CoreWeave's $21 billion AI computing deal with Meta is an expansion of an earlier $14 billion agreement, representing a massive commitment from Meta through 2032.
Meta's move signals a major, ongoing investment in training AI models to remain competitive, using both in-house efforts and external cloud providers.
This deal exemplifies a broader trend: companies are turning to "new clouds" like CoreWeave for AI-as-a-service when they cannot meet all demand in-house.
Hyperscalers (e.g., Microsoft, Amazon) are still capacity-constrained, unable to fully meet the explosive demand for AI infrastructure.
The total backlog of cloud/AI commitments across the industry is estimated to be "north of 1 trillion" and potentially as high as $1.5 trillion.
Microsoft alone has over $600 billion in cloud commitments on its balance sheet that need to be converted into sales.
Demand is unprecedented, driven not only by AI model companies but also by enterprises integrating AI into daily operations.
The pace of "tens of billions of dollars" data center deals is accelerating, with major announcements occurring almost weekly.