Disclaimer: This is not a bearish take on AI
Apparently, the word of last year was “AI Bubble”. It feels as though the frequency with which I encounter the phrase has increased exponentially. Likely guess for the word of 2026: AI megabubble.
But why has the discourse, whether on main street or the wall street, become so trapped within the language of the so called bubble? The concept has become something of a discursive dead-end.
There is a simple answer why. To explain, I will have recourse to one of the epistemological inventions of the great Nassim Taleb: Wittgenstein’s Ruler. The concept alludes to this: whenever you use something to measure another something, the measurement you get might say more about the measurement than the thing measured, especially if it is something domain-independent.
The ONLY test of intelligence that counts is to see if the person detects Wittgenstein's ruler situations: 1) When you use a ruler to measure the table, you are also using the table to measure the ruler. 2) The more unexpected the measurement, the more you apply W's ruler.
“Wittgenstein’s ruler: Unless you have confidence in the ruler’s reliability, if you use a ruler to measure a table you may also be using the table to measure the ruler. The less you trust a ruler’s reliability (in probability called the prior), the more information you are getting about the ruler and the less about the table.”1
You might anticipate where this is going.
It is a fact that AI has revolutionary potential to disrupt the way things are in our society. It is also a fact that AI-spending commands never-seen-before sums of money in the economy. Naturally, it is difficult to justify unprecedented sums being thrown at an unprecedented technology. That is hardly anything new.
What doesn’t follow from all these premises is this: prevailing habit of attributing the entire magnitude of the transformations now unfolding to AI itself. The “bubble” in question therefore, is not called a bubble due to the inherent potential of AI and the capex spending it brings about. It is called a bubble because of a gross misattribution. People attribute bubble qualities to AI because it is the next big thing.
AI can finally trigger an overhaul, for good or for worse, of the internal contradictions embedded within our economic system. The centre cannot hold. Under such conditions, almost any technology promising creative destruction on a grand scale could have been labelled a bubble. This is what I mean when I say AI bubble is not about AI.
Let me unpack what I mean by this by way of some concrete examples. The examples abound, here it is suffice to include a handful.
The purported financial AI bubble
It is well in order to worry about a financial bubble caused by exuberant AI spending, that is understandable. What I find intellectually disingenuous is this: Why is it that the same circles rarely ask how the actors behind these capital allocations came to place themselves in a position where they could wager seemingly unlimited sums on AI in the first place?
In 2026, the world put over $2.5 trillion into AI in a single year — split across three fronts: the chips, the data centers, and the labs racing to AGI. Nvidia sells more AI silicon per year than most countries' entire GDP. OpenAI and Anthropic raised $217 billion between them in
I am talking about the hyperscalers: the Amazons and Oracles, the chipmakers and CoreWeaves, the asset managers such as BlackRocks of the world. How did they end up in a position where they basically were already sitting on infinite amounts of cash?
Without accounting for the financial genealogy of these actors’ balance sheets, it is methodologically faulty to ascribe what is happening in the capital markets to an “AI bubble”. The innocuous-sounding terminology shrouds our recent economic history, making it harder to see how Big Tech acquired the economic and political power to drive such a vast reallocation of society’s resources.
The detailed story of how this came about would occupy volumes. I will therefore attempt to grossly oversimplify the story. For the purpose of the central argument, it is enough to recognise that the AI boom was made structurally possible by developments that long predate AI itself.
The first part of the story has to with the gradual privatisation of the internet by a handful of private companies. This is the origin of the phenomenon, defined as Technofeudalism by Yanis Varoufakis.
What happened to the internet is eerily reminiscent of the historical enclosures movement, through which commonly held land was consolidated into private property. This was one of the driving engines that helped conjure the capitalist mode of production into existence as per Marx.
An analogous enclosure of the online spaces occurred. Digital spaces that had once appeared open and decentralised were gradually consolidated into privately owned platforms. The companies controlling these platforms accumulated what Varoufakis calls cloud capital and acquired the ability to extract cloud rent from the users of the internet—that is, from us.
The second part of the story has to do with the financial infrastructure itself. In essence, after the 2008 Global Financial Crisis, a new-monetary regime was imposed on the world. The Federal state rescued the financial system and major corporations, while much of the cost of the crisis was socialised to the tax-payers through unemployment, austerity and deteriorating public services.
At the same time, central banks flooded the financial system with liquidity through quantitative easing. Interest rates were pushed towards zero, all the available capital then shifted to financial assets, driving their values to record highs, and the cost of borrowing collapsed.
I have written more extensively about this in my earlier pieces:
The 'so called' liberal capitalism died in 2008. Not with a bang but with a whimper.
The hyperscalers did not acquire their present investment power overnight. They were among the principal beneficiaries of the post-2008 economic order.
