What AACE estimate class is this CAPEX model?

The engine produces a Class 4 (Study / Feasibility) estimate per AACE International Recommended Practice 18R-97. Typical accuracy range is −30% / +50%. It is NOT a Class 2 or Class 1 detailed estimate and must not be used for contract procurement without further engineering definition.

How does the city $/W override work?

When a city is selected, the engine sets locMult = city.perW / 4.65 (the cityAnchorPerW). The 4.65 $/W anchor is the Americas baseline; Tokyo at 15.20 $/W gives locMult = 3.27×. Category costs are then multiplied by this city locMult instead of the region-group multiplier.

What drives the PUE output?

PUE comes from the DATA.capexDetail.pueByCoolingRack matrix (sourced from Uptime Institute 2026 survey cohorts). It is indexed by both cooling type and rack-density tier. Air cooling with standard-density racks gives PUE 1.8; liquid cooling with AI-density racks gives PUE 1.15. PUE feeds the annual energy cost estimate.

CAPEX Calculator — Technical Manual & Methodology

01 Purpose & engineering basis

The CAPEX calculator produces a parametric, Class 4 (Study / Feasibility) budgetary estimate of data-center construction cost. It answers five pre-design questions: what is the total build cost, how does cost break down across engineering disciplines, what floor area and rack count does the design imply, what construction schedule should the project plan against, and what annual energy cost will the facility carry.

The governing cost basis is the Turner & Townsend Data Center Cost Index (DCCI) 2025, supplemented by the Cushman & Wakefield Data Center Cost Guide 2025 for city-level $/W anchors and JLL 2026 escalation indices. Category cost factors are calibrated so that the Americas baseline (locMult 1.00) at 4.65 $/W matches the sourced composite anchor. The model is AACE Class 4: project definition maturity 1–15%, typical accuracy −30% / +50%. It is an engineering pre-design screening tool, not a contractor bid.

The complete model body lives in rz-engine.js under RZEngine.models.capex.detailed() (approximately line 8399). The constant block is DATA.capexDetail (approximately line 5266). The provenance sidecar is DATA.sources['capexDetail']. The calculator page (capex-calculator.html) is a pure presentation layer that reads DOM inputs and calls the engine — no cost arithmetic lives on the page.

02 Inputs

Seventeen core inputs across five groups — IT infrastructure, cooling & density, fire/life-safety, power, and location. Eighteen additional advanced parameters unlock in Advanced mode. All defaults below reproduce the page's value="" / selected attributes exactly.

