Risk assessment
Six material risk categories. Likelihood × impact ratings calibrated to the tier selected; figures assume Tier 4 exposure as the upper bound.
The participation framework carries six material risk categories. Likelihood × impact ratings are qualitative, calibrated against the tier selected; the figures below assume Tier 4 exposure as the upper bound.
| Risk | Likelihood | Impact | Mitigant |
|---|---|---|---|
| HUMAIN execution risk. The HUMAIN programme is at framework engagement stage with a proposed proof-of-concept structure. Full USD 10 billion deployment is not committed. Tier 3 and Tier 4 economics depend materially on HUMAIN realisation. | Medium | High | Probability assumptions in the model reflect the sequential commitment structure (34 per cent realisation probability on the full programme). Probability-weighted IRR provides a transparent view of expected returns. POC outcome by end of 2026 will inform the binary on Phase 2 and Phase 3. |
| Global Pipeline execution risk. The USD 25 billion Global Pipeline is expected but not signed. Multi-year realisation horizon. Tier 4 carries the most exposure. | Medium-high | Medium | Probability assumption of 25 per cent reflects conservative pricing of unsigned pipeline. Causara plays a similar role in Global Pipeline deals to CEI within HCC, leveraging the same capital origination and structuring capabilities. |
| GPU market and technology risk. Rental rate compression as supply increases; technology obsolescence as next-generation GPU architectures arrive. | Medium | Medium | Clusters under contract lock in revenue at fixed rates. Where clusters are uncontracted, Brokkr maintains greater than 90 per cent utilisation across all on-demand deployments. Conservative residual value underwriting at 40 per cent of MSRP. |
| Counterparty and structural risk. Offtake counterparty default; failure of one of the co-GP parties to perform; misalignment between Kai's interests and the broader Causara platform. | Low | High | Hydra requires substantial offtake prepayment (10 to 20 per cent of contract value), which naturally filters for high-quality end users. The signed Co-GP document establishes LP admission rules and allocation rights. The deferred payment mechanism ties Kai's interest to long-term platform performance. |
| Valuation methodology risk. The probability assumptions and discount rates used in the NPV methodology are calibrated; alternative assumptions yield different valuations. | Medium | Medium | The accompanying Excel model exposes all assumptions as inputs. Sensitivity tables provide direct visibility on the impact of alternative MOIC scenarios and HUMAIN probability. Tier valuations sit at 16 to 37 per cent of risk-adjusted NPV, well below the established-manager range, providing cushion against assumption changes. |
| Key person and platform continuity risk. Causara's analytical capability is concentrated in the founding team. Departure of key personnel would affect platform value, particularly for Tier 4. | Low | Medium | Causara's value is increasingly platform-encoded (Brokkr operating data, Causara Intelligence proprietary methodology, signed Co-GP framework). The contribution matrix structure means GP economics are programmatic rather than personality-dependent. The 24-month lockup on Kai's residual stake further aligns long-term commitment. |
Six material risks, with likelihood, impact and mitigant.
Likelihood on the vertical axis, impact on the horizontal. Severity = L × I.
Phased exposure; Series B reached MOU stage with two anchor LPs; tier 3/4 deferred mechanism absorbs slippage.
Pipeline diversified across Luxembourg / Singapore feeders; modelled at 60% realisation probability.
Senior debt covenants; offtaker pre-commitment reduces residual price risk; sensitivity bands published.
Investment-grade offtakers only; legal review of every SPA before signing.
Live model published; every assumption editable; benchmarked against four observable GP-stake transactions.
Three-person bench at CEI; six-month notice; documented operating cadence; co-GP structure mitigates single-point dependency.