Data-centre real estate as the only structurally tight major property type, and the two-speed sector it has created

Six alternative branches. These are subjective probabilities from the seed, not observed frequencies or investment recommendations.

Why this trend

Load-bearing assumption

That CBRE's primary-market series - 1.4% vacancy, 7,481 MW under construction, 80.4% preleased - is measuring contracted, building capacity rather than announced pipeline. Every quantity in this set is CBRE's; JLL's 66+ GW figure for the same object differs by roughly ninefold and neither firm publishes its inclusion rule, so no scenario here can be resolved against JLL's series.

Preconditions

Not recorded.

Two-speed sector persists

base · 40% · band 30-50%

CBRE preleasing stays between 70% and 85% in successive semiannual reports and primary-market vacancy stays under 3%. Existing-home sales recover slowly toward 4.5 million as sellers capitulate on price rather than waiting for rates, which the corpus notes does not require the Fed. Office CMBS delinquency grinds down from 11.91% without a clearing event, construction labour stays tight at or near the record-low 3.1% unemployment, and construction input inflation moderates from 8.9% toward 4-6% as the tariff level shift laps out of the twelve-month comparison.

Mechanism

The pace is set by power and permitting rather than by demand, so a demand surplus shows up as a longer queue rather than as a higher build rate. The two economies stay linked through a single labour force: with 325,000 open construction positions and fewer than half of US metros adding construction jobs, additional data-centre construction is substitution rather than addition, which is why total construction spending can fall 3.8% year over year while data centres drive all nonresidential growth.

Preconditions, early indicators and assumptions

Band Width Pp

Calibration Basis

base_rate

Thesis

Data-centre construction continues at a high but decelerating rate as power and permitting bind, while housing volume recovers slowly and office credit grinds down without a clearing event.

Preconditions

Hyperscaler capex guidance narrows rather than being cut
No federal preemption of local data-centre siting
Mortgage rates remain in a 6-7% band

Early Indicators

Indicator
CBRE preleasing share holding between 70% and 85% in successive semiannual reports while total US construction spending stops declining year over year
Source
S-18-09 CBRE North America Data Center Trends; S-18-01 Census Construction Spending
Would Be Visible By
2027-10
Indicator
Trepp office CMBS delinquency between 9% and 12% with the inclusive/headline wedge still above 100bp
Source
S-18-11 Trepp CMBS delinquency reports
Would Be Visible By
2027-08
Indicator
Construction-industry unemployment below 4.5% with job openings still above 250,000
Source
S-18-04 BLS Employment Situation; S-18-14 AGC
Would Be Visible By
2027-09

Affected Industries

What Businesses Should Do

Underwrite non-data-centre construction on labour availability rather than on demand, and price crew access as the binding input. For housing, plan for volume recovery driven by seller price capitulation, not by rate relief.

Assumptions

Text
Hyperscaler capex guidance narrows rather than is cut through 2027
Confidence
medium
Load Bearing
true
Basis
Macro brief and sector 01: Meta narrowed FY2026 capex to $130-145bn rather than raising it after Alphabet's guidance raise drew a 5% share-price fall
If Wrong
Mass moves to downside, and the preleasing indicator fires before vacancy does

Precedence Note

Distinguished from downside by the direction of preleasing share, not by the level of vacancy. Base has preleasing holding above 70%; downside has it falling through 70%.

Would Change Our Mind

CBRE reporting primary-market data-centre vacancy above 5% in any semiannual report.

The freeze breaks without a bust

upside · 12% · band 2-22%

The constraint that releases is the cost of debt, and the actor that releases it is the FOMC. Extensions start curing rather than deferring, which shows up as the 176bp wedge between Trepp's reported 7.86% and inclusive 9.62% delinquency closing. Existing-home sales pass 5 million as the 30-year fixed falls below 6%. Data-centre demand proves durable while frontier-market supply - already 77% of capacity under construction per JLL - relieves pricing pressure without collapsing it, and immigration policy partially reverses, easing crew scarcity.

Mechanism

Lower policy rates reduce the refinancing hurdle on 2021-25 vintage CRE debt struck on an assumption of cheaper money in 2026 that the corpus shows was false. Loans that can refinance stop being modified and start being repaid, which converts deferred loss into realised transaction volume and re-establishes price discovery in the asset classes that currently have none.

Preconditions, early indicators and assumptions

Band Width Pp

Calibration Basis

analogue

Thesis

A rate-cutting cycle restores refinance capacity, existing-home sales recover past 5 million, and the office maturity wall resolves through refinancing rather than foreclosure.

