Global AI-related investment is now running at about $850 billion in 2026, roughly $500 billion above the pre-AI trend, making it larger and faster than historic manias such as railways, canals, and the dot‑com boom, said Macquarie analyst Viktor Shvets in a report.

Corporations, especially US hyperscalers, are rapidly exhausting internal cash, with debt issuance expected to reach around $180 billion and capex-to-revenue ratios climbing above 50%, underlining how aggressively AI is being funded. Yet, annualised AI revenues are already estimated at close to $175 billion, enough to cover current operating expenses and depreciation, and growing roughly three times faster than previous IT waves, suggesting the boom is not purely speculative.

Also Read | Chris Wood’s big warning: The specific risk that will finally trigger the end of AI trade

BIS warning: overextended but not empty

Live Events

The Bank for International Settlements has warned that AI now exhibits classic bubble characteristics, with extremely rapid capital deployment and increasingly complex off‑balance sheet vehicles and circular investment structures, which Macquarie says likely make current investment figures understated and more fragile than they appear. “AI is a bubble that could suddenly derate, with considerable consequences for markets and economies,” the note cautions, framing the current cycle as historically extreme in both scale and speed.
However, Macquarie stresses that adoption is running ahead of typical bubble patterns, with a $2 trillion contract backlog and heavy spending on data centres, memory and logic chips already visible in hard orders rather than just hype.
Economic impact still small, labour strains rising
Despite its market prominence, AI still accounts for a relatively modest share of overall economic activity, even as it increasingly shapes expectations for GDP growth and productivity. Macquarie warns that the real pressure points are emerging in labour markets, where lower hiring rates, declining education premia and signs of rising social polarisation point to early evidence of AI-related disruption that is not yet fully captured in official statistics.

The report argues that AI risks driving “declining marginal utility and compensation of labor,” with job insecurity and wage pressures likely to intensify as automation scales.

China’s cost shock: commoditisation is coming
Macquarie sees a major structural threat in China’s push to commoditise the AI stack, much as it did in solar, electric vehicles and batteries. On the latest data, China’s Z.ai and Tulongfeng systems are now matching the cybersecurity features of leading US model Mythos, with the US technological lead potentially narrowed to around 10–15%. Given China’s structurally lower cost base, this helps explain the rapid proliferation of its open‑weight models, which are being deployed primarily as cost‑efficiency tools, and underpins Macquarie’s view that pricing power in large language models – and ultimately in chips – will erode sharply.

‘Rolling bubbles’: from LLMs to applications
Macquarie’s central thesis is that the AI cycle will break not through one big burst but via a sequence of overlapping bubbles across the value chain. “We view AI as a sequence of ‘rolling bubbles’: LLMs to facilitators and applications. As one bubble deflates … others will pick up the mantle, until these bubbles also deflate,” the report says, noting that the market’s leadership is already rotating.

The so‑called Magnificent Seven have fallen from 36% of US market capitalisation to about 32%, while broader indices such as the S&P 500 and NASDAQ are showing phases where relative performance periodically shifts as leadership passes between segments.

In Macquarie’s view, periods between bubbles and shifts in monetary policy – for example when the US Federal Reserve tightens – may briefly broaden equity returns, but these will be “exceptions not the rule” in a cycle characterised by persistent concentration. Against that backdrop, the house outlines three broad approaches for investors navigating the AI boom: “day trade around headlines”, “go passive” or “go thematic”, reflecting a market environment where timing, diversification and exposure to structural themes may matter more than traditional stock‑picking. With no “reset button” in what it describes as an “age of extremes”, Macquarie concludes that investors should expect elevated volatility and serial repricing rather than a single, definitive end to the AI story.