Case Studies · Part 10 of 12
PROVENANCE TAGS
[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

Blackstone Infrastructure Partners with Canada Pension Plan Investment Board acquired AirTrunk in September 2024 [P]. Enterprise value approximately A$24bn, or roughly US$16.1bn per Blackstone's own announcement [source]. CPP Investments took approximately a 12% stake. Blackstone describes it as its largest-ever Asia-Pacific transaction, and contemporary transaction coverage places it among the largest data-centre transactions globally at announcement.

Capacity at deal close [P]: AirTrunk had more than 800 MW of committed capacity to customers plus land holdings supporting more than 1 GW of additional future growth per Blackstone's announcement. The platform has since expanded, but transaction-date facts (rather than later-state facts) are what the underwriting was evaluated against.

Operating footprint [P]: Australia, Japan, Hong Kong, Singapore, and Malaysia. Anchor tenants are named hyperscalers on long-tenor triple-net capacity leases (specific counterparty mix not disclosed at the individual site level).

AI Power Chain scorecard: physical asset risk

LayerScore bandBasis
Layer 1 Grid interconnectionLOWMature APAC grid regimes on the operating pipeline; LGIA-equivalent signed on committed capacity. Land bank supports the >1 GW growth pipeline but each new-build site carries its own interconnection risk.
Layer 2 Facility distributionLOW-MEDIUMOperating sites use AC distribution with roadmap alignment for higher-density liquid-cooling deployment. Note: 800V DC is a 2026-2027 roadmap architecture, not a 2024-deal-date reality.
Layer 3 Rack + wide-bandgapLOW-MEDIUMMulti-vendor SiC qualification programme applied per platform (specific vendor mix not publicly disclosed at the site level).
Layer 4 On-package powerMEDIUMStandard merchant vendor mix (MPS with Vicor as documented substitute at the tenant equipment tier; platform-agnostic at the operator layer).
Layer 5 ThermalLOW-MEDIUMDLC-ready designs; post the 2024-2026 vendor consolidation wave (Eaton/Boyd, Ecolab/CoolIT) the merchant thermal vendor set is compressed but AirTrunk sits inside the consolidator ecosystem.
Layer 6 ModularLOWAPAC prefab partnerships documented; commissioning schedules include contingency.
Aggregate physical [D]~35 of 162 (LOW-MEDIUM band, ~22%)Below the LOW/MEDIUM threshold of ~40. The deal scores at the clean end of the framework's physical band.

Scoring methodology: individual layer scores are the framework author's assessment against public disclosures at deal-date. A sponsor running the scorecard with private DD access would produce a different per-layer distribution but the aggregate band would likely land in the same LOW-MEDIUM territory.

Investment Layer scorecard: sponsor thesis risk

SectionScore bandBasis
Section 1 Thesis coherenceLOWDeveloper-layer buyout playbook with clear precedents (Blackstone/QTS 2021, KKR+GIP/CyrusOne 2022). Blackstone Infrastructure and CPP execution track record supports mid-teens IRR on long hold. Regime-dependence question flags exposure to AI capex normalisation, but base-case demand thesis is diversified across 5 jurisdictions.
Section 2 M&A defensibilityLOW>800 MW committed capacity puts the platform above the 500 MW threshold that attracts strategic and sovereign co-investor interest at exit.
Section 3 Downside coverageLOW-MEDIUMContracted revenue with named hyperscaler tenants across jurisdictions; capex pause exposure exists but is diversified.
Section 4 Cross-framework linkageLOWBlackstone Infrastructure's DD process is well-documented in prior transactions.
Section 5 Capital structureLOWLeverage sized against contracted hyperscaler cash flow; institutional infrastructure and pension capital.
Uncontracted-principal vetoGREEN [D]Public disclosure supports the inference that contracted hyperscaler revenue covers debt service through the operating asset life; the recourse structure is clean. Explicit debt structure and covenant detail is not in the public record for this transaction.
Aggregate thesis [D]~28 of 105 (LOW band, ~27%)Below the developer-layer playbook PASS threshold (~25) with veto GREEN.

2x2 output and IC read

Framework 2×2 · AirTrunk position
AirTrunk on the framework 2x2 STRUCTUREPaper FortressClean asset, fragile thesis PASSDistressedBoth blocks show material exposure UNDERWRITEPremier institutional AI assetPhysical solid, thesis holds RESTRUCTUREMerchant Physics PlayCoherent thesis, ambitious asset AirTrunk · Sep 2024 HIGH LOW 40% HIGH INVESTMENT THESIS RISK → PHYSICAL ASSET RISK (POWER CHAIN) → Score: physical ~35 of 162 (LOW-MEDIUM, ~22%) · thesis ~28 of 105 (LOW, ~27%) Veto: uncontracted-principal GREEN · wired 2×2 does not fire
Both scores sit well inside the 40% LOW threshold. AirTrunk anchors the framework's UNDERWRITE reference point.
Veto 2×2 · AirTrunk position
AirTrunk on the uncontracted-principal veto 2x2 AMBERRecourse absent, tail lowOne leg high; one absorbed REDVeto firesBoth legs high & unabsorbed GREENBoth legs lowContracted tail; recourse present AMBERTail high, recourse absorbsRated-parent absorbs mismatch AirTrunk · Sep 2024 HIGH LOW 40% HIGH ECONOMIC REPAYMENT EXPOSURE ABSENT → UNCONTRACTED PRINCIPAL AT MATURITY → Leg 1: uncontracted principal LOW — leverage sized against contracted hyperscaler CF across 5 jurisdictions Leg 2: recourse PRESENT — institutional infra + pension capital, cross-jurisdictional operating asset. Veto GREEN.
Both legs low. Contracted hyperscaler CF plus institutional recourse; the wired 2×2 does not fire.

UNDERWRITE Premier institutional AI asset. Physical asset is solid, thesis stands up under stress, no structural veto triggered.

The transaction illustrates the level of physical and investment diligence the framework would regard as appropriate for a mature institutional AI infrastructure asset. It is the reference case for what UNDERWRITE looks like on the framework's calibration. The number of AI infrastructure transactions in 2024-2026 that would actually score UNDERWRITE on the framework is small, which is what makes AirTrunk a useful anchor point.

Assessment basis

  • Transaction date: September 4, 2024 (announcement)
  • Information cutoff: August 2026 (transaction-date facts only; later platform expansion excluded)
  • Public evidence: Blackstone press release, CPP Investments confirmation, contemporary DCD/Data Centre Dynamics coverage
  • Derived assumptions: Aggregate scores, per-layer bands, veto verdict all derived from framework applied to public disclosures
  • Unavailable information: Specific tenant contract terms, per-site debt allocation, covenant detail, individual-site LGIA status
  • Framework version: v2.3.2

Tools referenced