Alphabet, Meta, Microsoft, Amazon and Oracle have disclosed roughly $3.3T in future contractual commitments, according to a compilation of company filings.
These include equipment purchases, leases not yet commenced and other long-term obligations. They are not $3.3T of debt, nor all AI-related spending.
The more important shift is that hyperscalers are locking in years of future payments before the corresponding infrastructure generates revenue.
That creates a duration mismatch: long-term obligations against uncertain AI demand and pricing.
Supplier backlogs may gain visibility, but hyperscalers also lose flexibility if utilization or returns disappoint.
The risk isn't simply how much AI infrastructure costs today. It's how much future spending has already been committed.
$GOOGL.US $META.US $MSFT.US
These include equipment purchases, leases not yet commenced and other long-term obligations. They are not $3.3T of debt, nor all AI-related spending.
The more important shift is that hyperscalers are locking in years of future payments before the corresponding infrastructure generates revenue.
That creates a duration mismatch: long-term obligations against uncertain AI demand and pricing.
Supplier backlogs may gain visibility, but hyperscalers also lose flexibility if utilization or returns disappoint.
The risk isn't simply how much AI infrastructure costs today. It's how much future spending has already been committed.
$GOOGL.US $META.US $MSFT.US