Field / DOM id	Symbol	Unit	Range / default	Meaning
`itLoad`	`itLoadKw`	kW	100 – 100 000 · default 1 000	Total IT load (IT power budget). Drives every per-kW cost line and rack count.
`rackType` (radio)	`rackType`	enum	standard* / medium / high / ai	Rack-density tier. Sets `rackKw` (6 / 12.5 / 25 / 75 kW/rack) for rack count and `rackMult` for building, electrical, and cooling cost lines.
`coolingType` (radio)	`coolingType`	enum	air* / inrow / rdhx / liquid	Cooling architecture. Drives `coolingMult`, design PUE from `pueByCoolingRack`, WUE, and construction timeline cooling multiplier.
`redundancy` (radio)	`redundancy`	enum	n / n1* / 2n / 2n1	Uptime Institute redundancy tier. Drives `redMult` applied to electrical, UPS, generator, and commissioning lines. Also scales testing and space support ratios.
`fuelHours`	`fuelHours`	hours	8 – 168 · default 48	Generator fuel-tank autonomy. Fuel multiplier = 1 + (fuelHours − 24) × 0.008. Only hours above 24 accrue additional cost.
`buildingType`	`buildingType`	enum	warehouse / modular / purpose* / highrise	Building construction type. Applies `buildingMult` (0.70 / 0.85 / 1.00 / 1.40×) to the building and seismic lines. Also scales construction schedule civil phase.
`seismicZone`	`seismicZone`	enum	zone0 / zone1 / zone2* / zone3 / zone4	Seismic design zone. Multiplies the seismic line item by `seismicMult` (0.2 / 1.0 / 2.5 / 5.0 / 8.0×).
`fireType`	`fireType`	enum	fm200 / novec* / inergen / n2 / water	Fire-suppression agent. Applies `fireSuppressionMult` (1.0 / 1.3 / 1.2 / 1.8 / 0.6×) to the fire-suppression line.
`alarmType`	`alarmType`	enum	conventional / addressable* / vesda / hybrid	Fire-detection system. Applies `fireAlarmMult` (0.6 / 1.0 / 1.8 / 2.2×) to the fire-alarm line.
`upsType`	`upsType`	enum	standalone / modular* / distributed / rotary	UPS topology. Applies `upsMult` (0.9 / 1.0 / 1.2 / 1.5×) to the UPS line item.
`genType`	`genType`	enum	diesel* / gas / dualfuel / hvo	Generator fuel type. Applies `genMult` (1.0 / 1.15 / 1.3 / 1.2×) to the generator line.
`locationFactor`	`location`	enum	americas* / emea / middle_east / apac	Region group. Resolved via `regionGroupDefaults` to an internalRegion and a base `locMult` (1.00 / 1.15 / 0.90 / 0.85×). Overridden when a city is selected.
`cityMarket`	`city`	enum	none* or any of 58 cities	Optional city-specific override. Sets `locMult = city.perW / 4.65` and overrides `effectiveRegion`. City $/W values come from T&T DCCI 2025 + C&W 2025 anchor table.
`deepSeaEnable` (checkbox)	`deepSea`	bool	off*	Enables deep-sea water cooling physics path. Adds a `deepSeaCooling` capex line from `RZEngine.models.cooling.deepSea()` and overrides PUE with the physics-derived value.
`dsDepth`	`depthM`	m	300 – 1 200 · default 900	Seawater intake depth (active only when deep-sea is enabled). Drives pipeline head pressure and pump energy in the cooling physics model.
`dsPipeline`	`pipelineKm`	km	0.5 – 15 · default 3	Subsea pipeline length. Scales pipeline capital cost in the deep-sea cooling capex model.
`refrigerantType`	`refrigerant`	enum	auto* or R-410A / R-134a / R-513A / R-32 / R-454B / R-1234ze / R-1233zd / R-717 / R-290	Chiller refrigerant selection. When not "auto", `RZEngine.models.cooling.refrigerant()` applies a `capexMult` premium to the cooling line item based on refrigerant equipment cost and safety-class installation requirements.

Advanced mode adds 18 further parameters: projection year (2025–2030 escalation), market condition, delivery method, contractor availability, design fee %, PM fee %, contingency %, FOM (substation/grid-connection/switchgear), power distribution type, transformer type, PDU type and cabling type (per-rack costs), floor type, site condition, security level, fiber entry count, green certification, and renewable generation option. Each maps to a multiplier or additive cost in DATA.capexDetail.

03 Calculation methodology

All arithmetic is in RZEngine.models.capex.detailed(inp). The model has five sequential stages: resolve multipliers → compute 14 base cost lines → add advanced per-rack and fiber costs → apply soft costs and contingency → derive PUE, space, and timeline. Function location: rz-engine.js approximately line 8399.

