AI Will Destroy Everything

But not in the way you think it will.

Photo by hao wang on Unsplash

If you take the tech bros seriously, we will all be suffering in a Terminator-esque AI dystopia in just a few years because their AI models will be so damn good. And why shouldn’t we believe them? It isn’t like they have a huge vested interest in AI being perceived as an existential threat… Meanwhile, back in the real world, AI poses a gargantuan threat to us all; it’s just more like 2008 and 1933 rather than some big Hollywood production. You see, AI critics like myself have been comparing the AI bubble to the infamous dot-com bubble of the late ’90s, but we have been sorely wrong. It is far, far worse. The dot-com crash will seem like a picnic in the park in comparison to what is to come.

There is a good chance that many of you reading this weren’t even born when the dot-com bubble happened (my god, am I old!), so let’s quickly recap.

The dot-com bubble occurred in the late 1990s because e-commerce suddenly became viable. Vast swathes of the population had a personal computer and a slow dial-up internet connection (which involved accessing the internet via landline phone cables). At the time, capital was cheap too. So investors piled money into internet businesses, hoping to scoop up large portions of the market, as well as building infrastructure, such as much faster fibre optic broadband networks. This drove the speculative values of businesses behind this boom through the roof! But by the end of the year 2000, the bubble burst, these businesses’ valuations tanked, many went bankrupt, millions were laid off, the economy was in tatters, and a recession ensued.

Why?

Investors were keen to jump on this new market. They believed that early movers would dominate their respective sectors and that commerce would rapidly migrate to mostly online markets. So they speculatively valued small upstarts at astronomical levels, creating the bubble.

This bubble was then primed to pop by overinvestment in infrastructure and a lack of profitability. The adoption of the internet and e-commerce was significantly slower than predicted, and thus, the billions of dollars invested in building faster networks were largely wasted. Four years after the bubble burst, 85% to 95% of the fibre optic internet networks laid in the 1990s remained unused. Even the largest internet businesses, such as Pets.com, were losing hundreds of millions of dollars a year. To attract customers, they had to lower prices, and the cost of expanding their capacity was crippling, making profitability impossible. Because capital was cheap, many used loans backed against their extortionate valuations just to keep the lights on.

The pin that popped the bubble was a shift in the economy. Recession fears shook the world after Japan’s economy faltered, and to prevent inflation from gripping the US, interest rates were hiked. Fear of a recession meant investors’ appetite for risk reduced, as they looked for a safe place for their money to weather the storm, so speculative internet investments were less attractive. But the rising interest rates suddenly made borrowing money expensive, and as such, internet businesses couldn’t raise the funds they needed to pay for their losses. In one fell swoop, investors pulled their cash, major players declared bankruptcy, and the entire bubble collapsed.

If you have been paying attention to the generative AI industry, you can already spot the similarities. Now, we know this isn’t just similar but nearly identical to the dot-com bubble. From the investor hype to the infrastructure overspend to the lack of profitability to the economic forces that pricked the bubble, all the signs are there. Except this time, the bubble is so much bigger, and the damage it will cause when it inevitably bursts will be biblical. Let me explain.

I will cover how inflated AI stocks are in a minute. For now, let’s start with the surprising fact that no sector of the AI industry is profitable, from its infrastructure to operators to adopters — everyone along the chain is losing money.

A recent MIT study found that 95% of AI pilot projects fail to yield meaningful results. For the 5% where AI yielded results, the improvements were marginal, and the AI was used to augment highly constrained back-office jobs, not replace workers as promised. A METR report found that AI coding tools, which are meant to be the most promising application for generative AI, actually slow developers down. The inaccuracy of these models means they repeatedly make very bizarre coding bugs that are highly arduous to find and correct. As such, it is quicker and cheaper to get a developer to code it themselves. Research has even found that for 77% of workers, AI has increased their workload and not their productivity, leading to burnout and decreased performance.

In other words, all the data points to the fact that there is actually very little upside to a business adopting AI. Personal users are also struggling with adopting AI, as it seems to be highly damaging to their mental health, leading to a dramatic uptick in reports of “AI Psychosis”. It isn’t much of a surprise that AI adoption has lagged behind predictions for years now (with more on that in a bit).

And it isn’t like AI operators, such as OpenAI, are making bank either. In fact, they are losing money hand over fist! Despite having by far the largest income of any AI company, OpenAI still loses a dramatic amount of money for every one of its $200-a-month plans. Their models are just too expensive to build and run. In fact, there are suggestions that OpenAI would have to charge ten times price that just to break even.

