Most people presumably know by now that we’ve been building a gigantic bubble in AI, and that this bubble is destined to burst. The scale of risk involved is large, and some of the characteristics of that risk are extraordinary.
Its denouement – perhaps in conjunction with the bursting of the contemporaneous bubble in cryptocurrencies – might prove very much more value-destructive than the global financial crisis of 2008-09.
But what very few seem to have noticed, however, is the sequential character of these ‘bouts of irrational exuberance’.
As we shall see – and beyond its sheer size – the AI craze has some distinctive features all of its own.
But the dynamic at work here is one of successive attempts to delay and disguise the inflexion of the underlying economy of material products and services from growth into contraction.
The operative myth is that monetary innovation can reinvigorate a flagging material economy.
Illogical though this notion is, it’s proved far more palatable than the reality of post-fossil material economic contraction.
We cannot yet know whether the AI bust – with or without the help of cryptos – will be big enough to crash the global financial system.
But we can be assured that, if the AI and crypto bubbles aren’t big enough to bring down the structures of money and credit, we’ll carry on building ‘bigger and better’ bubbles until we find one big enough to get the job done.
Whilst the AI craze is fascinating in itself, our main focus here is on the mechanisms involved in the pursuit of financial disaster.
1
The bubble in almost everything AI-related is undoubtedly very large. Most estimates put the scale of capital at risk at between $1.5 and $2 trillion, though it’s no surprise that some sources have produced even bigger numbers, some as high as $6tn.
Even at the lower end of these estimates, the bursting of this bubble could destroy value at about seventeen times the scale of the dotcom crash of 2000-02, and exceed the losses of the 2008-09 global financial crisis by a multiple of four.
Beyond its sheer size, however, this bubble has a series of characteristics that distinguish it from previous such bouts of recklessness
First, massive investment in AI is concentrated in a small group of giant technology companies which together account for about one-third of the entire value of the American stock market.
Second, the investment interconnections between these big players appear to involve a great deal of circular financing.
The sums thus far committed vastly exceed the free cash flows of the companies involved, and one of the central players – Open AI – seemingly expects its losses to rise from $9bn this year to $47bn by 2028.
There exists, as yet, no demonstrable route even to profitability in AI, let alone to the earning of adequate returns on existing and planned capital investment.
If – or rather, when – the bubble bursts, scope for the recovery of residual value looks remarkably small. Most of the assets against which lending is secured consist of enormously expensive, fast-ageing GPUs and huge, single-purpose buildings ‘in the middle of nowhere’.
There’s every prospect of the bubble ending in fire-sales of out-dated chips and redundant data centres.
Perhaps the strangest characteristic of the lot, though, is that the AI model favoured by American Big Tech demands resources – including energy and water – that simply do not exist in the requisite quantities.
Even if they did exist, these demands could pitch deep-pocketed tech behemoths into one-sided competition with households, municipalities and other businesses for constrained supplies of necessities, including power and water.
Putting all of this together – the scale of investment, the paucity of current cash flows, vendor financing, minimal residual value, the lack of a persuasive business plan and unrealistic resource demands – has led some observers to conclude that the AI craze is a gigantic scam.
There is, though, an alternative hypothesis, one which blends hubris with a determination to follow a strategy which, whilst effective in the past, threatens to end disastrously in its latest iteration.
2
We need to be clear that the bursting of a financial bubble does not necessarily pass a negative verdict on the product, service or technology on which the bubble has been built.
Rail remained a very important means of transport after the collapse of the Victorian railway mania of the 1840s. The internet survived and prospered despite the dotcom fiasco, and real estate retained its utility after the subprime lunacy had ended.
AI, perhaps in some form which is thus far undefined, can remain a worthwhile technological breakthrough after the current bubble has burst.
But, if some form of worthwhile, potentially transformative AI does lie ahead, what’s driving the absurd bubble in the sector?
America’s major technology players developed a winning strategy in the aftermath of dotcom. The essential prerequisite is the accumulation of capital resources big enough to finance the making of large losses over a protracted period of time.
This ‘go big’ strategy of loss-leading is how you create a very large customer base, draw in advertisers and other commercial entities, destroy or acquire your competitors, and gain industry-standard dominance for your product.
The aim – whether in social media, internet search, software or on-line retail – is to leverage huge capital availability into the creation of a quasi-monopoly which leads, in due course, to very high levels of profitability.
There can, though, be no guarantee that this ‘go big’ strategy will work with artificial intelligence. For a start, AI could degenerate into an echo-chamber in which it regurgitates its own output in a process known as “AI slop”.
Competitors – most obviously in China – might out-compete Big Tech by preferring an innovative strategy of going smart to the brute force of ‘going big’.
There are no equivalents to the obvious revenue streams that were waiting to be tapped by social media, search and on-line retail twenty or more years ago.
This could quite simply turn out to be a case of ‘too big to work’, in terms less of invested capital than of energy and other resources.
There can be little reason to doubt the hubristic over-confidence of Big Tech, whose existing businesses may be at risk both from enshittification and from underlying economic changes that its leaders either misunderstand or choose to disregard
Is this a disastrous case of ‘going big’ in a completely unsuitable context?
3
Most people probably realise by now that the enormous bubbles in AI and in cryptos are destined to burst, though suggestions that this might cause “some” economic harm are likely to rank amongst the biggest under-statements of all time.
But few seem to recognise the dynamic that has been driving a sequence of ever-more-reckless financial excesses.