Stage 1 — Multiplier resolution

redMult = CD.redundancyMult[redundancy] { n:1.00, n1:1.25, 2n:1.85, 2n1:2.10 } coolMult = CD.coolingMult[coolingType] { air:1.00, inrow:1.20, rdhx:1.35, liquid:1.60 } rackMult = CD.rackMult[rackType] { standard:1.00, medium:1.10, high:1.30, ai:1.90 } buildMult = CD.buildingMult[buildingType] { warehouse:0.70, modular:0.85, purpose:1.00, highrise:1.40 } seismicMult = CD.seismicMult[seismicZone] { zone0:0.20, zone1:1.00, zone2:2.50, zone3:5.00, zone4:8.00 } fireSupMult = CD.fireSuppressionMult[fireType]{ fm200:1.00, novec:1.30, inergen:1.20, n2:1.80, water:0.60 } alarmMult = CD.fireAlarmMult[alarmType] { conventional:0.60, addressable:1.00, vesda:1.80, hybrid:2.20 } upsM = CD.upsMult[upsType] { standalone:0.90, modular:1.00, distributed:1.20, rotary:1.50 } genM = CD.genMult[genType] { diesel:1.00, gas:1.15, dualfuel:1.30, hvo:1.20 } locMult = regionGroupDefaults[location].multiplier (or city.perW / 4.65 if city selected) fuelMult = 1 + (fuelHours − 24) × 0.008 advGlobalMult = yearMult × marketMult × deliveryMult × contractorMult (1.0 in simple mode) Constants from DATA.capexDetail. locMult for city override: Dallas 14.30/4.65 = 3.075×; Tokyo 15.20/4.65 = 3.269×; Jakarta 11.21/4.65 = 2.411×.T&T DCCI 2025 · C&W 2025

Stage 2 — 14 base cost lines

costs[key] = CD.costFactors[key] × itLoadKw × multiplier × advGlobalMult Per-line multiplier breakdown (each × locMult as base): building × buildMult × rackMult × floorM × siteM seismic × seismicMult × buildMult electrical × redMult × rackMult × powerDistM × transformerM ups × redMult × rackMult × upsM generator × redMult × fuelMult × genM cooling × coolMult × rackMult fireSuppression× fireSupMult fireAlarm × alarmMult security × securityM commissioning × redMult testing × pr.testing × testingRedundancyMult[redundancy] permits × pr.permits bms, network × locMult only (no secondary multiplier) In the simple-mode loop: Object.keys(CD.costFactors).forEach(…) iterating 14 keys. pr = CD.permitRegionMult[effectiveRegion].rz-engine.js ~line 8481

Stage 3 — Advanced per-rack costs (advanced mode only)

racks = ceil( itLoadKw / CD.rackKw[rackType] ) pduTotal = racks × pduPerRack × advGlobalMult × locMult → added to costs.electrical cablingTotal = racks × cablingPerRack × advGlobalMult × locMult → added to costs.network fiberCost = CD.fiberEntryCost[fiberEntry] × advGlobalMult → added to costs.network pduPerRack: basic $800, intelligent $1 500, switched $2 500. cablingPerRack: copper $600, hybrid $1 200, fiber $2 000. fiberEntryCost: single $150 k, dual $350 k, multi $600 k.DATA.capexDetail · rz-engine.js ~line 8499

Stage 4 — Soft costs, contingency, and FOM

softCosts.design = total × (designPct / 100) (if designPct > 0) softCosts.pm = total × (pmPct / 100) (if pmPct > 0) greenPremium = total × (greenM − 1.0) (if greenM > 1.0) contingency = total × (contingencyPct / 100) simple mode: contingencyPct = 5% (CD.softCostDefaults.simpleContingencyPct) advanced mode: contingencyPct = user input (default 10%) fomTotal (if includeFOM): subCost = CD.substationCosts[substationType] × transformerLeadMult gridConnection = itLoadKw × 0.001 × CD.fom.gridConnectionPerMw ($500 k/MW) switchgear = itLoadKw × 0.001 × CD.fom.switchgearPerMw ($300 k/MW) fomTotal = (subCost + gridConnection + switchgear) × (1 + utilityRate/100) × yearMult renewableCost = CD.renewableCostPerMw[renewableOption] × (itLoadKw / 1000) × advGlobalMult total = sum(costs) + softTotal + greenPremium + contingency + fomTotal + renewableCost FOM = Facility Off-site MEP (substation, grid connection, switchgear). Substations: shared $1 M, dedicated 33 kV $4 M, dedicated 132 kV $7.5 M. Renewable: none $0, solar $1.2 M/MW, solar+BESS $2.5 M/MW.DATA.capexDetail · rz-engine.js ~line 8508

