We inhabit a recurring nightmare that keeps promising to be different this time. The dream of frictionless accumulation is rendered plausible by each new acronym—AI, VR, Web3—until the material contradiction it was designed to mask returns to violently reassert itself. Since 1989, what we call tech crises have not been moments of madness interrupting an otherwise rational system; they are the way capital periodically incinerates its own over-accumulation in order to resume the grim work of extracting value from living labour.

The early 1990s recession offered a preview of what was to come, though few recognised it as such. Unemployment climbed from 5.2 percent to 7.8 percent by mid-1992, even as GDP growth had technically resumed, and the tech sector shed 48,500 jobs in the forty-two months following the official end of the recession—a primary crisis of production in which the wave of automation that had swept through manufacturing and logistics during the 1980s finally exhausted its countervailing tendencies. Military-industrial spending collapsed with the Cold War's end, the old spatial fixes were closing, and something new was required, something large enough to absorb the growing mass of capital that could no longer find profitable outlets in the productive sphere.

What arrived was the internet, and the dot-com bubble was a temporary fix as surplus capital, starved of productive investment, hurled at claims on future profitability that might never materialise. This is fictitious capital in its purest form—wealth that appears on balance sheets with almost no basis in the actual production of goods and services. Between 1995 and 2000, venture capital flooded into any company whose business plan contained the word "platform" or whose domain name ended in .com, and the valuations were merely prayers addressed to a god that does not answer. When the NASDAQ peaked and collapsed, more than five trillion dollars in market value evaporated, tens of thousands of rank-and-file workers lost their livelihoods, and the productive forces the bubble had financed—fibre optic cables, server farms, routing infrastructure—remained standing, now cheap enough for the survivors to consolidate into what would become surveillance capitalism. The crash did not destroy the technology but the destroyed the workers and smaller firms that had built it, leaving the wreckage to Amazon, Google, and Meta.

By 2008, the centre of speculation had shifted from technology to housing, but the underlying dynamic was identical. The financial crisis was rooted in a declining rate of profit extending back through the 1990s, a tendency Marxist economists have demonstrated through empirical data with considerable rigour. When the productive sector can no longer generate sufficient surplus value, capital flows into secondary circuits—mortgage-backed securities, collateralised debt obligations, derivatives whose complexity served primarily to obscure their emptiness—and the crash wiped out over twenty trillion dollars in household wealth. But crucially, the tech sector emerged strengthened: the infrastructure of smartphones, mobile internet, and cloud computing that had been laid during the dot-com rubble suddenly became indispensable, and a crisis that gutted manufacturing and construction left the digital monopolies ascendant, gorged on the cheap capital and desperate labour the crash had produced.

The cryptocurrency crash of 2018 was a still more naked expression of fictitious capital, stripped of any pretence that the underlying asset performed useful social labour. Bitcoin's value collapsed by over eighty percent; Ethereum, Ripple, and thousands of other tokens followed it into the abyss. The libertarian rhetoric of decentralisation could not disguise the material reality that these were claims on future value backed by nothing but the willingness of later speculators to buy in, a structure indistinguishable from a Ponzi scheme except in its ideological self-presentation. When that willingness evaporated, so did the wealth. And yet blockchain technology survived, absorbed into the portfolios of major financial institutions, ready to fuel the next cycle of speculative fever, because the technology is simply the vehicle through which surplus capital attempts to valorise itself without passing through the tedious business of production.

2022 brought the most dramatic tech crash since the dot-com unwinding, and in some ways a more revealing one. The NASDAQ lost over five trillion dollars in value between its November 2021 peak and mid-2022, exceeding the dollar losses of the entire dot-com collapse. Global tech layoffs reached 161,411 in 2022 alone—a six hundred and forty-nine percent increase over the prior year—with another 155,462 cuts in just the first three months of 2023. What was being exposed here was the deeper structural problem that had been mounting since 2009: a decade of near-zero interest rates had inflated balance sheets far beyond any plausible connection to the value those companies actually produced through the exploitation of labour. Labour time determines market value, even in the digital economy, and without commensurate increases in productivity the valuations could not be sustained.

We are now living through the largest and most dangerous of these cycles: the AI bubble stretching from 2024 to the present. The scale of speculation defies historical comparison. NVIDIA alone lost six hundred billion dollars in market capitalisation in a single day in January 2025, and annual capital expenditure on artificial intelligence has reached three hundred and thirty-two billion dollars against just twenty-eight point seven billion in revenue—a ratio that would be called insanity in any other context but is called vision when enough capital is committed to sustaining the illusion. This is a bet that the future will be so radically transformed that its profits will justify any present expenditure, fictitious capital on a scale Marx could not have imagined but whose mechanisms he described. The AI boom absorbs surplus capital that has nowhere else to go while promising to revolutionise every sector of the global economy, and in doing so it reveals something about the current phase of accumulation: that capital has run out of productive terrain and is now speculating on the abolition of labour itself.

What makes this bubble far more dangerous than its predecessors is its concentration. The dot-com crash dispersed losses across thousands of firms; the AI crash will concentrate them in a handful of trillion-dollar companies and the financial institutions that have leveraged themselves to those companies. If the crash comes, the destruction will not be contained to Silicon Valley but will rip through pension funds, sovereign wealth vehicles, and the balance sheets of governments that have tied their fiscal futures to the promise of an AI-driven productivity miracle. Unlike the cryptocurrency crash, which could be dismissed as a sideshow for speculators and libertarian fantasists, this one sits at the centre of the global accumulation process.

These cycles do not produce innovation. The productive breakthroughs—the fibre optic cables, the smartphone infrastructure, the cloud computing architecture—were built before the crashes, often by workers whose wages were suppressed and whose labour was intensified because the speculative frenzy created conditions in which exploitation could be deepened under the guise of opportunity. The crashes destroy the claims on future value that have accumulated beyond any reasonable expectation of realisation, wipe out smaller competitors and upstart firms, render worthless the stock options that workers accepted in lieu of wages, and concentrate capital into fewer and fewer hands, leaving the survivors more powerful than before. What Schumpeter celebrated as creative destruction was always, from a materialist perspective, simply destruction—the violent devaluation of capital that has become unmoored from its productive base, followed by a reorganisation that deepens exploitation, entrenches monopoly, and prepares the ground for the next round of the same catastrophe dressed in new clothes.

We cannot predict the precise moment the AI bubble will burst, any more than we could have predicted the exact peak of the dot-com mania or the week in September 2008 when the financial system seized. But we can recognise the pattern with the clarity that comes from having watched it repeat for decades: the over-accumulation of capital seeking outlets, the turn to fictitious claims on future value, the suspension of disbelief that passes for investor confidence, and finally the reassertion of material reality against the fantasies capital tells itself. That pattern has not changed because the underlying contradiction has not changed—capital must constantly revolutionise the instruments of production while fleeing the falling rate of profit those very revolutions produce, and no technology, however transformative, can resolve a contradiction that is structural to the mode of production itself.

We are riding the same wave that has broken repeatedly across four decades, told each time that this wave is different because the technology is different or the market is larger or the future is closer. But the wave is not the technology; the wave is capital searching for somewhere to land before it drowns in its own excess, and when it breaks, as it always breaks, the ones who will pay are not the ones who rode it but the ones whose labour built the board.