How is the Build CAPEX calculated?

CAPEX = IT_MW × buildCost_per_MW × tierMultiplier. The market build cost ($/MW) comes from CBRE/JLL data. Tier II = 0.85×, Tier III = 1.00×, Tier IV = 1.35×. All figures are sourced from MARKET_DATA in tco-calculator.html (inline, not the RZEngine).

What drives the breakeven year calculation?

The calculator runs a year-by-year accumulator out to 15 years. In each year it adds Build OPEX (power + staff + maintenance + insurance + network + 4% other) to the running Build cumulative total, and adds Colo OPEX (kW/month lease × growth scale + staff + network + 3% other) to the Colo cumulative. The breakeven year is the first year Build cumulative falls below Colo cumulative.

How does the Monte Carlo work?

10,000 iterations. Each draws a macro scenario (baseline 50%, AI boom 25%, recession 15%, power crisis 10%) and five stochastic variables: correlated construction/power shocks (ρ = 0.65, Box–Muller + Cholesky), occupancy uncertainty (uniform), colo price variance (uniform), and demand growth variance (uniform). The engine prefers RZEngine.models.sim.monteCarlo (seeded, reproducible, seed 20260705); falls back to inline Math.random(). Outputs P5/P50/P95 of the winning option's TCO.

Home / DC Solutions / Technical Manuals / TCO Calculator

Technical Manual · Total Cost of Ownership

TCO Calculator — Methodology & Formulas

Every input, equation, constant, market constant, and output behind the Build vs Colo vs Cloud total-cost-of-ownership calculator. All cost arithmetic is inline in tco-calculator.html (no separate engine file). Market pricing data is sourced from CBRE, JLL, Cushman & Wakefield, and DC Byte (2025). The Monte Carlo delegates to RZEngine.models.sim.monteCarlo when the engine is present; otherwise falls back to the inline Box–Muller sampler.

Engine inline JS (WACC + Monte Carlo) Markets 12 Horizons 5 yr / 10 yr MC Iterations 10 000 Scenarios 4 macro

▶ Open the live calculator

01 Purpose & engineering basis

The calculator answers one question at two horizons (5 or 10 years): for a given IT load and region, which deployment model — owning your own facility (Build), renting power-and-space in a third-party data hall (Colo), or consuming compute from a hyperscaler (Cloud) — delivers the lowest total cost of ownership?

The comparison is user-driven: you supply the IT load (MW), region, Uptime Institute tier, PUE, cooling type, utilisation rate, and annual growth rate. The calculator returns deterministic CAPEX+OPEX totals for all three paths, a breakeven year where Build crosses below Colo, an effective $/kW/yr metric, and (for Pro users) a scenario-weighted Monte Carlo risk distribution and a WACC-adjusted NPV table.

The model is a financial pre-feasibility aid — not a substitute for a full techno-economic study. Market data reflects publicly available pricing surveys; individual project costs will vary. The calculation is fully deterministic given fixed inputs; the only stochastic element is the Monte Carlo risk simulation accessed via the Pro panel.

02 Inputs

Eight inputs across two groups — site & capacity, and analysis parameters. All validation is client-side via RZCalc.validateNumbers(). Missing numerics fall back to the listed defaults.

Field	ID	Unit	Default	Meaning
Region	`tcoRegion`	enum (12 markets)	n-virginia	Selects the MARKET_DATA block supplying build cost, colo rate, power rate, staff cost, and year-indexed projections.
IT Load	`tcoItLoad`	MW	5	Rated IT power draw. Drives all linear cost scaling.
Analysis period	`tcoPeriod`	years (5 / 10)	5	Accumulation horizon. Breakeven search extends to 15 years regardless of this setting.
Tier	`tcoTier`	enum (II / III / IV)	tier-3	Uptime Institute tier. Multiplies Build CAPEX: II = 0.85×, III = 1.00×, IV = 1.35×. Also sets staffing density.
PUE	`tcoPue`	dimensionless	1.3	Power Usage Effectiveness. Scales the total facility power draw in the annual power cost formula.
Utilisation	`tcoUtil`	%	70	Average IT load factor. Scales power consumption (0–100%).
Cooling type	`tcoCooling`	enum	chilled-water	Air / Chilled-water / Evaporative / DLC. Sets the default PUE reference (inline: air 1.40, chilled-water 1.25, evaporative 1.15, DLC 1.08).
Annual growth	`tcoGrowth`	%/yr	10	Compound capacity growth rate applied to Colo and Cloud spend via `(1 + growthRate)^y` scale factor per year.

