Revenue FY2025
$5.13B
+168% YoY (FY24: $1.92B)
RPO — Backlog
$60.7B
12× revenue · FY24: $15.1B
Net Loss FY2025
$(1.17)B
EPS Diluted: $(2.81)
Total Debt
$21.4B
FY24: $7.9B · +170% YoY
P&E, Net
$30.6B
+157% YoY · GPU + Data Centers
Active Power
>850 MW
3.1 GW total contracted
Customer A (Microsoft)
67%
of total FY2025 revenue
Analytical Framework
6D
Six-dimensional review
Market Cap
$61.98B
StockAnalysis · Apr 2026
P/E
N/A
TTM net loss · ratio not meaningful
EV/EBITDA
31.37
TTM · StockAnalysis
Gross Margin
69.38%
TTM · StockAnalysis
1 Company Profile
CoreWeave was founded in 2017 in Livingston, NJ, by three former capital-markets traders (Michael Intrator, Brian Venturo, Brannin McBee). It pivoted from a low-cost GPU provider for crypto mining (2017-2019) to a GPU cloud for AI training (2020+). Today it positions itself as a vertically integrated AI hyperscaler — building the stack from the gigawatt-level data center upward through Kubernetes orchestration.
The product: a purpose-built GPU cloud, not a general-purpose cloud. CoreWeave runs exclusively on NVIDIA GPUs (GB200, GB300, RTX Pro 6000) with NVIDIA InfiniBand and Spectrum-X networking. 98% of FY25 revenue came from committed take-or-pay contracts, weighted-average ~5 years.
IPO: March 2025 — priced at $40/share, $1.4B net proceeds. 20-for-1 stock split. Class B with 10 votes/share is held entirely by the three co-founders (auto-converts in 7 years or upon Mr. Intrator's departure). NVIDIA invested $2B in CRWV Class A at $87.20/share in January 2026 [Note 17 Subsequent Events].
2 Financial Performance — 3 Years
| Metric | FY2025 | FY2024 | FY2023 |
|---|
| Revenue | $5,131M | $1,915M | $229M |
| YoY Growth | +168% | +737% | — |
| Gross Margin | 71.7% | 74.3% | 69.9% |
| Operating Income (Loss) | $(46)M | $324M | $(14)M |
| Interest Expense (net) | $(1,229)M | $(361)M | $(28)M |
| D&A | $2,454M | $863M | $103M |
| Net Loss | $(1,167)M | $(863)M | $(594)M |
| SBC (% of Revenue) | $630M (12.3%) | $31M (1.6%) | $15M |
D&A of $2.45B + SBC $630M + Op Loss $(46M) implies an unstated Adjusted EBITDA of ~$3.0B, but CoreWeave does not present a formal Adjusted EBITDA reconciliation in this 10-K.
3 Balance Sheet & Capital Structure
| Item | 12/31/2025 | 12/31/2024 |
|---|
| Cash + Restricted Cash | $4.16B | $2.04B |
| Property & Equipment, Net | $30.56B | $11.92B |
| Goodwill (Weights & Biases) | $1.10B | $0.02B |
| Total Assets | $49.30B | $17.83B |
| Total Debt | $21.37B | $7.93B |
| Operating Lease Liabilities | $8.20B | $2.60B |
| Total Liabilities | $45.97B | $16.53B |
| Stockholders' Equity | $3.34B | $(0.41)B |
| Liquidity (Cash + ST + Undrawn) | $6.86B | — |
Debt structure (11 facilities, Note 10): DDTL 1.0 ($1.55B @ 15%), DDTL 2.0 ($5.04B @ 10%), DDTL 2.1 ($2.74B @ 9%), 2030 Notes ($2.0B @ 9.25%), 2031 Notes ($1.75B @ 9.0%), 2031 Convertibles ($2.59B @ 1.75%), OEM Equipment Financing ($4.17B @ 10%), Revolver ($1.0B), Magnetar Loan ($273M @ 12%), and others. Off-balance-sheet leases: $38.5B+ in operating leases not yet commenced, plus ~$13.5–14.4B for a 393MW site over 16 years (Note 8).
4 Customer Concentration & Major Contracts
The most critical section in CoreWeave's analysis. Customer concentration is not just a "risk" — it is the business structure. Microsoft = 67% of FY2025 revenue and 68% of accounts receivable.
| Customer | FY2025 | FY2024 | FY2023 |
|---|
| Customer A (Microsoft) | 67% | 62% | 35% |
| Customer B | <10% | 15% | 17% |
| Customer C | <10% | <10% | 21% |
Major contractual commitments (RPO of $60.7B as of 12/31/2025):
- OpenAI MSA #1 (March 2025): up to ~$11.9B through October 2030
- OpenAI MSA #2 (May 2025) + September 2025 order form: up to ~$6.5B through May 2031
- Combined OpenAI commitment ceiling: ~$18.4B
- Microsoft MSA (Feb 2023): specific size not disclosed; 67% of FY25 revenue
- Meta order form (Sept 2025): up to ~$14.2B through December 2031
RPO recognition schedule (Note 2): 43% over the first 24 months · 38% in months 25–48 · 19% in months 49–84.
5 Analytical Lens — The Questions We Ask
In professional company analysis, the question is not "is this good?" but rather "through which lenses must this company be examined so that we do not miss what matters most?" At Bakshi Finance, every analysis passes through six lenses.
This framework is intended to structure analysis, not to produce an investment conclusion.