Stage 5 — PUE, energy cost, space model, and timeline

pue = CD.pueByCoolingRack[coolingType][rackType] wue = CD.wueByCooling[coolingType] { air:0.0, inrow:1.8, rdhx:0.3, liquid:0.1 } energyRate = CD.energyRateByLocation.other = $0.12/kWh (legacy quirk: location key never matches) annualEnergy = itLoadKw × pue × 8 760 × energyRate racks = ceil( itLoadKw / CD.rackKw[rackType] ) rackFootprint = CD.space.rackFootprintM2[rackType] × liquidRowFactor liquidRowFactor = 1.08 if coolingType in {liquid, rdhx, deepSea}, else 1.0 whiteSpaceM2 = ceil( racks × rackFootprint ) supportM2 = ceil( whiteSpaceM2 × CD.space.supportRatioByRedundancy[redundancy] ) adminM2 = ceil( 300 + racks × 0.15 ) grossM2 = whiteSpaceM2 + supportM2 + adminM2 Timeline (parallel-phase model — timelineDetailed()): design = max(2, ceil(base.design × lMult)) permit = max(2, ceil(base.permit × pMult)) civil = max(2, ceil(base.civil × bMult × lMult)) mep = max(2, ceil(base.mep × cMult × lMult)) commission = max(1, ceil(base.commission × cMult × max(0.7, lMult))) totalMonths = commStart + commission (phases partially overlap) lMult scales with IT load: ≤1 MW = 0.55, ≤2 MW = 0.65, ≤5 MW = 0.80, ≤10 MW = 1.00, ≤25 MW = 1.15, ≤50 MW = 1.30, >50 MW = 1.50. PUE source: Uptime Institute Global Survey 2026 cohorts.DATA.capexDetail · rz-engine.js ~line 8524

Refrigerant premium (optional): when a specific refrigerant is selected (or auto-mapped), RZEngine.models.cooling.refrigerant(refKey, opts) computes a capexMult based on GWP and safety class. If capexMult > 1, the premium = costs.cooling × (capexMult − 1) is added to costs.cooling and the total. Deep-sea cooling adds a physics-derived costs.deepSeaCooling from RZEngine.models.cooling.deepSea() and replaces the PUE lookup with the physics-derived value.

04 Constants & data sources

Every constant lives in DATA.capexDetail (rz-engine.js ~line 5266) with provenance registered in DATA.sources['capexDetail']. The 14 base cost factors in $/kW IT:

Category key	$/kW IT	Primary multipliers	Notes
`building`	$800	locMult · buildMult · rackMult · floorM · siteM	Civil, structural shell, raised floor, fit-out.
`seismic`	$100	locMult · seismicMult · buildMult	Seismic bracing, isolation pads. Zone 4 = 8× base.
`electrical`	$1 200	locMult · redMult · rackMult · powerDistM · transformerM	Switchgear, MV/LV distribution, busway, PDU (+ per-rack in adv. mode).
`ups`	$600	locMult · redMult · rackMult · upsM	UPS systems and battery strings.
`generator`	$400	locMult · redMult · fuelMult · genM	Gensets and day-tank. fuelMult > 1 above 24 h autonomy.
`cooling`	$700	locMult · coolMult · rackMult	Chillers, cooling towers, CRAC/CRAH, CDU, pipe work.
`fireSuppression`	$150	locMult · fireSupMult	Agent cylinders, distribution pipe, nozzles.
`fireAlarm`	$80	locMult · alarmMult	Detector field devices and fire control panel.
`bms`	$120	locMult	BMS / DCIM controllers, sensors, integration.
`network`	$250	locMult (+ per-rack cabling + fiber in adv. mode)	Structured cabling, patch panels, ODF.
`security`	$80	locMult · securityM	Access control, CCTV, man-trap. Security-level mult 1.0 / 1.5 / 2.2×.
`commissioning`	$120	locMult · redMult	IST, integrated systems testing, 3rd-party witness.
`testing`	$90	locMult · pr.testing · testingRedundancyMult	Factory acceptance, site acceptance, compliance testing.
`permits`	$60	locMult · pr.permits	Building permits, environmental approvals, utility interconnect.