But OpenAI is only valuable because investors think it could one day make “agentic AI”, which is “smart” enough to do complex tasks reliably and autonomously. However, to improve their currently unreliable models, let alone reach this agentic myth, OpenAI needs to invest exponentially more money in them — so, they are actually getting further away from profitability! Analysts have found that OpenAI is set to post a loss of over $14 billion in 2026, which is a larger loss than many of the banks that folded in 2008. OpenAI’s own numbers, which use unbelievably optimistic AI adoption rates, suggest that by 2029, it will post losses in the hundreds of billions of dollars (read more here). Every other generative AI business is facing the same problem. They really aren’t profitable at all.

But remember how one of the big issues with the dot-com bubble was overinvestment in infrastructure? Well, AI has that too, and those building AI infrastructure, namely data centres, are also losing money hand over fist!

A recent report found that the AI industry will need to generate $2 trillion in annual revenue just to pay for the data centres they plan to build by 2030. The entire global AI market is expected to generate $390.91 billion in revenue by the end of 2025, meaning the AI industry needs to grow by 512% over the next five years just to break even on its current planned infrastructure build-out. But thanks to all the reasons we just covered, the AI industry growth is slowing. AI adoption rates have fallen short of predictions for two years running now. Consequently, the AI industry grew by roughly 50% from 2023 to 2024 but fell to just 40% growth from 2024 to 2025. This means the predicted demand these data centres were being built for simply isn’t there, just as with the fibre optic networks in the 1990s. Even optimistic predictions suggest that the AI industry can’t raise the $2 trillion annual revenue needed just to pay for its infrastructure build-out, and these data centres will run at an $800 billion loss per year.

However, unlike fibre optic networks, ‘macro-scale’ economics show that data centres are an utter money pit, even if the demand exists. Praetorian Capital CIO Harris Kupperman recently revealed that the AI data centres being built today will suffer $40 billion of annual depreciation while generating somewhere between $15 and $20 billion of revenue. And right now, there is a major data centre shortage! So each dollar invested in data centres, even when demand is sky high, returns a net loss of 50% — 62.5%! In other words, for this infrastructure to be even remotely sustainable, its cost of use has to more than double, which will make AI companies’ profitability issue even worse!

Yet, despite all of these very troubling issues, speculative investment in AI companies is out of control. OpenAI is now valued at nearly double that of Samsung, despite OpenAI’s revenue being 16 times smaller than Samsung’s and the fact that Samsung is legitimately profitable. These valuations are so inflated that even Sam Altman has had to admit that the AI industry operates in a bubble.

But AI advocates say all of this will be resolved. Soon, we will have “agentic AI”, which will address the significant issues with current models, driving AI adoption to new heights as large portions of the workforce are replaced by it. This will then significantly lower costs for businesses and generate substantial revenue for AI companies, making AI operators and data centres profitable. Sure, it might not be profitable to build AI infrastructure, develop AI or use AI right now, but it will be soon. Sadly, that is a whopping, barefaced lie.

All the science, trends and the technical realities of AI show that agentic AI is a total myth. It cannot be built.

Take the efficient compute frontier. This has been an immutable law of AI development since day one. This boundary describes how the maths involved in AI training has a serious problem with diminishing returns, meaning that if you want AI to develop at a linear rate, it requires exponentially more data, energy, resources and cash. But because that isn’t possible, it means that today’s AI models are about as good as they will ever be, given that even colossal investments into their development will only yield marginal improvements. We can already see this with ChatGPT-4 and 5, which, despite OpenAI significantly increasing the model size and training time, have only delivered very minor improvements.

Scaling AI up to make it agentic will increase costs exponentially. So even if the likes of OpenAI manage to do it, it will only push them further away from profitability.

But here’s the wild thing: we know that AI will never become agentic.

Take the Floridi conjecture, which explains that the maths that powers AI means that AI systems can either have great scope but no certainty or a constrained scope and great certainty. Crucially, Floridi’s Conjecture states that an AI absolutely can’t have both a great scope and great certainty (read more here). In other words, because an agentic AI is inherently broad in scope (they need to account for a lot to complete even simple tasks reliably; just look at Tesla FSD), it can never have high certainty, which is another way of saying it can’t be reliable. Therefore,it can never be truly autonomous and will always require human oversight, which makes the whole effort moot.

There is also the issue of the fact that for an AI to be “agentic”, it needs to be cognitive. It needs to conceptually understand the task at hand so that it can correct its own errors or adapt if something unexpected happens. Only then can it complete tasks, even simple tasks, autonomously. But we know that AI can’t do this. All it can do is replicate patterns — it isn’t cognitive, and it doesn’t “understand” what it is doing. Iris van Rooij, a professor of computational cognitive science at Radboud University, recently published a paper detailing this issue. They found that even if AI engineers have unlimited computation power and data, they could never produce a cognitive AI. This is because the fundamental programming structures underlying AI, such as neural networks (a terrible and misleading name) and transformers (the “T” in “ChatGPT”), are just statistical models capable of replicating basic patterns. No matter how much we try to scale them up, they won’t magically gain the ability to cognitively think, understand the world, or create novel thought.

