Meta to Sell USD 2 Billion in Data Centre Assets to Offset AI Infrastructure Costs
- tech360.tv
- Aug 4, 2025
- 2 min read
Meta Platforms plans to sell USD 2 billion worth of data centre assets as it seeks external partners to help fund its growing artificial intelligence infrastructure.
The move, disclosed in a quarterly filing on Thursday, marks a shift in strategy as the company looks to share the financial burden of building and powering data centres for generative AI.
Meta said it approved a plan in June to dispose of certain data centre assets and reclassified USD 2.04 billion in land and construction-in-progress as held-for-sale.
These assets are expected to be contributed to a third party within the next 12 months for co-developing data centres.
The company did not record a loss on the reclassification, which values the assets at the lower of their carrying amounts or fair value less costs to sell.
As of June 30, Meta’s total held-for-sale assets stood at USD 3.26 billion.
Chief Financial Officer Susan Li said on a post-earnings call Wednesday that Meta is exploring ways to work with financial partners to co-develop data centres.
While Meta still expects to fund most of its capital spending internally, Li said some projects could attract significant external financing and offer flexibility if infrastructure needs change.
The company has not finalised any transactions yet, but the filing suggests plans are progressing.
CEO Mark Zuckerberg has committed to investing hundreds of billions of dollars into AI data centre “superclusters,” describing one as covering a significant part of Manhattan’s footprint.
On Wednesday, Meta raised the lower end of its annual capital expenditure forecast by USD 2 billion, now projecting between USD 66 billion and USD 72 billion.
The increase follows stronger-than-expected advertising sales, driven by AI-enhanced targeting and content delivery, which executives said are helping offset rising infrastructure costs.
Meta to sell SGD 2.75 billion in data centre assets
Move aims to share AI infrastructure costs with external partners
Assets expected to be transferred within 12 months
Source: REUTERS
