The breakneck expansion of artificial intelligence is heading for a stark reality check in 2026, posing a significant and growing risk to the global economy. Despite soaring revenues, the astronomical costs of developing and running AI systems are creating what sceptics call unsustainable 'dogshit economics', fuelled by hundreds of billions in debt.
The Unsustainable Maths of AI Investment
Revenues from AI are climbing fast as more clients sign up, but they are failing to keep pace with the extraordinary levels of capital being burned. Investment in the sector reached a staggering $400bn (£297bn) in 2025, with even more forecast for the coming year. The fundamental problem, according to prominent AI critic Ed Zitron, is that the 'unit economics' – the cost of serving a single customer versus the price charged – simply do not stack up.
This is not typical for a frontier technology. While new industries often operate at a loss, profitability usually emerges as costs fall. For AI, the opposite is happening. Each new generation of large language models (LLMs) tends to be more expensive, consuming more data, energy, and highly-paid expert time.
The infrastructure itself is a major financial burden. The vast data centres needed to train and run AI models are so costly that many are built using debt secured against future, hoped-for revenue. Bloomberg analysis identified $178.5bn of data centre credit deals in 2025 alone, with a 'gold rush' involving new operators and Wall Street firms.
Debt, 'Slop', and a Looming Correction
This leverage is compounded by another classic bubble indicator: complex financial engineering. The situation is made more precarious by the limited shelf-life of the precious Nvidia chips that power these centres, which may become obsolete faster than the loans used to buy them are repaid.
The belief that generative AI will eventually justify its colossal price tag relies on grand narratives about imminent 'superintelligence' or AI companions replacing human friendship. Yet the current output often tells a different story. Merriam-Webster's 2025 word of the year, 'slop', defined as low-quality AI-generated content, highlights a widespread issue.
Author Brian Merchant, who compares the tech backlash to the Luddite movement, has collected testimony from writers, coders, and marketers replaced by AI. They frequently cite the bland or risky quality of the automated work. Real-world stumbles are becoming common, from UK lawyers citing fake AI-generated case law to a US police transcription tool absurdly claiming an officer turned into a frog.
As critic Cory Doctorow argues, AI is best seen as a 'grab-bag of useful tools' rather than a path to superintelligence. Viewed this way, it may boost productivity, but perhaps not enough to justify today's sky-high valuations.
Why the UK Economy Would Feel the Shock
A major rethink of AI's value would trigger chaos on financial markets. The Bank for International Settlements notes that the 'Magnificent Seven' tech stocks now make up 35% of the S&P500, up from 20% three years ago. A sharp correction would ripple far beyond Silicon Valley.
The UK's Office for Budget Responsibility has modelled a 'global correction' scenario where stock prices fall 35%. This would knock 0.6% off UK GDP and worsen public finances by £16bn. While less severe than the 2008 crisis, the impact would be keenly felt in an already fragile economy, affecting retail investors, lenders, and Asian tech exporters alike.
The message is clear: while some may relish the thought of hubristic tech giants being humbled, the economic aftershocks of an AI reckoning would be felt by everyone. The world is now living in their debt-fuelled world, and no one would escape the consequences of a pop.
