The situation
A compute-infrastructure operator is evaluating alternative ways to finance a large-scale equipment deployment through a ring-fenced special-purpose vehicle. The model spans a staged capex draw schedule, two delayed-draw term loans, parent and preferred equity contributions, operating cash flow, a required cash reserve, and residual equipment value at the end of the investment horizon.
The central credit question is how the tenor and durability of customer demand — long-dated contracted offtake versus a pool of shorter-duration contracts — should change the capital structure, credit support, amortization profile, and allocation of cash through the waterfall.
Two structures. Divergent risk paths.
The first structure pairs parent equity with a preferred-equity sponsor, long-dated contracted compute demand, and an incremental share pledge as additional credit support. The second relies on parent-funded equity and a diversified pool of shorter-tenor customer contracts, with different pricing, utilization, and cost-of-debt assumptions. Both model staged delayed-draw funding and full debt amortization across the investment period.
Parent common equity alongside a preferred-equity sponsor
Two delayed-draw term loans funded against the capex draw schedule
Longer-dated contracted compute demand
Contracted cash flows plus an incremental share pledge
The critical diligence question
Validate the share pledge’s collateral value and enforceability, the priority of preferred distributions and accrued unpaid dividends in the waterfall, and whether pledge sizing references equity ownership or total capitalization.
Why the analysis demands a multi-agent system
The answer is distributed across the transaction diagram, sources and uses, operating assumptions, facility schedules, distribution priorities, and return sensitivities. Each exhibit can appear internally consistent while a cross-document inconsistency remains unresolved. A planner agent decomposes the review into dependent tasks, specialist agents extract and reconcile each component with citation-level provenance, and a verifier agent tests the pieces against one another in a single working record.
Document-intelligence agent
Applies layout-aware parsing and table extraction to the deck, builds a schema-validated assumption register, and links every material term to its source page.
Cash-flow analyst
Rebuilds sources and uses, capex draws, facility utilization, DSCR, cash sweep, and distributions on a deterministic calculation engine — arithmetic is never delegated to the language model.
Adversarial credit reviewer
Stress-tests collateral, maturity alignment, lien priority, and downside cases, and routes open questions to the investment team and counsel with confidence scores.
Investment-memo editor
Assembles a side-by-side decision brief that separates source-stated assumptions from derived figures and unresolved issues.
The questions that deserve scrutiny
- Funding sufficiency: do committed delayed-draw availability and retained cash cover each tranche of the capex program while preserving the required liquidity reserve?
- Tenor alignment: how do facility-specific maturities reconcile with the overall investment horizon and the assertion that debt fully amortizes within the period?
- Collateral definition: is the pledged share amount expressed as a percentage of total capitalization, a collateral value, or an ownership interest? These are distinct concepts that must be defined consistently before credit support can be sized.
- Waterfall mechanics: how do senior debt service, the cash sweep, preferred distributions, accrued unpaid dividends, management fees, and terminal proceeds interact in each period?
- Downside resilience: what happens to DSCR and projected sponsor IRR and MOIC if utilization or pricing compresses while infrastructure costs remain largely fixed?
A decision package with provenance attached
The proposed workflow produces an assumption register, a funding and debt-service reconciliation, a structure-by-structure comparison, a prioritized issue list, and a draft investment brief. Every conclusion is traceable to its inputs through citation-level provenance, and every unresolved item is surfaced rather than smoothed over.
The value lies in clarity about where the structures diverge and which assumptions require confirmation. The source material does not establish that a transaction closed, that projected returns were achieved, or that an AI system produced a measured productivity gain.
Investment judgment stays with people
The investment team approves model assumptions and conclusions at explicit human-in-the-loop checkpoints. Counsel confirms collateral enforceability, contractual priority, and structural treatment. No source document or confidential model is available for download from this website.