Three large banks now forecast at least $1 trillion of AI-related capital spending in 2027. Morgan Stanley is highest at $1.3 trillion, a figure it raised from $700 billion last November, per Reuters. Goldman Sachs projects $1.01 trillion and JPMorgan about $1 trillion, per the I/O Fund's summary of both.
A $300 billion gap separates the highest 2027 forecast from the lowest bank figure, and definitions explain much of it. The published numbers, with what each one counts:
- Morgan Stanley: $850 billion in 2026, $1.3 trillion in 2027, $1.5 trillion in 2028. Data center capex.
- Goldman Sachs: $765 billion in 2026, $1.01 trillion in 2027, $1.64 trillion by 2031. AI capex, with $7.6 trillion cumulative from 2026 through 2031.
- JPMorgan: $800 billion in 2026, $1 trillion in 2027. Hyperscaler capex.
- Analyst consensus: $934.5 billion in 2027 for Alphabet, Amazon, Microsoft, and Meta combined.
Every one of these moved up in 2026. Morgan Stanley nearly doubled its 2027 number in eight months. Goldman had 2025 Big Tech capex at $253 billion as recently as September 2024, and its 2026 figure is now three times that. Forecasts that rise this fast are tracking company guidance. They carry little independent view of demand.
The spending has to be financed. Morgan Stanley expects AI-related debt issuance near $570 billion in 2026, which we covered in June. JPMorgan's work on bond supply is in our August piece. Third-quarter earnings later this month bring new guidance from all four hyperscalers.
Bottom Line
The 2027 consensus is about $1 trillion, and the range runs from $934.5 billion to $1.3 trillion depending on who is counted. If you sell into this build-out, plan against the four-company consensus, since it is tied to named budgets. Check each bank's definition before quoting its number next to another's.