AI buildout echoes dot-com and railway manias, BIS warns on bust risk and recession threat
The BIS framing puts a systemic lens on what markets have been treating as a sector-specific valuation debate. The warning that household equity exposure has risen materially relative to both wealth and income means a sharp AI-driven correction would transmit to consumption more forcefully than past tech busts, with spillovers well beyond US borders given the outsized share of American equities in global market capitalisation. The inflation dimension compounds the central bank dilemma: AI energy and chip demand is already pressuring input costs at a moment when Middle East conflict has pushed inflation above target in major economies, yet the BIS explicitly declined to recommend rate hikes as a policy response, citing uncertainty. For commodity and energy markets, the buildout bottleneck findings are directly relevant: electricity, advanced semiconductors and grid equipment shortages are amplifying the capex cycle, with long-dated supply contracts locking in exposure that magnifies the downside if AI demand disappoints. Credit markets face a secondary contagion risk through data centre construction and hardware supply chains, where leveraged borrowers depend on hyperscaler spending remaining intact.
—
The BIS warned Sunday that a $1 trillion AI investment boom risks an abrupt bust and recession if returns disappoint, comparing the cycle to the dot-com crash and railway mania of the 1840s.
Summary:
- The five largest hyperscalers are on track to spend more than $1 trillion on AI-related capital expenditure across 2025 and 2026, with the BIS warning the race for market share may have driven investment to excessive levels
- BIS general manager Pablo Hernández de Cos said the sector is vulnerable if AI under-delivers, potentially bringing the investment boom to an abrupt end with economy-wide recession consequences
- The BIS drew explicit parallels with canal construction in the 1830s, British railway mania in the 1840s, late 1920s electrification and the late 1990s dot-com boom, all of which ended in investment reversals and recessions
- Equity market valuations were described as reflecting investor complacency, with the BIS warning a large correction could produce more severe wealth effects and consumption pullbacks than past corrections of similar scale, given the growth in household equity exposure relative to income and total wealth
- AI’s voracious demand for electricity, advanced semiconductors and grid equipment is already pressuring input costs with potential inflation spillovers, while long-dated supply contracts are amplifying overinvestment risk across the supply chain
- Despite the warnings, the BIS stopped short of recommending central banks raise rates to cool the boom, with Hernández saying prescriptive guidance on policy response would be unwise given the level of uncertainty involved
The Bank for International Settlements has issued one of its starkest warnings in years about the trajectory of artificial intelligence investment, cautioning in its annual global economic outlook that a spending race among the world’s largest technology firms risks ending in an abrupt bust that could tip economies into recession if AI profits fail to meet the expectations now embedded in markets and capital allocation decisions.
Reporting by the Wall Street Journal (gated) on the BIS annual report, published Sunday, revealed that the five largest hyperscalers are on course to commit more than $1 trillion to AI-related capital expenditure across 2025 and 2026 combined. The Switzerland-based body, which serves as a coordinating institution for the world’s central banks, said the competitive pressure to secure market share may have driven investment beyond levels that can be justified by realistic returns.
BIS general manager Pablo Hernández de Cos framed the risk in direct terms: if AI under-delivers on its productivity promises, the financing that has sustained the buildout could be withdrawn suddenly, ending the boom with the kind of speed and severity that previous technology investment cycles have demonstrated. The historical comparisons the BIS reached for are not reassuring. Canal construction in the 1830s, British railway expansion in the 1840s, electrification in the late 1920s and the dot-com surge of the late 1990s all shared the same basic structure: large sums deployed rapidly into a transformative technology, followed by a reversal that induced economy-wide recessions.
The BIS was careful to distinguish AI’s longer-term potential from those predecessors. If AI systems eventually become capable of improving their own capabilities and autonomously generating new ideas and technologies, the macroeconomic consequences could be profoundly different from past innovations, potentially lifting the fundamental constraint on long-run growth imposed by the pace at which humans generate new knowledge. That upside, however, is precisely the expectation the BIS warned may be generating the current complacency.
Equity market valuations were identified as a particular concern. The BIS said investor pricing reflects a level of comfort with risk that is not warranted given the uncertainties involved, and that a large correction would carry more severe economic consequences than past episodes of comparable scale. The reason is structural: household equity exposure has grown significantly over recent decades relative to both total wealth and income, meaning a sharp fall in prices would produce more pronounced wealth effects and a sharper pullback in consumption than historical precedents would suggest. The contagion would not be contained within the United States despite most major AI players being American, given the disproportionate share of US equities in global market capitalisation.
The inflation and credit dimensions add further complexity. AI’s demand for electricity, advanced semiconductors and grid equipment is already generating input cost pressure with potential spillovers to broader inflation, at a moment when the Middle East conflict has pushed price growth well above central bank targets in major economies. Meanwhile, the shift from internally funded investment to debt financing has introduced financial stability risk: a slowdown or halt in hyperscaler capital expenditure could leave leveraged borrowers across the data centre construction and hardware supply chain unable to service debt or replace lost revenue.
Despite the weight of the warnings, the BIS explicitly declined to recommend that central banks should tighten policy to cool the boom, with Hernández saying that attempting to be prescriptive about the appropriate monetary policy response would be unwise given the scale of uncertainty. That judgment leaves markets and policymakers in the uncomfortable position of monitoring a risk the BIS has named clearly but for which it has no straightforward remedy to offer.
This article was written by Eamonn Sheridan at investinglive.com.