[P] Public evidence — disclosed in filings, press releases, or verified reporting. [D] Derived analysis — framework-applied inference from public inputs. [I] Illustrative assumption — chosen to demonstrate mechanics, not a claim about the target.
The deal
CoreWeave $2.6bn delayed-draw term loan facility entered July 28, 2025 (referenced here as DDTL 3.0, using framework-author sequencing convention). SEC filing crwv-20250728 [source].
Note on disambiguation [D]: two separate CoreWeave DDTL facilities near the $2.6bn size exist in the market. This case describes the 2025 secured facility maturing August 21, 2030 at SOFR + 4.00% (SEC filing crwv-20250728). A separate 2026 CoreWeave facility of approximately $2.6bn priced at SOFR + 5.50% appears in a subsequent SEC filing (ex991pr on the CoreWeave EDGAR index) and is not the subject of this case.
Key terms [P]:
- Facility amount: $2.6bn
- Interest rate: SOFR + 4.00% (not SOFR + 5.50%, which appears elsewhere in CoreWeave's financing history)
- Undrawn commitment fee: 0.50%
- Maturity: August 21, 2030
- Principal amortisation: monthly from April 2026
- Rating: Ba2 / BB+
- Structure: unconditionally guaranteed by CoreWeave (parent guarantee), secured by substantially all subsidiary assets per SEC filing crwv-20250930 [source]
- Use of proceeds: primarily to finance capex required to perform a specific customer contract (July 2025 announcement identified the customer as OpenAI for a long-term agreement)
Note on collateral counterparty [D]: earlier drafts of this case referenced Microsoft, Meta, and Nvidia as DDTL 3.0 collateral counterparties. That was analytical shorthand for CoreWeave's broader customer relationships and has been corrected. The specific customer contract this DDTL supports is disclosed in the July 2025 announcement.
AI Power Chain scorecard: underlying asset risk
The scorecard runs against the underlying infrastructure the DDTL is financing (compute capacity delivered from CoreWeave's colocation footprint across multiple third-party providers), not against any single greenfield asset.
| Layer | Score band | Basis |
|---|---|---|
| Layer 1 Grid interconnection | MEDIUM | Portfolio-averaged across colocation providers (CoreWeave leases capacity rather than developing greenfield). |
| Layer 2 Facility distribution | MEDIUM | High-density deployment; 800V DC transition is a 2026-2027 roadmap for greenfield, not a claim about existing colocation capacity. |
| Layer 3 Rack + wide-bandgap | MEDIUM-HIGH | SiC concentration exposure at the moment Wolfspeed emerged from Chapter 11 (September 2026). Multi-vendor qualification programme partially mitigates. |
| Layer 4 On-package power | HIGH | Specialist GPU cloud operators concentrate MPS exposure across the fleet. |
| Layer 5 Thermal | MEDIUM | Post the Ecolab/CoolIT and Eaton/Boyd consolidation vendor set. |
| Layer 6 Modular | MEDIUM | Colocation-inherited rather than sponsor-controlled. |
| Aggregate physical [D] | ~60 of 162 (MEDIUM band, ~37%) | Above the LOW/MEDIUM threshold; below the MEDIUM/HIGH threshold. The physical side is not what makes this case interesting. |
Investment Layer scorecard: DDTL 3.0 credit structure
The Investment Layer scorecard here evaluates the DDTL 3.0 structure itself as a credit facility, not a sponsor thesis in the traditional sense. This is a slightly different application of the framework: the "thesis" is the credit structure's ability to survive contract-life-versus-debt-tenor mismatch.
| Section | Score band | Basis |
|---|---|---|
| Section 1 Thesis coherence | MEDIUM | Specialist GPU cloud sponsorship playbook is genuinely novel; parent-guarantee mechanism makes recourse structure coherent. |
| Section 2 M&A defensibility | MEDIUM | Organic-build risk from hyperscaler internalisation is real over the debt tenor. |
| Section 3 Downside coverage | HIGH | Hyperscaler capex pause scenario cuts directly through the collateral base. |
| Section 4 Cross-framework linkage | MEDIUM | Physical DD is portfolio-inferred rather than site-by-site. |
| Section 5 Capital structure | HIGH | SOFR + 4.00% pricing reflects the market's read of the structural risk. Debt tenor (5 years) is meaningfully longer than typical customer contract tenor (industry norm 2-3 years, specific DDTL 3.0 supporting contract tenor not disclosed). |
Uncontracted-principal veto verdict: AMBER
Verdict: AMBER, not RED.
Legs of the 2x2:
- Uncontracted principal leg [D]: at the 2030 maturity, a meaningful share of outstanding principal would depend on re-leasing, refinancing, residual value, or contract renewal not contracted today. The exact percentage depends on the specific customer contract term (not fully disclosed in the SEC filing) and the amortisation schedule. Under the framework's typical AI-infra parameters (2-3 year customer contract, 5-year debt tenor, monthly amortisation from Y1), the share sits above the 40% framework threshold.
- Economic repayment exposure leg [P]: the facility is unconditionally guaranteed by CoreWeave (parent guarantee per SEC filing crwv-20250930). CoreWeave carries a Ba2 / BB+ corporate credit rating. The parent-recourse mechanism absorbs the tail risk.
Because one leg (parent recourse) is absorbed even though the other leg (uncontracted principal) is high, the verdict is AMBER, not RED. The wired combination that would fire the veto requires BOTH legs simultaneously (over 40% uncontracted AND economic repayment exposure absent). CoreWeave's DDTL 3.0 fails only the first test, not the second. The market rated and priced the structure at Ba2/BB+, which is the market's own read that the wired combination does not fire.
2x2 output and IC read
RESTRUCTURE Merchant Physics Play, with AMBER veto flag, not RED. The distinction matters. RESTRUCTURE means the deal is doable at appropriately-structured credit terms. A RED PASS would mean walk away.
The framework's teaching point: DDTL 3.0 is the market-tested case for how a rated-parent structure absorbs an uncontracted-principal exposure that would fire the veto in a thin-equity SPV. This is why CoreWeave sits as the anchor amber calibration example across the framework and the tools.
Corollary for the $8.5bn IG-rated facility (framework-author sequencing: DDTL 4.0): a different absorption mechanism operates via a different leg. The facility is non-recourse except for customary carve-outs but supports previously contracted cloud services. The contracted leg absorbs where the recourse leg would in DDTL 3.0. Both are AMBER. Neither is RED. The framework's 2×2 architecture is what allows the same veto to render two different absorption mechanisms as parallel amber cases rather than treating one as a workaround for the other.
Assessment basis
- Transaction date: July 28, 2025 (SEC 8-K)
- Information cutoff: Q3 2025 10-Q filing (September 2025)
- Public evidence: SEC filings crwv-20250728 and crwv-20250930, CoreWeave July 2025 announcement
- Derived assumptions: Framework-applied per-layer physical scoring, framework-applied per-section thesis scoring, veto AMBER verdict (one leg absorbed per parent-guarantee), uncontracted-principal share derivation (depends on assumed customer contract tenor)
- Unavailable information: Specific DDTL 3.0 supporting-customer contract tenor, exact amortisation waterfall, cash-sweep trigger definitions, per-site physical DD
- Framework version: v2.3.2
Tools referenced
- Investment Layer sponsor thesis scorecard (walk the veto question directly)
- The Investment Layer framework (canonical)