Preconditions

A federal funds target below 3.00% by mid-2027
The 30-year fixed mortgage below 6.00% for a full quarter
No new tariff action re-inflating construction inputs

Early Indicators

Indicator
Trepp's inclusive CMBS delinquency measure converging toward the headline measure by more than 100bp
Source
S-18-11 Trepp
Would Be Visible By
2027-12
Indicator
Freddie Mac PMMS 30-year fixed below 6.00% for eight consecutive weekly readings
Source
S-18-06 Freddie Mac Primary Mortgage Market Survey
Would Be Visible By
2027-09

Affected Industries

What Businesses Should Do

Hold refinancing optionality rather than locking 2026 terms, and prepare listings for a volume unlock that arrives through price rather than through rates - the months-supply figure at 4.9 against 4.0 pre-pandemic says the inventory is already there.

Assumptions

Text
The FOMC cuts rather than holds or hikes through 2027
Confidence
low
Load Bearing
true
Basis
Macro brief: the July 2026 hold was 9-3 with three dissents in favour of a hike, and the 2026-09-16 decision fell one day after the research date and is unresolved
If Wrong
Upside collapses into base; nothing moves to downside, because the freeze persisting is the base case

Precedence Note

Requires the inclusive/headline delinquency wedge to close, not merely the headline rate to fall. A falling headline with a stable wedge is base.

Would Change Our Mind

Two consecutive FOMC meetings with a hike, or the 30-year fixed above 7.25%.

AI capex retrenchment meets the CRE debt overhang

downside · 20% · band 10-30%

The 80.4% preleasing figure falls first, then backlog, then vacancy - in that order, because vacancy moves last in a preleased market. Data-centre debt reprices against an equity market that barely exists: $1.7bn of investment sales in H1 2026 against roughly $27.8bn of high-yield issuance, $4.9bn of single-asset CMBS on CBRE's count, and individual financings like the $16bn Related Digital and Blackstone construction facility that dwarf the entire equity transaction market. The labour crowded out of housing does not return fast enough to matter, and regional banks with CRE concentration take losses at the same time.

Mechanism

The parameter that deteriorates is contracted hyperscaler demand, and the first place it appears in public data is the preleasing share in CBRE's semiannual series. The transmission runs from capex guidance to lease cancellation to backlog erosion to securitisation spreads - and the FOMC has already named leveraged financing of infrastructure buildout as a financial-stability concern, which is the channel that carries it into the banking system.

Preconditions, early indicators and assumptions

Band Width Pp

Calibration Basis

base_rate

Thesis

Hyperscaler capex guidance cuts trigger lease cancellations, preleasing falls before vacancy rises, and data-centre securitisations reprice with no equity bid to clear into.

Preconditions

A guidance cut, not a narrowing, at any of the four largest hyperscalers
Preleasing share falling through 70%
Data-centre ABS or CMBS issuance stalling

Early Indicators

Indicator
CBRE preleasing share falling below 70% in any semiannual report
Source
S-18-09 CBRE North America Data Center Trends
Would Be Visible By
2027-10
Indicator
Data-centre ABS or single-asset CMBS new issuance stalling for two consecutive quarters
Source
S-18-11 Trepp; S-18-21 Data Center Frontier
Would Be Visible By
2027-12
Indicator
Multifamily CMBS delinquency at or above 8.00%, confirming the broader CRE credit channel is transmitting
Source
S-18-11 Trepp
Would Be Visible By
2027-12

Affected Industries

What Businesses Should Do

Watch preleasing, not vacancy. If you are a contractor, price data-centre work on the tenant's credit rather than the developer's, because the corpus is explicit that this is concentrated corporate credit with a building attached.

Assumptions

Text
Preleasing share is a leading indicator of delivered demand rather than a lagging indicator of lender requirements
Confidence
medium
Load Bearing
true
Basis
T-18-02; the dossier names preleasing as the early indicator specifically because vacancy lags in a preleased market
If Wrong
The indicator fires without the downside occurring - a false positive that would have to be published and scored

Precedence Note

Requires the preleasing leg. A vacancy rise alone, with preleasing holding, stays in base because it reflects delivery timing rather than demand.

Would Change Our Mind

Preleasing share rising above 85% while hyperscaler capex guidance is cut - which would mean the ratio is a financing requirement, not a demand signal.

The lease stops being a real-estate instrument

disruption · 10% · band 1-26%

CBRE already reports take-or-pay power floors requiring 60-85% minimum utilisation in Hillsboro, and utility queue-position financing emerging in Dallas-Fort Worth. If those terms generalise, the conventional commercial lease inverts: the tenant bears consumption risk on contracted electrical capacity rather than the landlord bearing vacancy risk on space. REIT structures become the wrong wrapper, infrastructure funds and private credit displace real-estate capital, and appraisal-based indices stop describing the asset class at all.