Growth
Revenue +168% YoY — but driven by Microsoft. Is OpenAI/Meta ramp on schedule?
Profitability
Gross margin 71.7% but Net Loss $(1.17)B due to interest expense. Where is "real" cash earning power?
Leverage
$21.4B debt + $38.5B off-balance-sheet leases. Effective rates 6%-15%. Is debt service sustainable?
Competitive Position
Microsoft is largest customer AND a competitor (Azure). NVIDIA exclusivity — strength or single-point dependency?
Management Quality
Class B (10 votes/share) entirely with founders. Capital allocation is structurally aggressive — execution track record?
Business Complexity
Class action lawsuits (Note 9), pending NVIDIA Rubin deployment, capex/D&A treadmill — multi-dimensional risks.
6 Scenario Framework
Scenarios are descriptive, not predictive. They outline possible conditions, not expected outcomes.
These scenarios carry no probability assessment, no preferred direction, and no expectation regarding which, if any, will materialise.
Constructive Scenario — diversification + debt servicing on track:
Microsoft revenue mix declines to 40-50% by FY27 (driven by OpenAI/Meta ramp); cash interest stabilises as a % of revenue; NVIDIA Rubin deployment on schedule; CoreWeave begins disclosing Adjusted EBITDA.
Base Scenario — current trajectory with friction:
Microsoft remains 60-70% of revenue through 2026; OpenAI ramp delayed 1-2 quarters but materialises; debt servicing on schedule without distress refinancing; NVIDIA relations stable; net loss narrows but persists.
Adverse Scenario — contract or debt structure breaks:
Microsoft reduces or renegotiates MSA; OpenAI seeks contract modifications under financial pressure; NVIDIA shifts supply to hyperscalers; DDTL 1.0 refinancing in 2028 at higher rates; class action verdict against the company; environmental rate hikes make cash interest unsustainable.
Scenarios describe conditions, not forecasts. There is no preferred direction and no probability assessment expressed in this framework.
7 How to Think About This Company
CoreWeave is not a software company and not a cloud company in the classic sense — it is a capital allocation business on AI infrastructure, with an aggressive debt structure matched by an equally aggressive contract structure. The entire structure rests on one arbitrage: an investment-grade-equivalent customer (Microsoft) signs a long-term contract; CoreWeave borrows against that contract; buys NVIDIA GPUs; delivers the service; and pockets the spread. This is not "cloud" in the AWS sense.
The thesis core: $5.13B revenue (+168% YoY), RPO of $60.7B (12× annual revenue — the largest such ratio in any public company we track). 98% of revenue is multi-year take-or-pay, meaning the customer pays whether they use or not. The classic operational risk of "demand drop" is contractually muted — the risk migrates to counterparty risk: if Microsoft, OpenAI, or Meta encounter financial stress, the on-balance asset ($30.6B GPU + $8.2B leases) becomes a giant warehouse with unclear takers.
Where the analysis may go wrong — first error: reading $60.7B RPO as locked-in revenue. It is not. RPO is contractual entitlement conditional on (a) contract validity continuing; (b) CoreWeave delivering capacity (subject to power, leases, GPU delivery); (c) counterparty financial standing. If, say, OpenAI struggles to meet obligations in 3 years, RPO does not "close" — it is reduced or cancelled. The correct read: $60.7B is a ceiling, not a floor.
Second error: assuming a Net-Debt/EBITDA of ~6.5× is comparable to a normal levered company. CoreWeave's EBITDA includes $2.45B of D&A on GPUs that must be replaced in 4-6 years or the service stops. So the "real" cash available for debt service is EBITDA minus replacement capex — and in this category capex can equal 100-150% of D&A.
Capital allocation — different chapter than NOW or MSFT. In FY2025: $10.31B in capex against $3.06B of operating cash flow. The $7.25B gap was financed by $11.83B of new debt + $1.49B IPO + $11.15B in OEM-supplied financing (not "debt" but supplier-supported credit). This is not a cash-generating business — it is a cash-consuming business building a physical asset. Analysis must separate "is the company profitable" (no, and not soon) from "is the capital structure stable" (depends on contracts, rates, NVIDIA roadmap).
Risks that could break the thesis: (1) Microsoft reduces or terminates MSA; (2) OpenAI cannot meet obligations; (3) NVIDIA shifts supply preference to hyperscalers; (4) DDTL 1.0 refinancing in 2028 at higher rates; (5) class action verdict.
Open litigation: January 12, 2026, putative class action (Raymond Masaitis v. CoreWeave) filed in District of New Jersey alleging false/misleading statements (Sec. 10(b)/20(a)). Two derivative actions filed February 10, 2026 [Note 9]. Company disputes the claims.
What separates professional analysis from a "tip": A tip says "AI stock". Professional analysis asks why 67% Microsoft? How does DDTL structure differ from a normal corporate loan? What if the Fed doesn't cut rates in 2026? How does the $11.15B "OEM financing" affect risk structure? The answers are in Note 10, Note 8 (off-balance leases), and Item 1A.
The difference between surface-level analysis and professional thinking often lies in the variables that are not immediately visible.
This site does not participate in the investment decision. The decision belongs to the client.
8 Sources & Data
| # | Source | Date | Type |
|---|
| 1 | CoreWeave 10-K FY2025 | March 2026 | Official — SEC EDGAR |
| 2 | CoreWeave 10-Q Q3 2025 | November 2025 | Official — SEC EDGAR |
