Mostly True
Meta plans to expand its GPU inventory to 1 million units, a scale of AI infrastructure buildup that, based on available evidence, largely holds up as accurate.
Meta has been publicly documenting its AI hardware expansion, including images of its Prometheus data center project published on its engineering blog showing Nvidia GB300-based systems. The 1 million GPU figure refers to the company's stated goal for its total GPU count across its data centers.
The plan is part of Meta's broader push into large-scale AI training infrastructure, with the Prometheus cluster positioned as a flagship build for training frontier LLMs.
The 1 million figure represents a target rather than a confirmed current inventory, and the timeline for reaching that scale has not been fully detailed. While the company's buildout trajectory and public disclosures support the direction of the claim, the exact final count and date remain dependent on execution.
Verdict: Mostly True