Parameter	Values / range	Source	asOf
City $/W anchor table (58 cities)	$6.20 (Chennai) – $15.20/W (Tokyo)	Turner & Townsend DCCI 2025 + Cushman & Wakefield DC Cost Guide 2025	2025
cityAnchorPerW	$4.65/W	Americas composite baseline (locMult normalisation anchor)	2025
Year escalation multipliers	2025: 1.000 / 2026: 1.060 / 2027: 1.115 / 2028: 1.165 / 2029: 1.210 / 2030: 1.250	JLL Data Center Outlook 2026	2026
PUE matrix (air / inrow / rdhx / liquid × density tier)	1.15 – 1.80	Uptime Institute Global Survey 2026 (cooling-type cohorts)	2026
Space: rackFootprintM2	standard 2.5 / medium 2.8 / high 3.2 / ai 4.2 m²/rack	Uptime/ASHRAE TC9.9 layout guidance; DLC rows +8% CDU clearance	2026
Space: supportRatioByRedundancy	N: 0.55 / N+1: 0.75 / 2N: 1.05 / 2N+1: 1.20	Published DC space programs; white space typically 40–50% of gross	2026
fuelMultPerHourOver24	0.008 per hour	Engine lineage (capex-calculator inline model v2.3.0)	2025
energyRateByLocation.other	$0.12/kWh	Blended rate for regions other than SEA/India/USA	2025
AACE estimate class	Class 4; accuracy −30% / +50%	AACE International Recommended Practice 18R-97	2020

05 Outputs

The engine returns a single object R. The calculator page reads these fields to populate the KPI summary, cost breakdown chart, space program table, and timeline strip.

Output field	Formula / source	Unit	Interpretation
`R.total`	sum(costs) + softCosts + contingency + FOM + renewable	USD	Total project capital estimate (Class 4, −30/+50%). Displayed as headline KPI.
`R.perKw`	`R.total / itLoadKw`	$/kW IT	Build cost intensity. Compare against AACE Class 4 band: ~$5–15 k/kW for Tier 2–4.
`R.costs`	Object of 14 category costs (+ deepSeaCooling if enabled)	USD ea.	Line-item breakdown for the cost pie / waterfall chart.
`R.contingency`	`pre-contingency total × contingencyPct/100`	USD	Risk reserve. 5% simple mode, user-set in advanced mode.
`R.softCosts`	design fee + PM fee (% of hard costs)	USD obj.	Professional fees. Zero in simple mode.
`R.fomTotal`	substation + grid-connection + switchgear costs	USD	Off-site infrastructure. Zero unless FOM checkbox is enabled.
`R.racks`	`ceil(itLoadKw / CD.rackKw[rackType])`	count	Rack count driving space program and per-rack advanced costs.
`R.pue`	`CD.pueByCoolingRack[coolingType][rackType]`	ratio	Design PUE used for annual energy estimate. Replaced by physics PUE in deep-sea mode.
`R.wue`	`CD.wueByCooling[coolingType]`	L/kWh	Water usage effectiveness. 0 for air and liquid; 1.8 for in-row evaporative; 0.3 for RDHX.
`R.annualEnergy`	`itLoadKw × pue × 8 760 × energyRate`	USD/yr	Annual electricity cost at the blended energy rate. Screening-grade opex proxy.
`R.space`	racks, footprint, whiteSpaceM2, supportM2, grossM2, grossSqft, halls, kwPerM2	various	Space program output. grossM2 = white + support + admin. Screens land-take and floor loading.
`R.timeline`	design, permit, procurement, civil, mep, commission, totalMonths	months	Parallel-phase construction schedule. Phases overlap per fast-track logic.
`R.deepSea`	physics model output (null if not enabled)	object	Deep-sea cooling capex, PUE, pump power, flow rates.
`R.refrigerant`	refrigerant model output (null if auto resolves to none)	object	Refrigerant identity, GWP, capexMult, and capexPremium applied to cooling line.

