global financial crisis, in which UK institutions were leading players. But it would still be keenly felt in an economy struggling to find its feet.
So while it is perhaps understandable to anticipate a frisson of schadenfreude at the thought of big tech’s super-rich boss class being humbled, we’re all living in their world, and we would not escape the consequences.
