AI will revolutionise everything. It’s the next mega Ai industry and it will usher in a new, robot-dominated industrial revolution. It will divorce growth from work and create massive wealth. That’s the line that’s been pounded down our sense of reality by drug-addled techbros and the cowardly financebros who try to make them look good.
This article has been so compelling that we have piled on debt and invested in AI with such fervor that we are now effectively betting the entire Western economy on its success. But there is only one problem: the tale just isn’t accurate, and increasingly it seems the AI industry is finding that out the hard way.
Take Microsoft. It has deployed many hundreds of billions of dollars into AI infrastructure and OpenAI since 2020. Its first product spawned from this titanic investment is Copilot, an “agentic” AI designed to help make you more productive in Windows, sort of a grown-up version of Clippy (God I miss Clippy; he was just terrible and wonderful).
But it seems no one is buying Copilot.
Microsoft has reportedly cut Copilot sales targets in half and is struggling to attract buyers, according to The Information. One place this can be seen is in Microsoft’s financials, where they’re barely making any money at all from their massive AI push.
Microsoft disputed the Information’s reporting, saying it conflated sales and growth quotas. But that is akin to saying “It’s not horse shit on my face — it’s bull shit.” No matter how you slap a label on it, a setback is horrifically bad in front of either target! That is, it suggests that the product they have invested hundreds of billions of dollars in is a dud!
But why are Copilot’s growth and/or sales so terrible? Well, one set of researchers put these ‘agentic’ AIs, including Copilot , to the test and found that, more often than not, they flat-out failed even basic tasks 70% of the time — essentially making them somewhere between useless and a liability.
The same is true of Microsoft’s cash baby, ChatGPT. It’s also frequently functionally useless, as with Copilot. Some researchers have found that newer models are ‘hallucinating’ in 33 percent to 79 percent of simple tasks. Even the very latest ChatGPT-5 — which is, like, hilariously computationally intensive and blindingly expensive to run — doesn’t ever try it! — hallucinates 10% of the time on rudimentary tasks. Yeah, that’s enough to make it worthless, as you will spend more time running for examples rather than just fact-checking the machine.
So it is no wonder that studies show that only 5 percent of ChatGPT users pay for the service. But OpenAI has to undercharge these customers; they are still losing serious money on each paying customer! And yes, for those $200-per-month plans. That total customer revenue of zero is a big part of why OpenAI is on pace to report an annual loss of tens of billions of dollars in 2025.
And this cautionary tale of zero customer demand is not unique to Microsoft and OpenAI; it applies across the entire generative AI market. But guess what? It is set to get even worse.
With all this talk of hyperscaling and hypergrowth, there are, small as they might be, at least a few more AI customers today than in the past.
But contrary to that idea, spending on and use of AI are actually declining — at least in corporate America.
The number of Americans using A.I. to do this was 11 percent in October, The Economist reported. That’s down from 12% in their last survey. That may not seem like much, or perhaps within the margin of statistical error, but it is consistent with other patterns in this data. The same survey indicated that the percentage of firms with 100 to 249 employees who said they do not use AI was 74.1% in March, but has yielded negative results for hires who do not use any! For larger companies with 250 or more workers, that figure now stands at 68.6%, up from 62.4% in February.
And corporate A.I. projects are being canceled at an all-time high. In 2024, only 17% were scrapped, but the percentage has ballooned to 42%!
None of this should come as much of a surprise. After all, the widely known MIT report states that 95% of AI pilots yielded no meaningful results. Other reports have since come out that support the findings of this one, like BCG’s report, which found that across companies only 5% were seeing a return on their AI investment, and Forrester Research’s corporate survey responses, which showed that 15% increased profit margins over the last year thanks to AI.
And that AI slump is poised to worsen in the coming year. After conducting that survey, Forrester Research analysed the responses and predicted that in 2026, companies will delay about 25% of their planned AI spending by a year. In other words, because AI is so damn useless, more than a quarter of the corporate money previously predicted to be spent on it in 2026 will be held back!
This is compounded by the fact that most of the AI revenue for outfits like OpenAI comes from corporate, not private, customers. After all, that’s where AI is meant to deliver its massive economic dividends, so if any of this were true, corporate AI spending would be through the roof! But it isn’t true. Corporations are seeing how shitty and worthless this technology is and are beginning to shun it.
That is awful news, given that the only thing holding up the entire AI bubble at this point is a tsunami of debt (link to read about it here). The AI industry is founded on a situation in which it desperately needs corporate spending on AI to explode, and soon, to cover the debts it is raising without signing off as ready for financial stability (the first quarter was supposed to be unauditable)! In other words, this economy-wide corporate sea change in attitudes toward AI is the canary in the coal mine. It’s a reminder that the propaganda fueling this bubble is falling apart as reality intrudes. It’s a harbinger that the financial booby trap Big Tech has rigged is about to blow up in its collective face.
