CRWV
CoreWeave (IPO Mar 2025)
Summary
What they do:
CoreWeave operates the largest independent GPU cloud platform, renting AI compute infrastructure to model developers, enterprises, and hyperscalers across 40+ data centers with 850MW of active capacity — the pure-play alternative to AWS, Azure, and GCP for GPU-intensive training and inference workloads.
Why they matter:
CRWV is the fastest cloud platform in history to reach $5B in annual revenue, with an $87.8B revenue backlog (40% Meta) and NVIDIA's strategic backing ($2B investment) — the independent GPU cloud that hyperscalers themselves rent from when their own capacity runs short.
Recent performance:
FY2025 revenue $5.1B (+170% YoY); Q4 2025 revenue $1.57B (+110% YoY); net loss $1.17B. Stock ~$117, market cap ~$61B. April 2026 deal blitz: Meta $21B expansion, Anthropic multi-year agreement, Jane Street $7B commitment ($6B cloud + $1B equity at $109/share).
Our Verdict
The purest public-market AI infrastructure bet — 100% GPU cloud with an $87.8B backlog converting at 170% revenue growth — but $21B in debt at 11% weighted average interest, NVIDIA single-supplier dependency, and 67% Microsoft revenue concentration in 2025 create execution risk that the $61B market cap assumes away.
Structural trends
Structural
66
/ 100
Moat
6/10
NVIDIA partnership + scale + backlog depth
AI Exp.AI Exposure
Pure Play~100% AI
Play Type
EmergingAI Growth
~170% YoY
Rel. Value
74
COMPELLINGPriceLIVE
$117.20
+6.28%
Live via Yahoo Finance · refreshes every 5 min
Market Cap
$61.6B
P/E Ratio
N/A
P/S Ratio
12.0x
52W High
$187.00
52W Low
$33.52
52W Chg
249.7%
Beta
N/A
CoreWeave emerged from an unlikely origin — a cryptocurrency mining operation that pivoted to GPU cloud infrastructure when founder Michael Intrator recognized that the same hardware optimized for mining could serve the exploding demand for AI compute. The company launched its cloud platform in 2020, won NVIDIA's strategic backing (culminating in a $2B investment in January 2026 at $87.20/share), and went public in March 2025 at $40/share in what became one of the most closely watched IPOs of the AI era.
The business model is straightforward: CoreWeave buys NVIDIA GPUs (H100, H200, and now Blackwell/Vera Rubin architecture), installs them in GPU-dense data centers (100+ kW per rack, roughly double traditional cloud density), and rents compute capacity to AI companies under long-term contracts. Customers connect via standard cloud APIs — the same Kubernetes and PyTorch interfaces they would use on AWS or Azure — and spin up GPU clusters on demand. The pricing advantage is typically 30-40% below hyperscaler alternatives because CoreWeave runs no legacy business to cross-subsidize.
The scale has moved fast. CoreWeave ended 2025 with 850MW of active power capacity across 40+ data centers, with 3.1GW of contracted capacity expected online by 2027. The company targets 5+ GW by 2030 through an expanded partnership with NVIDIA. Revenue grew from $1.9B in FY2024 to $5.1B in FY2025 (+170% YoY), making CoreWeave the fastest cloud platform in history to reach that milestone. Management guided FY2026 revenue of $12-13B, with an annualized run-rate of $17-19B exiting 2026 and $30B+ by end of 2027.
April 2026 was transformational. In a six-day span, CoreWeave announced three landmark deals: Meta expanded its commitment to $21B through December 2032 (total Meta commitments now ~$35B), Anthropic signed a multi-year agreement for production-scale Claude inference workloads, and Jane Street committed $6B in cloud services plus $1B in equity at $109/share. These deals pushed the revenue backlog from $66.8B (year-end 2025) toward an estimated $88B. The customer base has broadened — CoreWeave now counts nine of the ten leading AI model providers as customers, including OpenAI and Microsoft — though Microsoft still accounted for 67% of 2025 revenue.
Supply Chain Dependencies
Upstream Suppliers
Downstream Customers
The Catch
CoreWeave's fundamental challenge is that it must grow faster than its debt compounds. At $21.4B in total debt carrying 11% weighted average interest, the company pays roughly $2.4B annually just to service its capital structure — nearly half of FY2025's $5.1B revenue. The $30-35B in planned 2026 capex will add more debt before revenue catches up. Meanwhile, the entire business rests on a single supplier (NVIDIA) whose pricing power can compress CoreWeave's margins at will, a single customer (Microsoft at 67% of 2025 revenue) whose purchasing decisions are existential, and an AI spending cycle that has quadrupled in three years and may not sustain this trajectory. CoreWeave is building the right infrastructure at the right time — but it is doing so with the most leveraged balance sheet in cloud history, leaving almost no margin for execution error.
If They Win
If CoreWeave converts its $87.8B backlog into recurring revenue, refinances its debt stack at lower rates, diversifies beyond Microsoft, and scales to 5+ GW of capacity by 2030, it becomes the fourth hyperscaler — the independent GPU cloud that sits alongside AWS, Azure, and GCP as a default platform for AI workloads. In this scenario, CoreWeave generates $25-30B in annual revenue with 65%+ gross margins and 20-30% operating margins, the NVIDIA partnership evolves into a durable co-dependent relationship (NVIDIA needs CoreWeave as much as CoreWeave needs NVIDIA), and the stock trades at $300-400 as the market values CoreWeave as a $150-200B cloud infrastructure company — the public market's purest expression of the AI infrastructure megatrend.
Others in Run the Workload
Not financial advice. All scores generated via AI algorithms using public data.