Jefferies warns of massive capital destruction from hyperscaler AI spending

Chris Wood, Global Head of Equity Strategy at Jefferies, warned on September 18, 2026 of massive capital destruction from the AI investments of American hyperscalers. "My base case is that there will be massive capital destruction in the US," Wood said in a recent statement. The central question is when markets will begin to price this in.

Trillion-dollar investments in AI infrastructure

The four largest US hyperscalers – Microsoft, Alphabet, Amazon and Meta – are investing approximately 695 billion US dollars in 2026 combined, according to Wood's analysis from July 2026. For 2027, the strategist forecasts 870 billion US dollars, which over both years amounts to a total of around 1.57 trillion US dollars. Across all six major hyperscalers – which also include Oracle and SpaceX – Wood expects approximately 916 billion US dollars in the next twelve months and nearly 1.2 trillion US dollars in the following period.

Economist Torsten Slok estimated in August 2026 that AI spending could potentially reach 3.1 percent of US gross domestic product for 2027. This would make the share three times larger than previous comparable values. Wood, who correctly predicted previous financial bubbles like the dot-com boom, Japan's credit bubble and the US housing bubble, now sees a comparable pattern.

Critical shift to debt financing

According to Wood, a central risk factor lies in the financing structure: After three and a half years of the AI boom, companies are increasingly financing their spending through debt rather than operating cash flow. The capital expenditures of the four largest hyperscalers reached 92 percent of projected operating cash flow in 2026 by Wood's calculation – a level he considers "remarkably high."

This development makes companies more vulnerable to credit risks. Wood points to Oracle, which experienced a credit rating downgrade in July 2026 that was tied to high AI spending. The Jefferies strategist sees the danger that hyperscaler bond spreads will widen if markets assess the financing structure more critically.

Returns question increasingly in focus

"My base case is that in the end they will waste a lot of money on AI capex and a lot of capital will be destroyed. But the key question is when the market begins to price this in," Wood said on September 18, 2026. According to Wood's observations from July 2026, investors are increasingly asking where returns on these capital expenditures should come from, after initially welcoming the spending.

The market is now reacting negatively to announcements of further capital expenditure increases, as Wood observed on August 1, 2026. This could point to a turning point in sentiment. Wood emphasizes that any news raising doubts about the entire cycle could serve as a trigger for a revaluation.

Competition from China as additional burden

Wood warned in July 2026 that cheaper Chinese open-source models could undermine the economic foundations of the American investment wave. His long-term base case assumes that market share will shift toward Chinese large language models.

He cited data from OpenRouter as evidence: In the week through July 19, 2026, the leading Chinese AI models on the global platform processed 36.39 trillion tokens, compared to just 7.39 trillion tokens from leading US models. According to Wood, China is increasingly perceived as a technological competitor on equal footing in the field of artificial intelligence. On July 17, 2026, Moonshot AI launched the open-source model Kimi K3.

Risk of major credit event

Wood warned on September 18, 2026 that the AI investment boom "could trigger a major credit event if the cycle turns." This points to systemic financial risks that go beyond individual technology stocks and could affect broader credit markets. The strategist compared the vulnerability to a real estate squeeze akin to the 2008 crisis, triggered by a credit crunch.

The market's decision to withdraw credit could mean a rapid end to the trend, in Wood's assessment. "The central macroeconomic question for equity markets right now is the outlook for AI capital expenditures," Wood said on September 18, 2026. He did not provide a clear timeline for such a scenario.

Impact on semiconductor stocks

According to Wood, semiconductor stock valuations depend directly on whether markets continue to believe in hyperscaler spending. As long as markets do not aggressively question capital expenditures, chip stocks would benefit. A collapse of the AI cycle would, according to Wood's statement on September 17, 2026, have implications for other markets as well – he cited as an example that only an AI implosion could bring significant foreign capital back into markets like India.

The transformation of hyperscalers from asset-light to capital-intensive business models represents a structural shift, according to Wood in May 2026. The spending level has now reached dimensions where it absorbs an increasingly large portion of cash flows, particularly for chip and memory purchases.