Vose SoftwareModelChoice decision tree · Technology
Data Center Build
Build a Tier IV or a Tier III data centre, lease colocation space, or move to the cloud?
From the model’s decision tree
Lowest expected cost: Colocation Partnership
The expected value of each first option when every later decision is also made well (the tree rolled back).
| First decision | Expected value | |
|---|---|---|
| Build Tier IV Data Center | −$28,020,000 | |
| Build Tier III Data Center | −$16,917,750 | |
| Colocation Partnership | −$14,100,000 | ✔ best |
| Cloud-Only Infrastructure | −$54,442,500 |
| Measure | Value |
|---|---|
| Expected value of the best strategy | −$14,100,000 |
| Best outcome on the best strategy | −$12,600,000 |
| Worst outcome on the best strategy | −$15,600,000 |
S-curve (cumulative probability) of Colocation Partnership from the saved simulation.
The decision
A managed services provider needs to expand its infrastructure capacity to serve growing demand for hosting, cloud, and colocation services. The company must choose between building its own facility, partnering with a colocation provider, or going cloud-only. The options are:
- Build Tier IV data center — highest reliability (99.995% uptime)
- Build Tier III data center — good reliability at lower cost
- Colocation partnership — lease space from an existing provider
- Cloud-only — use public cloud infrastructure exclusively
How the inputs are linked
The tree sheet has a parameter table above the decision tree. Cost per rack, power costs, and customer pricing are defined once and referenced by multiple branches. Changing a single parameter instantly updates all linked values.
The inputs that drive several branches at once:
- Cost per Rack — scales build cost for both Tier IV and Tier III
- Price per Rack/Month — revenue driver in all build and colo scenarios
- Power Cost per kW — operating expense across all build options
- Number of Racks — scales all capital and operating costs
- PUE — cooling efficiency affects power costs in all build paths
The decision tree
The tree has 1 decision, 4 chance events and 11 outcomes. The best path is open: at every decision the best choice is marked, and the expected value beside each branch counts every cash flow from the start. Open any other branch to see what follows it.
DecisionChance event+$ cash flow on the branch
Build Tier IV Data Center −$22,500,000expected −$28,020,000
Chance: Tier IV Demand Scenario- 40% High Demand (90% Fill) +$5,280,000payoff −$17,220,000
- 40% Medium Demand (65% Fill) −$8,220,000payoff −$30,720,000
- 20% Low Demand (40% Fill) −$21,720,000payoff −$44,220,000
Build Tier III Data Center −$17,250,000expected −$16,917,750
Chance: Tier III Demand Scenario- 35% High Demand (90% Fill) +$10,086,000payoff −$7,164,000
- 45% Medium Demand (65% Fill) −$1,389,000payoff −$18,639,000
- 20% Low Demand (40% Fill) −$12,864,000payoff −$30,114,000
Colocation Partnershipexpected −$14,100,000 ✔ best
Chance: Colocation Outcome- 50% Scales Well −$12,600,000payoff −$12,600,000
- 50% Margin Squeeze −$15,600,000payoff −$15,600,000
Cloud-Only Infrastructureexpected −$54,442,500
Chance: Cloud Cost Scenario- 40% Costs Stable −$35,700,000payoff −$35,700,000
- 35% Costs Rise −$57,750,000payoff −$57,750,000
- 25% Vendor Lock-in −$79,800,000payoff −$79,800,000
Change the numbers. Open the workbook with ModelChoice, change a probability or a cost in the tree, and the best decision updates at once.
Key assumptions
| Assumption | Value | Notes |
|---|---|---|
| Cost per Rack (Build) | $15,000 | Infrastructure cost per rack position |
| Power Cost per kW/Month | $120 | Electricity cost per kW per month |
| Number of Racks | 500 | Total rack positions to deploy |
| Cooling Efficiency (PUE) | 1.40 | Power Usage Effectiveness (1.0 = perfect) |
| Customer Price per Rack/Month | $1,800 | Monthly revenue per leased rack |
| Land and Building Cost | $15,000,000 | Site acquisition and construction |
| Network Connectivity Cost/Month | $50,000 | Fiber, transit, and peering costs |
| SLA Penalty Rate | 10.0% | Revenue penalty per SLA violation |
| Power per Rack (kW) | 8 | Average power draw per rack |
| Colocation Price per Rack/Month | $2,200 | Third-party colo provider rate |
| Cloud Cost per Rack Equivalent/Mo | $3,500 | Cloud infrastructure equivalent cost |
| Analysis Period (Months) | 60 | 5-year analysis horizon |
The figures are illustrative: the companies and numbers are invented for teaching.
Try it in ModelChoice
- Increase Power Cost per kW. Watch build options become less attractive
- Raise Customer Price per Rack. See all build scenarios improve
- Change Number of Racks. Observe economies of scale
- Run Sensitivity Analysis. Tornado chart shows which parameters matter most
Find similar models
- Product
- Type
- Application
- Uses
Further reading
Download the workbook and open it with ModelChoice. The free trial opens every example.
Related example models
- ModelChoice★★★★★Fleet ElectrificationConvert a 100-vehicle diesel fleet to electric all at once, in phases, in part, or not at all?Two decisions and four chance events, driven by linked inputs. Best first decision: Hybrid Fleet (30% EV), expected value $1.1m.
- ModelChoice★★★★★Retail Store RolloutOpen 50 stores at once, pilot 10 first, open one flagship, or stay put?Two decisions and four chance events, driven by linked inputs. Best first decision: National Rollout (50 stores), expected value $83.0m.
- ModelChoice★★★★★Software Platform Build vs BuyBuild a software platform, buy a licence, buy and customise one, or use open source?Two decisions and four chance events, driven by linked inputs. Best first decision: Buy Commercial License, expected value $9.4m.
An illustrative example: the figures are invented for teaching. Last updated 9 October 2026.