Meta 825B AI Commitments

Meta has agreed to spend about $825 billion on future leases and contracts, much of it for AI compute and data centers.

In July the number was almost $700 billion. It grew by roughly $125 billion in one quarter.

The breakdown matters more than the headline. $355.7 billion is in leases that aren't on the balance sheet yet, running through 2036. Another $454.5 billion is in non-cancellable contracts for cloud capacity, energy and hardware.

Non-cancellable is the key word. An AI buildout used to be a bet you could scale back if demand softened. At this size, the spending is largely committed before the revenue is proven.

For leaders buying AI, there are two takeaways. First, the hyperscalers are structurally motivated to fill that capacity, so expect aggressive pricing and bundling. Second, anyone building a multi-year strategy on one vendor's roadmap should ask how much of that vendor's future is already locked in, and what happens if demand grows slower than the commitments.

Capacity is being built on a decade-long timeline. Enterprise adoption still moves in quarters.

When the infrastructure is committed this far ahead of the demand, who carries the risk if the two don't line up: the builders, or the customers who signed on to the roadmap?