Mechanism

The substitute is an infrastructure-credit underwriting model and the capability delta is that it prices the thing that is actually scarce - dated, transferable electrical capacity - rather than the thing that is abundant, which is land and shell. At least two independent parties are pursuing it: the financings the corpus records at Related Digital and Blackstone ($16bn), Meridian Arc ($5.7bn at a 6.25% yield) and Blackstone's QTS ($2.1bn CMBS) are already credit instruments rather than property equity. The incumbent asset that becomes worth less is the appraisal-based index and the REIT wrapper built on it.

Preconditions, early indicators and assumptions

Band Width Pp

Calibration Basis

judgement_only

Thesis

Minimum-utilisation covenants, behind-the-meter generation and financeable queue positions generalise until data centres are underwritten as infrastructure credit rather than property.

Preconditions

Take-or-pay power covenants appearing in at least two markets beyond Hillsboro
At least two large platforms financing outside a REIT structure
A rating agency publishing a data-centre-specific methodology

Early Indicators

Indicator
A major data-centre platform financing or listing structured explicitly as infrastructure rather than as a REIT
Source
S-18-21 Data Center Frontier; S-18-11 Trepp securitisation data
Would Be Visible By
2028-12
Indicator
A rating agency publishing a distinct data-centre methodology that abandons property comparables
Source
S-18-21 Data Center Frontier; S-18-13 NCREIF for the index-side consequence
Would Be Visible By
2028-12

Affected Industries

What Businesses Should Do

If you underwrite data centres, build the model on contracted power and tenant credit rather than on cap rates, and assume the appraisal-based index you currently benchmark against stops being informative.

Assumptions

Text
Minimum-utilisation covenants generalise beyond the single market where CBRE has observed them
Confidence
low
Load Bearing
true
Basis
T-18-15, single-source on CBRE's Hillsboro observation, scored as the lowest-evidence trend in the sector at composite 36.2
If Wrong
This branch stays a local contracting practice and its mass returns to base

Precedence Note

Requires a second market and a second financier. One market with take-or-pay terms is a local utility negotiation, not a sector change.

Would Change Our Mind

Two further large data-centre financings closing in conventional REIT form with no utilisation covenants.

Land use becomes the binding national constraint

regulatory · 13% · band 3-23%

The 2026 instruments are already specific: New York's July 2026 one-year pause on state environmental permitting for projects requiring 50 MW or more; Hillsboro's temporary land-use moratorium; Oregon HB 4084 restricting enterprise-zone property-tax exemptions; Arizona's three-year suspension of new data-centre sales-tax exemptions; New Jersey weighing further restrictions; Phoenix municipal restrictions narrowing viable sites. CBRE now rates local opposition as an obstacle on par with power procurement. If tax-incentive withdrawal becomes the norm, the frontier-market strategy that put 77% of under-construction capacity outside established clusters loses its economic basis.

Mechanism

The instruments are state statutes and municipal land-use ordinances, issued by state legislatures and county commissions, currently at the enacted and in-force stage in at least five jurisdictions. The direction is restricting. The decisive feature is speed: a county commission vote operates on a far faster timetable than a power interconnection study, and no federal preemption exists for data-centre siting, so there is no route to override it.

Preconditions, early indicators and assumptions

Band Width Pp

Calibration Basis

base_rate

Thesis

Municipal and state moratoria spread from the handful of 2026 jurisdictions to a majority of high-growth markets, and capacity growth slows for reasons unrelated to demand or power.

Preconditions

Continued residential electricity price increases attributable to large loads
No federal preemption statute for data-centre siting
At least five further states or major counties acting

Early Indicators

Indicator
The count of US states with active data-centre permitting restrictions or incentive suspensions exceeding ten
Source
S-18-20 Construction Dive; S-18-21 Data Center Frontier; S-18-09 CBRE market commentary
Would Be Visible By
2028-06
Indicator
A federal data-centre siting preemption bill receiving a committee vote
Source
S-18-20 Construction Dive
Would Be Visible By
2028-12

Affected Industries

What Businesses Should Do

Site selection should now price political risk explicitly - entitlement timelines and incentive durability - alongside power and fibre, and should assume tax abatements are revocable within the asset's payback period.

Assumptions

Text
Local opposition is a durable political movement rather than a transient reaction that fades once specific projects are decided
Confidence
low
Load Bearing
true
Basis
T-18-09, single-source through one brokerage's market commentary; the corpus lists this explicitly under things nobody can know
If Wrong
Mass returns to base and frontier-market siting continues

Precedence Note

Regulatory outranks base when the state count passes ten, even if capacity and preleasing still look base-like, because the constraint has then changed from power to permission.

Would Change Our Mind

Two of the 2026 moratoria lapsing without renewal and the affected projects proceeding.