Three additional inputs are available to Pro (authenticated) users: power rate override ($/kWh), staff headcount override (FTE), colo escalation override (%), cloud reserved-instance discount (%), and debt-to-equity ratio for the WACC computation.

03 Calculation methodology

All arithmetic lives in the inline calculate() function in tco-calculator.html. There is no external TCO engine file. The inline constants are tagged INLINE below; market rates sourced from the in-page MARKET_DATA object are tagged MARKET_DATA.

3a — Build TCO: CAPEX

CAPEX = IT_MW × buildCost × tierMult × 1,000,000 [USD] buildCost = market construction cost per MW ($/MW) from MARKET_DATA. tierMult: Tier II = 0.85, Tier III = 1.00, Tier IV = 1.35. MARKET_DATA · CBRE/JLL 2025 tierMult · INLINE

CAPEX is a one-time charge, added at year 0. It also seeds the maintenance and insurance annual rates.

3b — Build TCO: annual OPEX components (Year 1)

annualPower = IT_MW × PUE × 8,760 h/yr × util × powerRate × 1,000 [USD/yr] annualStaff = staffCount × staffSalary [USD/yr] annualMaint = CAPEX × 0.03 [USD/yr] annualInsurance= CAPEX × 0.007 [USD/yr] annualNetwork = IT_MW × 20,000 × 12 [USD/yr] annualOther = (power + staff + maint + ins + net) × 0.04 [USD/yr] yr1Opex = power + staff + maint + ins + net + other powerRate = $/kWh from MARKET_DATA (Pro: overridable). staffCount = ceil(staffPerMW × IT_MW); staffPerMW: Tier II = 3, III = 5, IV = 8 (INLINE). staffSalary = $/FTE/yr from MARKET_DATA. Maintenance 3% of CAPEX, insurance 0.7% of CAPEX, network $20k/MW/month, other 4% of all other OPEX items — all INLINE constants. power · MARKET_DATA maint/ins/net/other · INLINE

3c — Build TCO: multi-year accumulation

For y = 0 .. years-1: yPowerRate = MARKET_DATA.projections.powerCostByYear[y] (fallback: powerRate × (1 + PROJECTIONS.powerInflationByYear[2025+y])^y) yStaffCost = MARKET_DATA.projections.staffCostByYear[y] (fallback: staffSalary × 1.05^y) yPower = IT_MW × PUE × 8,760 × util × yPowerRate × 1,000 yStaff = staffCount × yStaffCost yMaint = annualMaint × 1.03^y yOther = (yPower + yStaff + yMaint + yIns + yNet) × 0.04 buildTotal += CAPEX + ∑(yPower + yStaff + yMaint + yIns + yNet + yOther) Year-indexed power and staff costs come from per-market projection arrays (2025–2030, capped at the 2030 value for years beyond index). Global PROJECTIONS.powerInflationByYear fallback: 4% (2025), 5% (2026), 4% (2027), 3% (2028–2029), 2% (2030). Staff inflation fallback 5%/yr, maintenance inflation 3%/yr — INLINE. projections · MARKET_DATA inflation rates · INLINE/PROJECTIONS

