TAmerican dictionary Merriam-Webster's word of the year for 2025 is “slop,” which the dictionary defines as “low-quality digital content, typically produced in large quantities by artificial intelligence.” This choice underscores the fact that while AI is widely accepted, especially among business executives keen to reduce labor costs, its downsides are also becoming clearer. In 2026, economic risks are increasing given the reality of AI.
Ed Zitron, the most vocal poster child of AI skepticism, makes a pretty convincing case for the “unit economics” of the industry as it stands: the cost of fulfilling a single customer request is not enough for the price a company can charge. In typical colorful language, he calls them “shitholes.”
Revenues from AI are increasing rapidly as more paying clients sign up, but so far it hasn't been enough to cover the breakneck level of investment underway. It is predicted to reach $400bn (£297bn) in 2025 and rise even more in the next 12 months.
Another ardent skeptic, Cory Doctorow, argues, “These companies aren't making money. They can't be. They're keeping the lights on by siphoning hundreds of billions of dollars of other people's money and lighting it on fire.”
It is not uncommon for frontier companies to continue to lose money, sometimes for years. However, profitability tends to increase as costs decrease. So far, each iteration of large-scale language models (LLMs) tends to be more expensive, consuming more data, energy, and time of highly paid technical experts.
The vast data centers required to train and run models are very expensive to build and provision, and are often financed with debt for future revenues.
A recent analysis by Bloomberg suggests that these data center credit transactions will reach $178.5 billion in 2025 alone, creating a “gold rush” of inexperienced new players into Wall Street firms.
But the valuable Nvidia chips in your data center have a limited lifespan, which could be shorter than your loan agreement.
In addition to leverage, another bubble indicator is increasingly involved in this boom: It is financial engineering involving complex, circular financing arrangements with eerie echoes of corporate failures of the past.
As with all bubbles, believing that generative AI will eventually generate enough returns to match the huge investments depends on telling a big, dramatic story about the scale of the transformation underway.
Therefore, LLM is not only a great tool for analyzing and synthesizing large amounts of information; They are rapidly approaching “superintelligence,” as OpenAI CEO Sam Altman puts it. According to Mark Zuckerberg, it is replacing human friendship.
They certainly seem to be replacing unfortunate human employees in certain areas. Brian Merchant, author of Blood in the Machine, which likens the backlash against big technology to the Luddite revolt of the 19th century, has collected numerous first-hand accounts from writers, programmers, and marketers who were fired for championing AI-generated products.
However, many of them highlight the bland quality of works produced by digital substitutes, or worse, the risks that arise when sensitive tasks are moved outside of human control.
Indeed, the dangers of going head-on with replacing human workers have become increasingly clear in recent months.
In the UK, the High Court warned against the use of AI by lawyers after two cases cited completely fictitious precedents.
A Heber City, Utah, police officer learned how to manually check the transcription tool he was using to create stories from body camera footage after it falsely claimed officers had turned into frogs. Disney's “The Princess and the Frog'' was playing in the background.
These specific examples do not take into account the cost of what merchants call the “underlayer” of AI-generated content that flows through all online spaces, making it difficult to identify what is real and true.
Doctorow argues that “AI is not an 'imminent wave of superintelligence,' nor will it provide 'human-like intelligence.' AI is a collection of useful (sometimes very useful) tools that can sometimes make workers' lives better if they decide how and when to use them.”
When you think about it this way, these technologies may still have significant productivity benefits, but they're probably not important enough to justify today's top valuations and ongoing tsunami of investment.
A second thought would cause chaos in financial markets. As the Bank for International Settlements (BIS) recently pointed out, the “Magnificent Seven” of tech stocks now account for 35% of the S&P 500, up from 20% three years ago.
The correction in stock prices will have real-world implications far beyond Silicon Valley, with repercussions for retail investors on both sides of the Atlantic, Asian technology exporters, and financiers, including the lightly regulated private equity firms that financed 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 global stock prices fell by 35% next year would reduce the country's GDP by 0.6% and cause a £16bn deterioration in public finances.
This will be relatively manageable compared to the 2008 global financial crisis, in which British financial institutions played a leading role. But it will still be keenly felt as the economy struggles to get back on its feet.
So while it's probably understandable to expect a Schadenfreude rage given the debasement of the ultra-wealthy boss class of big tech companies, we all live in their world and cannot escape the consequences.