06 Worked example

1 000 kW IT, standard rack (6 kW/rack), air cooling, N+1 redundancy, 48 h fuel autonomy, purpose-built building, seismic zone 2 (moderate), Novec 1230 fire suppression, addressable fire alarm, modular UPS, diesel generator, Americas region (locMult 1.00), no city override, simple mode (no advanced options). All numbers below are the literal output of RZEngine.models.capex.detailed() run against rz-engine.js via Node.js vm.runInContext — engine run verified.

Multiplier resolution: redMult (N+1) = 1.25 · coolMult (air) = 1.00 · rackMult (standard) = 1.00 · buildMult (purpose) = 1.00 · seismicMult (zone2) = 2.50 · fireSupMult (novec) = 1.30 · alarmMult (addressable) = 1.00 · upsM (modular) = 1.00 · genM (diesel) = 1.00 · locMult (americas) = 1.00 · fuelMult = 1 + (48−24) × 0.008 = 1.192 · advGlobalMult = 1.00
Category costs (= costFactor × 1000 kW × multiplier):
building: 800 × 1.00 × 1.00 × 1.00 = $800 000
seismic: 100 × 1.00 × 2.50 × 1.00 = $250 000
electrical: 1200 × 1.00 × 1.25 × 1.00 = $1 500 000
ups: 600 × 1.00 × 1.25 × 1.00 = $750 000
generator: 400 × 1.00 × 1.25 × 1.192 = $596 000
cooling: 700 × 1.00 × 1.00 = $700 000
fireSuppression: 150 × 1.00 × 1.30 = $195 000
fireAlarm: 80 × 1.00 × 1.00 = $80 000
bms: 120 × 1.00 = $120 000
network: 250 × 1.00 = $250 000
security: 80 × 1.00 = $80 000
commissioning: 120 × 1.00 × 1.25 = $150 000
testing: 90 × 1.00 × 1.00 × 1.00 = $90 000
permits: 60 × 1.00 × 1.00 = $60 000
Pre-contingency sum: $5 621 000
Contingency (5% simple mode): 5 621 000 × 0.05 = $281 050
Total CAPEX: 5 621 000 + 281 050 = $5 902 050
Cost per kW IT: 5 902 050 / 1 000 = $5 902/kW
Rack count: ceil(1000 / 6) = 167 racks
Design PUE (air, standard): pueByCoolingRack.air.standard = 1.80 · WUE = 0.0 L/kWh
Annual energy cost: 1000 × 1.80 × 8760 × $0.12 = $1 892 160/yr
Space program: whiteSpaceM2 = ceil(167 × 2.5) = 418 m² · supportM2 = ceil(418 × 0.75) = 314 m² · adminM2 = ceil(300 + 167×0.15) = 326 m² · grossM2 = 1 058 m² (11 388 ft²) · kwPerM2 = 2.4 kW/m² · 1 data hall suggested
Construction timeline (N+1, purpose-built, air, USA, 1 MW IT → lMult = 0.55): design 3 mo · permit 3 mo · procurement 2 mo · civil 6 mo · MEP 5 mo · commissioning 3 mo · total 15 months

Verification method: all numbers above are the literal return value of

RZEngine.models.capex.detailed({itLoadKw:1000, rackType:'standard', coolingType:'air', redundancy:'n1', fuelHours:48, buildingType:'purpose', seismicZone:'zone2', fireType:'novec', alarmType:'addressable', upsType:'modular', genType:'diesel', location:'americas'})

executed via node -e with vm.runInContext against the live rz-engine.js file. Not hand-computed — engine run verified.

