Murray Hunter
SPECULATIVE manias have repeatedly inflated and deflated around transformative technologies and assets. The question of whether today’s artificial intelligence boom will end in a sharp crash remains actively debated.
Warning signs have intensified in late 2026, with major tech stocks experiencing sharp pullbacks and China’s AI-related markets already suffering a severe correction.
Yet the underlying technology’s potential and the durability of the physical infrastructure being built complicate any simple “bubble will burst completely” narrative.
Historical Context: Bubbles That Left Something Behind
Bubbles are not new. In the Dutch Tulip Mania of the 1630s, rare tulip bulbs traded at extraordinary prices fueled by speculation and easy credit before collapsing.
The episode is often cited as a pure speculative frenzy with limited lasting economic infrastructure.
More relevant is the late-1990s dot-com boom.
Companies with little or no revenue achieved enormous valuations on the promise of future disruption.
The Nasdaq Composite peaked in March 2000 and then fell roughly 78% by October 2002.
Many firms failed, capital was destroyed, and a recession followed.
Yet the underlying internet infrastructure with fibre optic networks, data centres, and protocols all survived and became the foundation for later giants such as Amazon, Google, and the broader digital economy.
Excess capacity and bankruptcies effectively subsidised later growth.
Railroad booms in the 19th century and other infrastructure-driven episodes followed similar patterns.
This is where speculation financed overbuilding, prices crashed, investors suffered, but useful physical assets remained.
As investor and historian William H. Janeway has noted, speculative bubbles frequently leave behind valuable infrastructure even as they bankrupt many of the financiers who paid for it.
AI may follow a comparable path.
Current Signs of Market Strain
By late October 2026, pressure is evident on multiple fronts:
China’s sharp correction: The STAR50 index, heavy in AI and semiconductor names, plunged roughly 40% in under four months from its July peak, erasing about $480 billion in market value.
Broader Chinese tech and blue-chip indexes also declined significantly (CSI 300 down around 20% from June highs). Valuations had reached extremes (STAR50 peaked near 250 times trailing earnings; even after the drop, it remained elevated around 100 times, versus roughly 30 times for the Nasdaq 100).
Factors included stretched prices, questions about demand and profitability, tax pressures, and risks of export restrictions or bans on Chinese equipment in data centres.
AI-related stocks had come to represent a large share of Chinese market capitalisation and trading volume.
U.S. market volatility: The tech-heavy Nasdaq faced major pressure, with reports of roughly $2 trillion in value wiped from the Nasdaq 100 in October amid scrutiny of AI capital expenditures.
Alphabet experienced a historic single-day market-value drop exceeding $500 billion after earnings that showed weakness or misses in core search advertising alongside accelerated cash burn from AI infrastructure spending (negative free cash flow in the billions for the quarter).
Other hyperscalers such as Amazon, Meta, and related names faced intense investor questions about hundreds of billions in planned and ongoing capital expenditures.
Chipmakers, data-centre operators (including “neocloud” firms), and related stocks also saw sharp moves.
Financing risks: Analysts highlight growing reliance on borrowed money, private credit, and circular arrangements (chipmakers and hyperscalers effectively supporting each other’s customers and demand).
Rising interest rates increase the cost of this leverage.
Some observers compare elements of the financing to past credit-fueled excesses.
Global AI-related capital expenditure is projected in the high hundreds of billions for 2026 alone, with multi-trillion-dollar cumulative figures discussed for the decade
Prominent voices have issued warnings. Ray Dalio has described AI as a “classic bubble” approaching a potential bursting point, citing debt accumulation and the eventual need to convert paper wealth into cash.
Michael Burry and others have flagged similar concerns about timelines into 2027 or beyond.
Former regulators and strategists have pointed to possible reckoning windows in the coming 18–30 months or into 2027–2028 if returns fail to materialise.
Arguments That a Crash (or Major Correction) Is Likely
Critics emphasise the gap between investment scale and near-term commercial returns.
Enormous sums are flowing into data centres, chips, power, and models, yet broad consumer and enterprise monetisation lags.
Subscription rates for many AI tools remain limited relative to the hype; transformative applications and products that could generate the expected revenue streams are still emerging or unproven at scale.
Much of the activity has a speculative character, where valuations are priced for perfection and rapid, sustained growth.
Speculative markets have a long history of correcting violently once sentiment shifts, liquidity tightens, or earnings disappoint.
New technologies in data storage, more efficient models, or alternative architectures could also render portions of today’s heavy infrastructure investments less competitive or obsolete sooner than expected.
Heavy leverage and circular financing raise the risk that problems in one part of the ecosystem cascade.
China’s experience is viewed by some as an early warning that patience for delayed returns is finite.
If free cash flow remains deeply negative while debt service rises, equity valuations could face prolonged pressure even if the technology itself advances.
Arguments Against a Complete Burst
Optimists and many industry participants argue that AI is a genuine general-purpose technology with multi-decade potential, more akin to electricity or the internet than pure tulips.
Demand for cloud and AI services is already showing up in strong growth for certain hyperscaler divisions.
The physical infrastructure (data centres, power capacity, specialised chips) is real and durable; even if stock prices correct sharply, the assets can support future applications that have not yet been invented or scaled.
Janeway and others note that the infrastructure may outlast the speculative financing that built it.
Much early financing was equity-based rather than pure bank credit, which historically reduces systemic banking crisis risk compared with debt-fueled bubbles (though this is shifting).
Corrections can be healthy, pruning weak players, forcing efficiency, and resetting valuations without destroying the long-term trajectory.
Past technology cycles show that after busts, survivors and new entrants often create far more value than was destroyed.
Not every high valuation is irrational if productivity gains eventually materialise at scale across industries.
Some strategists and fund managers continue to view current volatility as a valuation reset within a longer boom rather than the start of a multi-year collapse.
Whether the AI boom ends in a full-scale crash comparable to the dot-com bust, a painful but contained correction, or a more gradual digestion of excess is still unresolved.
China’s markets have already experienced a sharp deflation of expectations. U.S. markets in late 2026 are showing clear strain from the mismatch between massive front-loaded capital spending and the slower path to broad, profitable adoption.
Historical parallels suggest that speculation often overshoots, pain follows for many investors, and yet transformative infrastructure and capabilities can endure and eventually deliver value.
The decisive variables will be the speed at which real revenue and productivity gains appear, the sustainability of financing amid higher rates and potential cash-flow pressure, and whether technological progress continues to outpace or fall short of the extraordinary expectations already priced in.
Both the bull and bear cases have substance, where the coming quarters and years of earnings reports, capital-spending guidance, and actual AI deployment will provide the clearest evidence. – November 2, 2026