3d — Colo TCO

coloSetup = IT_MW × 50,000 (one-time setup fees) [USD] INLINE For y = 0 .. years-1: scale = (1 + growthRate)^y yColoPrice = MARKET_DATA.projections.coloPriceByYear[y] (fallback: coloKW × (1 + coloEsc)^y) yColo = IT_MW × 1,000 × yColoPrice × 12 × scale [USD/yr] yColoStaff = coloStaffCount × staffCostByYear[y] [USD/yr] yColoNet = IT_MW × 15,000 × 12 [USD/yr] INLINE yColoOther = (yColo + yColoStaff + yColoNet) × 0.03 [USD/yr] INLINE coloTotal += coloSetup + ∑(yColo + yColoStaff + yColoNet + yColoOther) coloKW = $/kW/month from MARKET_DATA. coloStaffCount = ceil(1.5 × IT_MW) — reduced headcount vs own-build (INLINE). coloEsc = colo escalation rate, default 3%/yr (INLINE; Pro: overridable). Network $15k/MW/month (INLINE). Growth scale is applied to the colo lease only (workload grows, you rent more kW). Setup fees do not escalate. coloKW · MARKET_DATA setup/staffing/net/other · INLINE

3e — Cloud TCO

CLOUD_BASE = $420 /kW/month (list-price equivalent) INLINE yr1Cloud = IT_MW × 1,000 × CLOUD_BASE × 12 × (1 - cloudDiscount) cloudEgress = yr1Cloud × 0.15 INLINE cloudSupport = yr1Cloud × 0.10 INLINE For y = 0 .. years-1: scale = (1 + growthRate)^y cloudCumRate = ∏(1 + PROJECTIONS.cloudPricingByYear[2025..2025+y]) yCloud = yr1Cloud × cloudCumRate × scale yEgress = yCloud × 0.15 ySupport = yCloud × 0.10 yCloudOther = yCloud × 0.02 INLINE cloudTotal += ∑(yCloud + yEgress + ySupport + yCloudOther) cloudDiscount = reserved-instance/committed-use discount, default 30% (INLINE; Pro: overridable). PROJECTIONS.cloudPricingByYear: 0.0% (2025), −5% (2026), −8% (2027), −3% (2028), −2% (2029), −1% (2030) — cloud pricing is modelled as declining year-on-year (INLINE). Egress 15%, support 10%, other 2% — INLINE multiples on the compute line. No CAPEX for Cloud. CLOUD_BASE · INLINE cloudPricingByYear · INLINE/PROJECTIONS

3f — Breakeven year

Extend to 15 years. For y = 0 .. 14: beCum += Build_OPEX_year_y (same formula as 3c, using market projections) ceCum += Colo_OPEX_year_y (same formula as 3d) If beCum < ceCum for the first time: breakevenYear = y + 1 break The Build accumulator starts at CAPEX (year 0). The Colo accumulator starts at coloSetup. The first year y where the Build running total falls below Colo is reported as the breakeven. If no crossover by year 15, the output reads "N/A". INLINE algorithm. INLINE

3g — Effective cost per kW per year

costPerKw = winnerTCO / (IT_MW × 1,000) / years [$/kW/yr] Divides the winning option's total TCO by the installed kW capacity and the number of years. INLINE.INLINE

3h — WACC-adjusted NPV (Pro)

WACC = D/V × kd × (1 - T) + E/V × ke kd = 0.06 (cost of debt), ke = 0.12 (cost of equity), T = 0.25 (tax rate) INLINE D/V = debtEquity (default 0.60), E/V = 1 - D/V Default WACC ≈ 0.60 × 0.06 × 0.75 + 0.40 × 0.12 = 0.027 + 0.048 = 7.5% NPV_Build = CAPEX + ∑(Build_OPEX_y / (1 + WACC)^(y+1)) for y = 0..9 NPV_Colo = coloSetup + ∑(Colo_OPEX_y / (1 + WACC)^(y+1)) NPV_Cloud = ∑(Cloud_OPEX_y / (1 + WACC)^(y+1)) NPV_diff = max(NPV_Build, NPV_Colo, NPV_Cloud) - min(...) The NPV spread is displayed as the "NPV advantage" of the cheapest option over the most expensive one. The WACC uses the Modigliani–Miller tax-shield form for the debt term. All rates are INLINE (not from FINEngine — the TCO calculator does not call FINEngine). D/V is the Pro user's debt/equity input, defaulting to 60%. INLINE · Modigliani–Miller