As regular readers will know, we need to start by recognizing that money has no intrinsic worth, but commands value only in terms of those physical things for which it can be exchanged. What this means is that money is nothing more than “an exercisable claim on the material”.
We have an infinite ability to create these monetary “claims”, but we cannot similarly create a corresponding increase in the material supply by which alone money is validated.
This obvious “claim” characteristic of money compels the drawing of a distinction between two economies. One of these is the “real” economy of material products and services, and the other is the parallel “financial” economy of money, transactions and credit.
Growth in the “real” economy has long been decelerating towards contraction. Depletion of low-cost carbon energy is reflected in relentless rises in the proportionate Energy Cost of Energy, and there exists, thus far, no alternative to fossil fuels which can tame, let alone reverse, the upwards march of ECoEs.
Meanwhile, the rate at which energy converts into a flow of material value has been trending downwards. This tells us that the non-energy natural resource base – which includes minerals, non-metallic mining products, biomass and accessible water – has been degrading more rapidly than the efficiency of conversion techniques has been able to advance.
At the same time, the vital resource of environmental tolerance for human economic activity has been deteriorating, latterly at a disturbingly rapid pace.
There are two main reasons why this process of deceleration towards economic inflexion has thus far escaped general recognition, despite the growing abundance of evidence to support it.
First, the economics orthodoxy blithely ignores the material, assuring us that no obstacles exist to the use of money to deliver ‘infinite growth on a finite planet’.
This pseudo-science purports to find “laws” of economics in what are in fact nothing more than behavioural observations about the human artefact of money, observations which are not in any way analogous to laws of physics.
Second, an absolute refusal to accept the reality of physical finality has led to the equally fallacious proposition that a flagging material economy can be reinvigorated using monetary tools.
There is a pronounced tendency for cornucopian fantasies to triumph over material realities.
4
Recent history graphically illustrates these processes at work.
We tried to counter the “secular stagnation” of the 1990s by pouring abundant borrowed liquidity into the system. When this “credit adventurism” led, with due inevitability, to the GFC of 2008-09, we doubled down with the “monetary adventurism” of QE, ZIRP and NIRP.
These exercises in financial gimmickry have had a massively distortionary effect on the relationship between asset values and all forms of income. The resultant increases in inequalities have created levels of social instability which, in the absence of enlightened reform, might end only in revolution, anarchy or authoritarian rule.
But our immediate concern must be with the ever-worsening financial risk which these gambits have created. These hazards fall into the distinct categories of scale and complexity risk.
Over the past twenty years, in which global reported real GDP has increased by $96tn, debt has expanded by $284tn, and broader financial liabilities by not less than $770tn – and even the latter number excludes enormous “gaps” in the adequacy of provisions to meet pensions promises.
Moreover, most reported “growth” has been nothing more substantial than the spending of huge amounts of borrowed money. SEEDS analysis indicates that material economic prosperity has increased by only 25% – rather than the reported 96% – since 2004.
Indeed, it has latterly become routine for reported “growth” in GDP to be significantly less than sums borrowed by governments alone.
The worsening of complexity risk has been even more dramatic than the rapid expansion of aggregate liabilities. The financial system has become a truly Byzantine structure of cross-collateralization, in which nobody really knows which component, perhaps small in itself, could, by failing, bring down the whole house of cards.
This can best be considered as an inverted pyramid, in which a massive and super-complex tophamper of liabilities sits atop a remarkably small foundation of monetized value.
The assets of NBFIs – the non-bank financial intermediaries known colloquially as “shadow banks” – have, in modern times, dramatically out-grown those of the regulated banking system.
It’s a disturbing reflection that the extent of broad liabilities isn’t even known with any certainty, since the reporting of data to the FSB is voluntary, and non-reporting jurisdictions include a string of massively-exposed specialist financial centres.
What this means is that, with the locus of risk migrating from the regulated centre of this system to its opaque and hazardous periphery, ever-increasing complexity risk has been added to the quantitative risk of unsustainably rapid expansion in the aggregates of exposure.
5
The mechanism at work here is one by which the consequences of each bout of financial excess lead on to ever-greater exercises in recklessness.
We can trace a sequence which began in the 1990s with attempts to reinvigorate a flagging economy with the super-rapid creation of credit. When this culminated in the GFC, the authorities were more or less compelled to respond with ultra-loose monetary policies.
These sequential processes have been described here as an arc of inevitability.
At the same time, 2008 was a massive exercise in moral hazard – investors who have once been rescued from the consequences of their follies or misfortunes naturally assume that they will be bailed out from any future excesses, even though a repeat of the rescue of 2008 has long since ceased to be a practical possibility.
Once these processes are understood – as a self-driving dynamic of excess in conflict with material economic contraction – it becomes self-evident that the creation of bubbles in a climate of recklessness cannot end until the global financial system collapses.
The AI craze – together with the bubble in crypto-currencies – might prove big enough to crash the system.
If they are not, we can be assured that we’ll keep building ‘bigger and better bubbles’ until this result has been achieved.The last working-class hero in England. Clio the cat, ? July 1997 - 1 May 2016 Kira the cat, ? ? 2010 - 3 August 2018 Jasper the Ruffian cat ??? - 4 November 2021 Georgina the cat ???-4 December 2025
"a repeat of the rescue of 2008 has long since ceased to be a practical possibility"
on October 23, 2025, 8:22 pm