Risk register: power, policy and phantom demand head the ranking
Ten risks, ranked, each with its early-warning indicator. Power, policy and phantom demand head the near-term cluster; obsolescence and demand-shift are the slow-burners that break a thesis rather than delay it.
Risk register (1 of 2): power, policy and phantom demand head the ranking
R1 Power availability: the binding constraint
- 44 GW requests vs ~6 GW needed
- 8+ yr core-market waits
- ~25% of pipeline deliverable
- Net-generator cost/schedule on every large project
- AEMO IASR totals
- Substation commitments
- First NSW prohibition decision
R2 Regulatory / policy
- Net-generator legislation (2027)
- HCF reform paused since Nov 2025
- AEMC cost-recovery could apply mid-connection (KWM)
- Exposure draft scope
- Transition relief
- HCF reform terms
The ten risks, in register order; colour as plotted
Power availability
Regulatory / policy
Phantom demand / double-counting
Cost of capital / valuation
Concentration (demand & geography)
Obsolescence (density step-change)
Construction & labour
MoU execution risk
Water & social licence
Inference-at-edge / efficiency shift
Likelihood x impact, inashanu judgement
Phantom demand, valuation and concentration complete the near-term, high-severity cluster (R1-R5)
R3 Phantom demand / double-counting
- ~6 of 7 requested MW won't materialise (Oxford Economics)
- 11.4 GW NSW pipeline inflated vs build-out
- Request withdrawals
- Project silent-period extensions
R4 Cost of capital / valuation
- NXT ~45x EV/EBITDA and AirTrunk 21x contracted price flawless conversion
- -14% month shows fragility
- Stack APAC sale tests exit depth
- REIT IPO pricing
- NXT order-book conversion rate
- Credit spreads on SLLs
R5 Concentration (demand & geography)
- One 555 MW contract = ~40% of 2025 operating capacity
- NSW+VIC pipeline in ~4 suburbs
- CDC = ~40% of Infratil value
- Tenant counterparty ratings
- Precinct-level grid notices
The evidence in numbers
How to read this register
Position on the matrix is inashanu judgement, not measurement: likelihood over the next 24 months against impact on sector economics. Every risk's evidence base is cited in the tables; R1-R5 (pages 67-68) are the near-term, high-severity cluster.
Risk register (2 of 2): obsolescence and demand-shift are the slow-burners the market underprices
Air-cooled legacy stock vs 40-100+ kW/rack AI loads
GPU generation turnover (B200 -> GB300 -> Rubin) shortens economic life
PwC flags obsolescence as key yield driver
Liquid-cooling retrofit capex announcements
Legacy vacancy divergence
R7 Construction & labour
- Sydney/Melbourne costs just below Tokyo/Singapore
- 22% vacancy for senior M&E/cyber engineers
- DCs compete with housing for trades
- Tender price indices
- IDA trade-mobility decisions
R8 MoU execution risk
- OpenAI-NEXTDC is an MoU, not a lease
- AirTrunk SYD4 power commitments short of nameplate
- Anthropic plans unconfirmed
- MoU-to-lease conversions
- SYD4 connection agreement
R9 Water & social licence
- 40 ML/day single-site applications
- Melbourne's west community politics
- DigiCo LAX1 delayed on energy/water concerns (template)
- Council-level objections
- Sydney Water determinations
R6-R9: the evidence in numbers
R1-R5 are timing risks; R6 and R10 are the genuine thesis-breakers
R10 Inference-at-edge / efficiency shift
- If inference migrates to edge/device or tokens-per-watt improves faster than demand grows, training campuses face stranded capacity
- US construction decline is the analogue
- Edge share of new commitments
- GPU utilisation disclosures
- US starts data
Portfolio-level read
Risks R1-R5 are timing risks: they bear on when megawatts arrive, and none of them bears on whether demand exists.
R6 and R10 are the genuine thesis-breakers: both attack the revenue durability of capacity already built.
Monitoring cadence
R1-R5 indicators are quarterly-or-better (AEMO processes, policy drafts, listed marks).
R6-R10 are annual reviews unless an early indicator fires, with the CBRE North America construction series as the standing global tripwire for R10.
Underwrite lease tenor against technology half-life; building life is the wrong benchmark.
Sources
- M3, as at Nov 2025
- IMARC, as at 2026
- The Good Builder, as at May 2026
- Next West Observer, as at Aug 2026
- CBRE via UVRN, as at Jul 2026
- KWM, as at Aug 2026
Related
- How much of Australia's announced data centre pipeline will actually be built?
- What are Australia's new data centre energy rules?
- How can you invest in Australian data centres on the ASX?
- What data is missing on the Australian data centre market?
The full primer is published as an 85-page PDF (download) with the underlying market model as a workbook (download).
This page is one of 30 in Australian data centres: the research, indexed by the question each one answers, with the four pages where the published sources disagree set first.
The research is published by inashanu, a studio that produces fixed-scope market research for Australian private equity, priced and dated before work starts. The sample engagement is this primer in full.
This standard, on your question. Six fixed-price deliverables, A$14,000 to A$45,000 excluding GST, opening at the A$14,000 market map in 10 business days. State the decision and the date it has to be made; the written assessment of whether the evidence base will carry it follows within two business days and carries no charge. Request a brief assessment · the catalogue and its prices.