3i — Monte Carlo risk simulation (Pro)

N = 10,000 iterations Each iteration draws: scenario ~ categorical({baseline:0.50, aiBoom:0.25, recession:0.15, powerCrisis:0.10}) conZ, pwrZ ~ N(0,1) with Cholesky correlation ρ = 0.65 (Box–Muller / engine stdNormal) occU, colU, dmdU ~ Uniform(-0.5, 0.5) Per-iteration winner TCO (e.g. Build path): CAPEX_i = CAPEX × (1 + conZ × mktVariance × scenario.costMult) [clamped 0.7–1.5] yPower_i = yPower × (1 + pwrZ × mktVariance × scenario.costMult) [clamped 0.7–1.5] yStaff_i = yStaff × (1 + occU × mktVariance × 1.5) [clamped 0.75–1.25] (Colo) colPrice_i = coloPrice × (1 + colU × mktVariance × 2.0) [clamped 0.7–1.4] (Cloud) demandScale_i = scenario.demandMult × (1 + dmdU × 0.10) Output: sorted array bestArr[0..9999]; P5 = bestArr[500], P50 = bestArr[5000], P95 = bestArr[9500] Macro scenarios multiply both demand and cost simultaneously. Construction cost and power cost shocks are positively correlated (rho = 0.65): energy-intensive supply crunches tend to hit both together. Market-specific variance (mkt.variance) scales the shock magnitude; emerging markets (Jakarta 0.22, Mumbai 0.20, São Paulo 0.25) carry wider distributions than mature markets (Frankfurt 0.09, Chicago 0.10).

Preferred path: RZEngine.models.sim.monteCarlo(fn, distributions, 10000, 20260705, {correlations:[{a:'conZ',b:'pwrZ',rho:0.65}]}) — seeded LCG (seed 20260705), reproducible across sessions. Fallback: inline Math.random() Box–Muller, non-seeded, non-reproducible. Histogram rendered in 40 bins; P5 zone green, P95 zone red. RZEngine.models.sim.monteCarlo · rz-engine.js:8990 scenario weights · INLINE/SCENARIOS

04 Constants & data sources

The 12-market dataset (MARKET_DATA) is the single largest source of externally-sourced data in the calculator. Every field is attributed below. Purely structural coefficients (tier multipliers, percentage ratios) are inline constants with no external provenance and are so labelled.

Market	Build $/MW	Colo $/kW/mo	Power $/kWh	Staff $/FTE/yr	CAGR	Variance
N. Virginia	$12.0 M	$215	$0.065	$85 K	25%	0.12
Dallas	$9.5 M	$160	$0.055	$75 K	22%	0.15
Phoenix	$9.0 M	$150	$0.058	$72 K	20%	0.13
Chicago	$10.5 M	$175	$0.072	$80 K	15%	0.10
Singapore	$14.5 M	$390	$0.180	$92 K	12%	0.14
Frankfurt	$11.6 M	$195	$0.150	$88 K	10%	0.09
Tokyo	$15.2 M	$270	$0.160	$95 K	8%	0.11
Jakarta	$8.5 M	$160	$0.080	$35 K	18%	0.22
Sydney	$12.5 M	$180	$0.140	$90 K	14%	0.11
London	$12.0 M	$200	$0.170	$92 K	10%	0.10
Mumbai	$7.5 M	$125	$0.085	$28 K	20%	0.20
São Paulo	$10.0 M	$165	$0.100	$42 K	16%	0.25

Source attribution in code: /* Source: CBRE H1 2025, JLL 2026, Cushman & Wakefield 2025, DC Byte 2025 */. Each market also carries 6-year projection arrays (2025–2030) for build cost, colo price, power cost, vacancy, and staff cost; year-indexed values take precedence over the flat-rate fallbacks.

