CoreWeave Secures $21B Additional Meta Deal, Pushing Revenue Backlog to $87.8B
Summary
- • Meta commits an additional $21B to CoreWeave, extending the partnership through December 2032
- • CoreWeave's total revenue backlog reaches $87.8B following the announcement
- • Meta and OpenAI together represent roughly 65% of CoreWeave's guaranteed revenue
- • CoreWeave reported $5.13B in 2025 revenue, up 2.7x year-over-year, with a $1.17B net loss
Details
$21B new Meta deal via SEC 8-K
CoreWeave filed an 8-K with the SEC announcing Meta's additional $21B AI compute commitment through December 2032, following the original $14.2B deal signed in September 2025 that runs through December 2031 — two separate contracts with staggered end dates.
Backlog hits $87.8B; top-2 customers = ~65%
Meta now represents 40.1% of CoreWeave's $87.8B backlog. OpenAI holds $22.4B in total commitments (~25.5% of backlog). Together, just two customers control nearly two-thirds of guaranteed revenue — a concentration risk that mirrors 2024, when Microsoft alone was 62% of CoreWeave's $1.92B revenue and top-3 customers were 77%.
2025 revenue $5.13B, net loss $1.17B
CoreWeave grew revenue 2.7x in 2025 to $5.13B, with an operating loss of only $46M — but stock compensation and other charges pushed the net loss to $1.17B. The company holds just under $4B in cash and spent $14.9B on capital equipment in 2025.
43 datacenters, ~600K GPUs, 850MW active
CoreWeave operated 43 datacenters at year-end 2025 with an estimated 600,000 GPUs — primarily Nvidia Hopper H100s and H200s, with a growing share of Blackwell B200s and B300s — and 850 megawatts of active power. The company has 3.5 megawatts of total power under contract.
Neocloud model: rent new GPUs to titans, recycle old iron
CoreWeave's business model keeps ~24 named customers on long-term GPU leases — hyperscalers and frontier labs occupy the newest hardware while older capacity cascades down to enterprises, governments, and startups. The model requires very high utilization, as idle GPUs burning electricity without revenue are the primary financial risk.
Key facts, financials, and context from CoreWeave's latest Meta deal
What This Means
CoreWeave's ballooning backlog signals that hyperscalers and frontier AI labs are locking in GPU capacity years in advance, treating neocloud compute as strategic infrastructure rather than a spot commodity. However, the extreme customer concentration — two clients representing nearly two-thirds of $87.8 billion in backlog — means CoreWeave's financial health is tightly coupled to the sustained AI investment appetites of Meta and OpenAI. If either scales back spending or shifts toward more in-house capacity, CoreWeave faces significant revenue risk despite its impressive headline numbers.