07 References & standards

T&T DCCI 2025 — Turner & Townsend Data Center Cost Index 2025: primary source for regional build-cost benchmarks, cost-category composition, and the city $/W anchor table (58 markets).
C&W 2025 — Cushman & Wakefield Data Center Cost Guide 2025: supplementary city $/W values (Americas and APAC markets) cross-checked against T&T.
JLL 2026 — JLL Data Center Outlook 2026: year-on-year construction cost escalation indices (2025 baseline → 2030 at +25%).
AACE 18R-97 — AACE International Recommended Practice "Cost Estimate Classification System": Class 1–5 estimate definitions, project definition maturity %, and accuracy ranges. Engine is Class 4 (−30% / +50%).
Uptime Institute 2026 — Global Data Center Survey 2026: PUE distribution by cooling-architecture cohort (air, in-row, RDHX, liquid); staffing benchmarks; Tier availability definitions (T2 99.741% / T3 99.982% / T4 99.995%).
ASHRAE TC9.9 (5th ed. 2021) — Thermal Guidelines for Data Processing Environments: supply-temperature envelopes (A1–A4, H1), maximum rack density by cooling type, floor-loading guidance. Referenced in space model and cooling architecture selection.
TIA-942-C — Telecommunications Infrastructure Standard for Data Centers: Rated-1 through Rated-4 topology descriptions corresponding to Uptime Tier 1–4.
Engine source — rz-engine.js (RZEngine v2.5.1): models.capex.detailed() ~line 8399; DATA.capexDetail ~line 5266; DATA.sources['capexDetail'] ~line 6589. Minified twin: rz-engine.min.js (loaded by 52 pages including capex-calculator.html).

08 Assumptions & limitations

The model is a Class 4 parametric estimate (AACE 18R-97). Its accuracy is inherently −30% / +50% with 1–15% project definition completeness. Specific limitations and assumptions:

Cost basis: category $/kW factors are anchored to the T&T DCCI 2025 composite benchmark, calibrated so that Americas locMult 1.00 × all-defaults ≈ 4.65 $/W. They represent a market-median build; owner-specific procurement, site conditions, and design choices can shift individual lines by ±50%+.

Energy rate: the annual energy cost uses a fixed $0.12/kWh fallback for all regions except SEA ($0.08), India ($0.07), and USA ($0.10). The USA condition is a legacy key-mismatch quirk documented in the engine JSDoc — the location input never equals the string 'usa' (it maps to 'americas'), so the $0.12 fallback applies globally in simple mode. This is a documented non-defect preserved for calculation continuity; future reconciliation requires a deliberate data-model change.

Space model: rack footprint and support-to-white-space ratios are planning rules of thumb from Uptime/ASHRAE TC9.9 layout guidance. They produce a reasonable gross building footprint for programming purposes but are not a substitute for a detailed space plan or architectural layout study.

Timeline: the parallel-phase schedule model uses base durations calibrated from industry benchmarks (T&T + JLL project delivery data). Critical long-lead equipment (transformers 60–120 weeks, switchgear 50–80 weeks in the 2024–26 market) can be the dominant schedule driver and is not fully reflected in the model's phase durations — the engine's DATA.construction.longLeadWeeks table documents these but they are advisory, not input to the timeline formula.

Scope exclusions: the estimate does not include land acquisition, IT equipment, furniture and fixtures, moving/relocation, tenant improvement works above the base shell, or ongoing operating costs beyond the energy screening estimate. FOM (substation, grid connection, switchgear) is modelled only when the Advanced mode FOM checkbox is enabled.

This calculator and manual are an engineering education and pre-design screening tool. They do not constitute a professional engineering estimate, a quantity surveyor's report, or investment or procurement advice. Final capital budgets must be validated by a qualified cost engineer or quantity surveyor against site-specific project definition documents.