Constant	Value	Location	Provenance
Tier II CAPEX multiplier	0.85	`TIER_MULT` inline	INLINE — structural
Tier III CAPEX multiplier	1.00 (base)	`TIER_MULT` inline	INLINE — structural
Tier IV CAPEX multiplier	1.35	`TIER_MULT` inline	INLINE — structural
Staff per MW: Tier II/III/IV	3 / 5 / 8 FTE	`TIER_STAFF_PER_MW` inline	INLINE — engineering judgement
Colo staff per MW	1.5 FTE	`coloStaffPerMW` inline	INLINE — reduced vs own-build
Maintenance rate	3% of CAPEX/yr	inline	INLINE — industry convention
Insurance rate	0.7% of CAPEX/yr	inline	INLINE — industry convention
Other OPEX overhead	4% of (power+staff+maint+ins+net)/yr (Build); 3% for Colo	inline	INLINE
Build network cost	$20,000/MW/month	inline	INLINE
Colo network cost	$15,000/MW/month	inline	INLINE
Colo setup fees	$50,000/MW	inline	INLINE
Cloud list-price base	$420/kW/month	`CLOUD_BASE`	INLINE
Cloud reserved discount	30% (Pro: overridable)	inline default	INLINE
Egress as % of compute	15%	inline	INLINE
Cloud support as % of compute	10%	inline	INLINE
Cloud other	2% of compute	inline	INLINE
Cost of debt (WACC)	6%	inline WACC constants	INLINE
Cost of equity (WACC)	12%	inline WACC constants	INLINE
Tax rate (WACC)	25%	inline WACC constants	INLINE
MC corr. construction/power	ρ = 0.65	`renderMonteCarlo()`	INLINE — Cholesky pair
MC seed	20260705	`E.models.sim.monteCarlo(...)`	INLINE
Construction inflation 2025–2030	8/10/7/6/5/4 %	`PROJECTIONS` inline	INLINE/McKinsey 2025
Cloud pricing 2025–2030	0/−5/−8/−3/−2/−1 %	`PROJECTIONS` inline	INLINE
AI demand growth 2025–2030	33/30/28/25/22/20 %	`PROJECTIONS` inline	McKinsey 2025 (attributed in code)

05 Outputs

Output	Formula ref	Unit	Interpretation
Build TCO	§3c sum	USD	CAPEX + all OPEX over the analysis horizon. Year-indexed power and staff costs.
Colo TCO	§3d sum	USD	Setup + lease + staff + network + other over horizon. Lease grows with capacity growth rate.
Cloud TCO	§3e sum	USD	Compute + egress + support + other over horizon. Compute price declines per cloudPricingByYear; spend grows with capacity.
Best option	min(Build, Colo, Cloud)	name	Lowest-TCO model at the selected horizon and growth rate.
Savings vs #2	(#2 − winner) / #2 × 100	%	Relative advantage of the winning model over the second-cheapest.
Build CAPEX	§3a	USD	One-time construction cost at Year 0.
Annual OPEX (Year 1)	§3b yr1Opex	USD/yr	Year-1 run cost for own-build; useful for budget planning.
Effective $/kW/yr	§3g	USD/kW/yr	Winner TCO amortised per installed kW over the analysis period.
Breakeven year	§3f	year (1–15)	First year Build cumulative total falls below Colo. N/A if never crosses within 15 yr.
P5 / P50 / P95 (MC)	§3i sorted array	USD	Risk distribution of the winning option's TCO across 10 000 stochastic scenarios. Pro only.
WACC-adjusted NPV spread	§3h	USD	Present-value advantage of cheapest vs most expensive option at the computed WACC. Pro only.
Breakeven (NPV projection)	§3h year loop	year (1–10)	Year Build NPV cumulative crosses below Colo NPV cumulative in the 10-year WACC projection. Pro only.