The flooding of the system with cheap money reduced the cost of debt. Thanks to this cheap money circulating in the system, the cost of borrowing was artificially brought down to zero.
Rather than producing a broad revival of productive investment, much of this liquidity flowed into financial assets, mergers, acquisitions and share buybacks. Instead of productive investments what these companies did was to conduct buybacks of their own shares. These buybacks increased returns to shareholders, supported equity valuations and strengthened the companies’ shares as acquisition currencies and gave them protection against any future financial shock. This cheap capital also allowed firms to distribute capital to shareholders while preserving their strategic position and consolidating their market power.
The post-crisis financial system continually inflated the value of financial assets while weak demand and austerity imposed on the world discouraged ordinary productive investment. Big Tech was uniquely positioned to absorb this financial abundance because it could deploy capital into platforms, algorithms, cloud infrastructure and network effects.
This is what eventually made possible the extraction of “cloud capital” according to Varoufakis.
The CEO of BlackRock, Larry Fink, admits that the trillions of dollars being used to build data centers and power grids will come from ordinary people’s savings accounts and pension funds, and says it is mandatory. He says America needs trillions in AI infrastructure spending,
It would be remiss not to mention a few other key turning points that further exacerbated this trend.
One is the Tax Cuts and Jobs Act, enacted as soon as Trump took office (again a nice indicator as to whose interests the state serves). The legislation largely removed the old incentive to keep accumulated foreign profits abroad and imposed a transition tax on previously unrepatriated earnings.
Thanks to legislation US corporations repatriated approximately $777 billion in 2018.
Another turning point is the even greater rounds of QE during 2020 Covid Pandemic. The Fed printed even more money during the pandemic. Once again, vast quantities of liquidity entered the financial system. As expected, the money circulating in the system found itself in the balance sheets of these same companies. The same process was repeated on an even bigger scale.
By the time the AI infrastructure race began, the hyperscalers had already spent fifteen years operating not merely as corporations, but as private financial institutions with enormous political and industrial muscle.
Set against this backdrop, the investments pouring into AI seem less to do with AI than a economic system slowly reaching an implosion point. It is I argue, saying more about the system than AI.
⚠️Wall Street is starting to panic over the AI debt bubble: Hyperscalers have issued $194 billion in debt so far this year across currencies, already accounting for a record ~9% of all US investment-grade (IG) bond supply. This share has QUADRUPLED since 2022 and has more than
What about the labor market?
Another part of society now confronting the destructive potential of AI is, of course, the labour market. A quick look into how the labor market was doing before the advent of AI strengthens the argument of this piece. Structural contradictions had been accumulating for decades. AI merely exposed and accelerated them.
The median real wages for the large parts of the populations had been stagnant, if not on a downward trajectory, for a long time. The financialisation of the economy and unfettered globalisation had already decimated the position of the average worker.
This is widely cited as one of the factors leading to Trump’s 2016 election. Nor did the crisis pertain only to blue-collar labour. Even a quick reading of David Graeber’s 2018 book Bullshit Jobs, funnily exposes the state of crisis most white collar work was in prior to AI.
It is also no secret that before the advent of generative AI, new graduates, software included, were on a downward employment trend.
Before vibe coding became a thing, programming was already evolving in that direction. It already increasingly consisted of installing and configuring stuff other people wrote, without reading the source.
Look at a typical headline of this sort for a moment: “AI is killing the cover letter - but here’s when it still pays to send one” . It is no secret that anyone in their right mind could have pointed out the inconsistencies of the cover letter system if someone actually wanted to hire a productive worker. It was performative to its core. The HR-Industrial complex however, was adamant about not letting it go.
AI did not make the cover letter absurd. It merely automated the absurdity to the point at which its continued existence became impossible to defend.
So, once again, the fact that AI is bulldozing over the labor market says in my opinion, more about the existing miserable state of the labor market than the revolutionary capabilities of AI.
Readership and text production were already in decline
I also want to address the foremost pedagogical anxiety surrounding the deployment of AI: the apparent decline of reading, writing and textual expression.
Here too, the crisis long predates generative AI. Just look at the statistics published by the National Endowment of the Arts: The proportion of adults who had read at least one book during the previous year fell from 54.6% in 2012 to 52.7% in 2017 and 48.5% in 2022.
A whole different discussion was underway, before AI was even introduced, as to how we get our kids to engage with texts again. Not to mention the whole crisis around attention spans.
To repeat, AI is not the culprit. There was something desperately needing to be addressed in our societies. AI is just making it harder to sweep it under the carpet.
We can of course extend this list of course to include more examples. this piece after all was written in haste. I wanted to get the argument out there so that we can actually talk about AI when we are talking about AI.
I say this as someone who is quite bullish on AI.
taken from: Taleb, N. N. (2005). Fooled by randomness: The hidden role of chance in life and in the markets (2nd ed.). Random House