The information problem produces a credit event

failure · 5% · band 1-26%

Three separate 2026 measures of commercial property value point different ways and none is a clean transacted price: NCREIF's appraised capital value fell almost 20% between Q2 2022 and Q4 2024 without a recession; Green Street's REIT-implied index is up 5.0% year over year; and transacted evidence in data centres is $1.7bn for a half year against tens of billions of debt. If a forced sale establishes a genuine arm's-length price materially below carrying value, the repricing runs through data-centre ABS, the CMBS stack and bank CRE books at once, and the FOMC's leveraged-infrastructure-financing concern is validated.

Mechanism

The thesis that fails is that an asset class can be financed at scale without transaction evidence. Who is left holding the asset: lenders, ABS holders and the regional banks that hold the 80-85% of US CRE mortgage debt that is not securitised and is far less transparent than CMBS. Historical analogue: the 2007-09 episode, in which transaction volume collapsed before appraised values moved and the appraisal-based indices were the last to reflect it - a four-to-eight-quarter lag the corpus describes as the structural behaviour of an appraisal-based index, not an occasional failure.

Preconditions, early indicators and assumptions

Band Width Pp

Calibration Basis

judgement_only

Thesis

The gap between appraised, REIT-implied and absent transacted values persists until a forced sale establishes a clearing price far below carrying values, and the repricing cascades.

Preconditions

A forced or distressed sale of a large data-centre portfolio
Continued absence of arm's-length transaction evidence in the interim
Bank CRE concentration remaining undisclosed at loan level

Early Indicators

Indicator
The first large, genuinely arm's-length data-centre portfolio sale - whatever price it clears at
Source
S-18-09 CBRE investment sales; S-18-12 Green Street; S-18-13 NCREIF
Would Be Visible By
2029-12
Indicator
Green Street CPPI and NCREIF appraised values diverging by more than 15% in the same quarter
Source
S-18-12 Green Street; S-18-13 NCREIF
Would Be Visible By
2028-06

Affected Industries

What Businesses Should Do

Mark data-centre exposure to the debt market rather than to the appraisal, and treat the absence of transaction evidence as a risk premium rather than as an absence of risk.

Assumptions

Text
No scheduled event generates an arm's-length clearing price before the leverage matures
Confidence
medium
Load Bearing
true
Basis
The dossier states directly that there is currently no scheduled event that would generate one, which is precisely the problem
If Wrong
A voluntary large portfolio sale produces price discovery in an orderly market and this branch's mass returns to base and downside

Precedence Note

Failure outranks all branches on occurrence. Distinguished from downside because the trigger is a valuation event, not a demand event - demand can be intact and this branch still fires.

Would Change Our Mind

Two or more arm's-length data-centre portfolio sales clearing within 10% of carrying values.

Additional scenario notes

Precedence Rule

Assign an outcome to the branch whose distinguishing indicator fires first. Preleasing share, not vacancy, is the first mover in every branch - vacancy is a lagging indicator in a preleased market. Regulatory outranks base when the count of US states with active data-centre permitting restrictions or incentive suspensions exceeds ten. Failure outranks all branches on a genuinely arm's-length data-centre portfolio sale that establishes a clearing price materially below carrying values.

Probabilities Sum

What Must Be True To Grow

Contracted hyperscaler demand converts into delivered load rather than withdrawing, which the corpus cannot verify: sector 05 records Exelon expecting only 22% of its own 65 GW pipeline to materialise
Construction labour remains available at all - at 3.1% unemployment the sector cannot build more in aggregate regardless of capital, so additional data-centre construction is substitution
Local jurisdictions continue to permit, with no federal preemption available if they stop
Debt markets continue to fund an asset class with almost no arm's-length equity transaction evidence

What Could Stop It

Demand deflation - hyperscaler capex guidance cuts feeding through preleasing to backlog to vacancy
Physical constraint - construction labour at a record-low 3.1% unemployment with 325,000 open positions
Political licence - municipal moratoria and incentive withdrawal, which CBRE already rates on par with power procurement
Measurement revision - the CBRE/JLL ninefold gap on capacity under construction, and the three conflicting property-value measures, either of which could reprice the sector without anything changing physically

Uncertain Assumptions

That CBRE's 7,481 MW and JLL's 66+ GW measure comparable objects - they do not, the gap is roughly ninefold, neither firm publishes its inclusion rule, and every quantity in this set is CBRE's
That Census 'Office' construction remains the category in which data-centre construction is reported - the corpus confirms it is, which is why Census office construction rose 21.3% in an office-distress narrative, and any reclassification would break the series used in the base and regulatory branches
That the immigration shock's labour-supply effect is real - the corpus says the demand side is well modelled by Harvard JCHS while the labour-supply side is inferred from record-low construction unemployment and is not evidenced, and calls it the most important single gap in the dossier

Review Required

false

Authored

2026-09-15

Subject Trend Evidence Quality

Subject Trend Composite Score

90.2

Subject Trend Continuity Prior

0.949