06 Worked example

Market: Dallas · IT load: 5 MW · Period: 5 years · Tier: III · PUE: 1.3 · Utilisation: 70% · Cooling: chilled-water · Growth: 10%/yr. All numbers computed by hand from the inline formulas and market constants extracted above. Year-1 values use the Year 0 projection array entries.

Build CAPEX: 5 MW × $9.5 M × 1.00 (Tier III) = $47,500,000
Build Year-1 power: 5 × 1.3 × 8,760 × 0.70 × $0.055/kWh × 1,000 = $2,185,650/yr (using Dallas power $0.055/kWh Y0)
Build Year-1 staff: ceil(5 × 5 FTE/MW) = 25 FTE × $75,000 = $1,875,000/yr
Build Year-1 maint: $47.5 M × 0.03 = $1,425,000/yr · Insurance: $47.5 M × 0.007 = $332,500/yr
Build Year-1 network: 5 × $20,000 × 12 = $1,200,000/yr
Build Year-1 other: (2,185,650 + 1,875,000 + 1,425,000 + 332,500 + 1,200,000) × 0.04 = $280,726/yr
Build Year-1 OPEX total: 2,185,650 + 1,875,000 + 1,425,000 + 332,500 + 1,200,000 + 280,726 = $7,298,876/yr
Build 5-yr TCO: CAPEX + sum of 5 annual OPEX rows (years escalate slightly). Using Year 0 rate for all 5 years as a conservative lower bound: $47,500,000 + 5 × $7,298,876 ≈ ~$84,000,000 (live calculator will be modestly higher due to year-indexed power/staff escalation from Dallas projections: power $0.055→$0.068, staff $75K→$95.7K by 2030)
Colo Year-1 lease: 5 MW × 1,000 kW/MW × $160/kW/mo × 12 × 1.0 (scale Y0) = $9,600,000/yr
Colo staff: ceil(1.5 × 5) = 8 FTE × $75,000 = $600,000/yr
Colo network: 5 × $15,000 × 12 = $900,000/yr
Colo other Y1: (9,600,000 + 600,000 + 900,000) × 0.03 = $333,000/yr
Colo setup: 5 MW × $50,000 = $250,000
Colo 5-yr TCO: Setup + 5 years (lease grows 10%/yr from year 0 scale=1 to year 4 scale=1.464). Sum of lease years: 1.0+1.1+1.21+1.331+1.464 = 6.105. $250,000 + 6.105 × $9,600,000 + 5 × ($600,000 + $900,000 + $333,000) ≈ ~$67,800,000
Cloud Year-1: 5 MW × 1,000 kW/MW × $420/kW/mo × 12 × (1 − 0.30) = $17,640,000/yr
Cloud 5-yr TCO (with 15% egress + 10% support + 2% other = 1.27× multiplier, cloud pricing cumulative rate Y0–Y4: 1.00×0.95×0.87×0.84×0.82 ≈ 0.574, growth scale cumulative sum ≈ 6.105): $17,640,000 × 0.574 × 1.27 × 5 ≈ (each year priced independently) ~$70,000,000 (cloud pricing declines partially offset growth)
Winner at 5 years, 10%/yr growth: Colo ≈ $67.8 M < Cloud ≈ $70 M < Build ≈ $84 M → Colo wins
Breakeven (Build vs Colo): With Build CAPEX of $47.5 M vs Colo $250 K setup, the Colo annual spend ($11.4 M/yr Y1, escalating) exceeds Build annual OPEX ($7.3 M/yr Y1) each year. The gap narrows only as growth compounds the colo lease. At 10% growth the colo lease grows faster than build OPEX, so the crossover is driven by growth rate vs savings per year. Rough breakeven: Build cumulative ≈ Colo cumulative → at flat rates, ΔY1 = $9.6 M − $7.3 M = $2.3 M/yr savings on colo, but CAPEX excess = $47.25 M → payback ≈ $47.25 M / $2.3 M ≈ ~20 yr (beyond the 15-yr search horizon, so reported as N/A at this scale and growth rate)

Verification note: The worked example was computed by hand directly from the inline formulas and MARKET_DATA constants extracted from tco-calculator.html. The live calculator applies year-indexed projections from the Dallas projections arrays (powerCostByYear, staffCostByYear, coloPriceByYear), so actual calculator output will differ modestly from the flat-rate approximation used here. The order of magnitude and winner (Colo) are consistent at 5 MW / 10% growth / 5 years in Dallas.