So, not only do we know that “agentic AI” wouldn’t solve the profitability problem, but it is also impossible to build. “Agentic AI” is a myth, and sadly, that means that so too is the idea that the AI industry in its current guise can ever be profitable.

Just like the dot-com bubble, infrastructure build-out is vastly outstripping the rise in demand, AI businesses are valued miles above where they should be, and the profitability of both AI infrastructure and AI is an complete pipedream. But, back in the ’90s, the lack of profitability stemmed from the industry’s attempts to rapidly expand, not the technology itself, allowing the industry to grow gradually and flourish after the collapse. Meanwhile, with AI, the lack of profitability is built into the technology. So, when this bubble pops, the damage will last a lot longer, as there will be no correcting and bouncing back.

Just two factors are missing from our comparison, the factors that eventually popped the bubble. Back in the ’90s, that was the dual prong of inflation driving interest rate hikes and the risk of a recession. This is where many criticise this comparison, as interest rates are currently being cut, not raised, and the economy is ticking upwards, so there is no sign of a recession. While no one is doubting that the AI bubble exists, many don’t think it will pop in the same way or have the same impact as the dot-com bubble.

But, it’s looking more and more like they are wrong.

Yes, interest rates are being cut right now, but that could end very soon. The inflation rate in the US is 2.9%, which is nearly 50% higher than it should be. It is likely going to get worse too, as tariffs, trade wars, international boycotts, and war-driven supply chain disruptions set in, which will all almost certainly push inflation rates higher. How can a government stop an inflation rate this high? They have two main tools: increasing taxes or hiking interest rates. Guess which one the right-leaning Western world will choose?

So, that is a tick for upcoming interest rate hikes.

But what about a recession?

Well, there is a reason the US is still cutting interest rates despite inflation getting out of control. Governments slash interest rates to stimulate investment in the economy and avoid recessions. And the US is a lot closer to a recession than you think.

Deutsche Bank recently discovered that, even with these interest rate cuts, the AI bubble is the only thing helping the US economy avoid a recession. Essentially, Deutsche Bank found that the AI industry’s utterly gargantuan investment in data centres is artificially inflating the US’s GDP. In fact, around half the current growth of index funds like the S&P 500 has been driven by the tech companies expending effort into data centres. It isn’t normal for a single industry to do that! If you exclude this expenditure from the US’s GDP, its growth has actually dropped off rapidly to near zero, which is a key warning sign of an upcoming recession.

So, there is a looming recession, too — it is just being hidden by how insanely large this bubble has grown.

Effectively, the pin has already popped the AI bubble. It is just so big that it hasn’t noticed yet.

As a side note, this is why it “feels” like we are in a recession when the markets say we aren’t. This cutting of interest rates isn’t stimulating our economy; instead, it is just inflating this bubble owned by the elite. For example, one report found that over 70% of US venture capital is going towards AI this year. This “cheap” money is overwhelmingly flowing into a dead-end financial bubble; meanwhile, the rest of us are actually facing recession-like economic conditions. This is just a sad continuation of “socialism for the rich and capitalism for the poor”, but that is a conversation for another day.

This is where the difference between the dot-com bubble and the AI bubble becomes apparent.

At its peak, the dot-com bubble only accounted for 15.6% of US GDP growth. Yet when this bubble burst, the economy faltered, causing a recession, and millions lost their jobs.

By comparison, the AI bubble accounts for nearly 75% of GDP growth.

When it pops, the damage will be unprecedented. Not just because this bubble is arguably five times bigger than the dot-com bubble, and its recession will likely be just as large, but because of our current economic climate. Sure, a big recession is bad, but we are living through a second wave of fascism and a new robber baron era. How do you think Trump will react to this pop? Will the government bail out big tech and gain control over it? Will they “fix” the economy so that it only works for the 1%? Will they use it as an excuse to undermine workers’ rights under the guise of “helping the economy”? All three, and more horrors, are wildly likely.

So when I say the AI bubble will destroy everything, I mean it. Our way of life, society’s very structure, the economy functioning for everyone, and how we understand the world works will all be torn down. A false sacrifice to appease the economic gods and subjugate ourselves to the new ruling class.

Brace for impact, I guess?

---

Thanks for reading! Don’t forget to check out my YouTube channel for more from me, or Subscribe. Oh, and don’t forget to hit the share button below to get the word out!

---

Sources: Fortune, Tech Spot, Will Lockett, Will Lockett, FRB, PRN, Bain & Company, GVR, NNG, Reuters, UN, World Bank Group, CMC, Will Lockett, RU, Reuters, Reuters, Trade Economics, Statista