Dimon Puts Hyperscaler AI Spending Near $1 Trillion Next Year

Artificial intelligence hyperscalers could spend $1 trillion next year, JPMorgan Chase chairman and chief executive Jamie Dimon told CNBC¹. Dimon said on Monday that investment across the artificial intelligence infrastructure ecosystem could reach $1 trillion². He said the surge in spending shows little sign of slowing down⁶.

Dimon also warned that the spending could add a little to inflation, and noted a pivotal shift in the US economy as spending moves away from its earlier pattern³.

Why it matters to investors

The published forecasts run higher than Dimon’s number. S&P Global Ratings projected combined hyperscaler capex could exceed $1.3 trillion in 2027, while UBS expects total AI capex to reach roughly $1.4 trillion⁴. A separate account of the same S&P work put hyperscaler capex above $1.3 trillion by 2027⁵.

Company guidance points the same way. Hyperscaler capital expenditure is projected to hit $800 billion in 2026 and $1.3 trillion in 2027, yet Nvidia, which reported $96 billion in revenue despite AI fears, can supply only roughly 70 percent of it⁷.

What to watch

Scale has turned capex into a macro variable rather than a sector story. AI investment has surpassed US housing, driven by hyperscalers such as Amazon and Microsoft⁵. Spending of that size competes for the same credit, power and equipment as everything else being built^5,7.

The markers from here are whether the 2027 forecasts hold as guidance rather than aspiration^4,5, whether supply rather than demand becomes the binding constraint⁷, and whether Dimon’s inflation caveat shows up in the price data³. Dimon’s own framing is that the spending is not slowing^1,6, which makes the financing mix the next thing to examine².

The spread between the forecasts is itself informative. Dimon’s figure for next year sits at $1 trillion^1,2; S&P Global Ratings puts combined hyperscaler capex above $1.3 trillion in 2027 and UBS puts total AI capex near $1.4 trillion⁴. Company guidance bridges the two, with $800 billion in 2026 stepping up to $1.3 trillion the following year⁷. None of these are the same measure, and the difference between hyperscaler capex and total AI capex is doing real work in the comparison^4,5. The supply-side counterpart is the constraint that Nvidia can cover only roughly 70 percent of it, which is why the bottleneck question is not academic⁷.