07 References & data sources

CBRE H1 2025 — Data Center Market Trends: build cost $/MW, colo $/kW/month baseline pricing across N. Virginia, Dallas, Phoenix, Chicago markets.
JLL 2026 — Data Center Outlook: construction cost global average $10.7 M/MW (2025), AI-ready premium 1.8×. Also London, Singapore, Frankfurt, Sydney market data.
Cushman & Wakefield 2025 — Data Center Outlook: vacancy rates, capacity pipelines, CAGR by market.
DC Byte 2025 — Global Data Center Tracker: operational MW, construction MW, planned MW, equilibrium colo price by market.
Uptime Institute 2025 Annual Survey — Global average PUE 1.55, target 2030: 1.30; rack density trends (8→50 kW/rack 2025→2030). Referenced in BENCHMARKS.
IEA Electricity 2026 — Global DC power demand TWh projections 2024–2030. Referenced in BENCHMARKS.dcPowerDemand.
Synergy Research Q1 2025 — Hyperscale count: 1,189 operational, 504 planned, 122,200 MW total. Referenced in BENCHMARKS.hyperscaleCount.
Dell'Oro Group 2025 — Hyperscaler CAPEX forecast: $224 B (2024), $413 B (2025), $500 B (2026). Referenced in BENCHMARKS.hyperscalerCapex.
McKinsey 2025 — AI demand growth trajectory 33%→20% (2025–2030). Referenced in PROJECTIONS.aiDemandGrowth.
STL Partners 2025 — Edge DC CAGR 16.5%. Referenced in PROJECTIONS.edgeDCGrowth.
Turner & Townsend / Arcadis — Regional construction cost benchmarks (earthquake-resistant premium Tokyo +15–20%; Sao Paulo construction inflation 20%+). Referenced in market notes.
Engine source — Inline JS in tco-calculator.html; Monte Carlo via js/rz-engine.min.js (RZEngine.models.sim.monteCarlo, seeded LCG, seed 20260705).

08 Assumptions & limitations

The model assumes a greenfield build at the stated IT load from Year 0 (no phased capacity expansion within the horizon). Power consumption is modelled as flat at IT_MW × PUE × utilisation throughout; the growth rate only scales the Colo lease and Cloud compute spend — it does not model incremental Build CAPEX for new capacity. This is conservative for Build (no expansion CAPEX) and aggressive for Colo/Cloud (full growth-rate escalation of the lease/compute line).

Colo pricing uses market-specific year-by-year arrays where available (2025–2030). Beyond 2030 the 2030 value is clamped. Mean-reversion and S-curve capacity models exist in the code (predictColoPrice, predictCapacity) but are display-only for market intelligence chips and do not feed into the TCO accumulator.

The Cloud base rate of $420/kW/month is a list-price equivalent for IaaS compute-intensive workloads; storage, managed databases, and other services are not separately modelled. Reserved-instance / committed-use discounts default to 30% (a typical 1-year commitment discount from the major hyperscalers as of 2025). The 15% egress charge is a mid-market estimate; egress costs vary substantially by provider and architecture.

WACC parameters (6% debt, 12% equity, 25% tax) are illustrative defaults for a mid-market infrastructure borrower. Real projects should compute the WACC from their actual capital structure and jurisdiction tax rate via the Pro override fields. This calculator is an engineering education and pre-feasibility aid — not a substitute for a full financial model or independent techno-economic study.

▶ Open the live TCO Calculator