AI · 2026-10-23
Meta’s Future Spending Commitments Hit $825 Billion, Raising the Cost of Its AI Infrastructure Push
Meta said it has committed to spending about $825 billion in the future through leases and various contracts. The figure covers data centers, energy infrastructure and outside cloud computing, and is up from the roughly $700 billion it disclosed three months ago.

Meta has revealed just how large its long-term artificial intelligence (AI) infrastructure plans have become. Meta Platforms said it has committed to spending about $825 billion in the future through leases and contracts. This does not mean the company has already spent that amount in cash or that it is all recorded as debt on the balance sheet. Rather, it should be understood as the total value of contractual and lease commitments to be fulfilled over a long period of time.
### What is included in the leases and contracts
According to Meta’s disclosure, the company has $355.7 billion in operating and finance leases that have not yet been reflected on the balance sheet. These include leases related to data centers and energy infrastructure that begin from the remainder of this year through 2036.
Separately, non-cancellable short- and long-term contracts totaled $454.5 billion as of September 30. Much of that is tied to external cloud computing capacity, the energy needed for data centers, and consumer hardware products from Reality Labs. Meta also said it has agreed to buy up to $14.7 billion in cloud capacity over five years. However, the company noted that this obligation could decline if the service provider sells that capacity to other customers.
If those three items are added together, they come to about $824.9 billion, which is essentially in line with the $825 billion total commitment mentioned in the disclosure. Data centers and computing resources account for a large share of that amount, but the full figure should not be interpreted as AI-only spending. Energy contracts and consumer devices are included as well.
### About $125 billion added in three months
Back in July, Meta said its future spending commitments were about $700 billion. Compared with the newly disclosed $825 billion, that represents an increase of roughly $125 billion in just three months. In its recent quarterly earnings announcement, the company also raised its capital expenditure plan for this year and unveiled a new class of non-voting shares that can be used to help fund acquisitions and other costs. However, the report provided here does not include a specific figure for the revised capital spending plan.
### Execution and profitability matter as much as the size of the investment
This disclosure shows how Meta is locking in computing resources, data centers and power infrastructure for the long haul as it expands AI services. At the same time, such a large commitment also means investment spending and contract execution will likely continue for years to come. For that reason, it is more accurate to view the $825 billion as a multi-year commitment to operations and infrastructure, not as a short-term cash outlay.
Going forward, the key question will not only be the size of the investment, but also how effectively the infrastructure translates into growth for AI services and Meta’s existing businesses, and whether the company can maintain profitability while absorbing the cost burden. The figures and disclosure details were compiled based on a Bloomberg report carried by NewsPim.
← Back to articles### What is included in the leases and contracts
According to Meta’s disclosure, the company has $355.7 billion in operating and finance leases that have not yet been reflected on the balance sheet. These include leases related to data centers and energy infrastructure that begin from the remainder of this year through 2036.
Separately, non-cancellable short- and long-term contracts totaled $454.5 billion as of September 30. Much of that is tied to external cloud computing capacity, the energy needed for data centers, and consumer hardware products from Reality Labs. Meta also said it has agreed to buy up to $14.7 billion in cloud capacity over five years. However, the company noted that this obligation could decline if the service provider sells that capacity to other customers.
If those three items are added together, they come to about $824.9 billion, which is essentially in line with the $825 billion total commitment mentioned in the disclosure. Data centers and computing resources account for a large share of that amount, but the full figure should not be interpreted as AI-only spending. Energy contracts and consumer devices are included as well.
### About $125 billion added in three months
Back in July, Meta said its future spending commitments were about $700 billion. Compared with the newly disclosed $825 billion, that represents an increase of roughly $125 billion in just three months. In its recent quarterly earnings announcement, the company also raised its capital expenditure plan for this year and unveiled a new class of non-voting shares that can be used to help fund acquisitions and other costs. However, the report provided here does not include a specific figure for the revised capital spending plan.
### Execution and profitability matter as much as the size of the investment
This disclosure shows how Meta is locking in computing resources, data centers and power infrastructure for the long haul as it expands AI services. At the same time, such a large commitment also means investment spending and contract execution will likely continue for years to come. For that reason, it is more accurate to view the $825 billion as a multi-year commitment to operations and infrastructure, not as a short-term cash outlay.
Going forward, the key question will not only be the size of the investment, but also how effectively the infrastructure translates into growth for AI services and Meta’s existing businesses, and whether the company can maintain profitability while absorbing the cost burden. The figures and disclosure details were compiled based on a Bloomberg report carried by NewsPim.