The bank sees Big Tech capital spending climbing from about $800 billion to nearly $2 trillion as AI infrastructure demand keeps compounding
JPMorgan thinks the AI building boom has a long way to run. The bank expects hyperscaler capital spending to climb from roughly $800 billion in 2026 to nearly $2 trillion by 2030.
The numbers behind the forecast
Hyperscalers are the giant cloud operators that rent out computing power at enormous scale. JPMorgan’s list includes Amazon, Alphabet, Microsoft, Meta, and Oracle.
JPMorgan projects that growth will slow even as total spending keeps rising. The bank sees the annual growth rate moderating from 66% in 2027 to 22% in 2028.
JPMorgan estimates that 2028 alone could add between $292 billion and $321 billion in hyperscaler spending.
The bank also broke the outlook down by company. Amazon and Alphabet are each projected to contribute roughly $470 billion, Microsoft about $330 billion, Meta $290 billion, and Oracle $100 billion.
The cash flow question
The bank expects a cumulative free cash flow burn of about $284 billion across the major hyperscalers through 2028. Free cash flow is the money left after a company pays for operations and investments, so a burn means spending is outrunning that cushion.
Then the picture flips, according to JPMorgan’s model. The bank anticipates operating margins expanding to 35%, with free cash flow projected to reach $624 billion by 2030.
How we got here
Earlier estimates from the bank put hyperscaler CapEx at $342 billion in 2025, reflecting a 62% year-over-year increase.
Broader projections suggest global AI and data center infrastructure spending might exceed $5 trillion cumulatively by 2030. Debt financing is expected to play a crucial role in meeting that demand.
Commentary from Amazon’s Andy Jassy has aligned with JPMorgan’s optimistic view of continued AI growth, reflecting broadly bullish sentiment among executives about capital investment in the sector.
What this means for investors and the AI supply chain
Power is the first constraint to watch. The research flags potential strains on existing power infrastructure, since data centers at this scale need enormous amounts of electricity.
Financing is the second pressure point. JPMorgan’s analysis points to the need for diversified funding, including bond issuance and project-specific debt.
Project-level debt ties borrowing to a specific asset, like a single data center campus, rather than to the parent company’s general balance sheet.
The free cash flow forecast is where the thesis lives or dies. A $284 billion burn through 2028 followed by $624 billion in free cash flow by 2030 requires margins to actually expand to 35%.
Company-level spending guidance will be the real-time scoreboard. Each earnings season will show whether Amazon, Alphabet, Microsoft, Meta, and Oracle are tracking toward JPMorgan’s roughly $470 billion, $470 billion, $330 billion, $290 billion, and $100 billion projections, or drifting away from them.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.

2 hours ago
29








English (US) ·