Public Research Scenario Model
Economic Availability Calculator Methodology
The calculator is designed to show how a small Economic Availability gap can translate into large recoverable economic output at AI factory scale. It is not based on Synestra customer operating data and it does not claim guaranteed outcomes.
Core definition
Economic Availability measures the percentage of a facility's potential economic output that is successfully converted into productive work.
Economic Availability Gap = Target EA − Current EA
Recoverable Capacity Equivalent = Campus IT MW × EA Gap
Campus value recovery formula
Total Recoverable Economic Output = Campus IT MW × EA Gap × Value Intensity per Recovered MW-Year
The value intensity is a scenario input derived from public market proxies. It combines landlord-side capacity economics and tenant-side AI production economics into one campus-level recoverable value estimate. It should not be read as energy savings alone.
Conservative case
$20M per recovered MW-yearUsed when tenant monetization is uncertain or when only a portion of the compute value can be captured.
Base case
$38M per recovered MW-yearA 500 MW campus improving from 92% EA to 95% EA recovers 15 MW equivalent. 15 MW × $38M = $570M per year as a base-case scenario, subject to site-specific validation.
High case
$55M per recovered MW-yearUsed for high-density AI factories with high GPU-hour value, high utilization pressure and strong tenant monetization.
Value allocation
The calculator shows total recoverable value first, then allocates it between landlord and tenant.
Landlord Capture = Total Recoverable Value × Allocation %
Tenant Capture = Total Recoverable Value × (1 − Allocation %)
The default 30% landlord / 70% tenant split is a scenario assumption. The calculator allows 25/75, 30/70 and 35/65 because the split depends on lease terms, power structure, tenant economics, SLA exposure and commercial negotiation.
Research foundation
- Illustrative hyperscaler reference (synthetic): public campus scale, 1.4 GW critical IT capacity, 10 data centers, 3.7M square feet, 1,200 acres and 250 kW+ rack capability.
- McKinsey: AI data center capital intensity and the scale of global AI infrastructure investment.
- Uptime Institute: outage cost, operational risk, availability pressure and AI density challenges.
- NVIDIA: liquid-cooled rack-scale AI infrastructure and AI factory architecture.
- IEEE / ACM / Google / LBNL research: utilization gaps, energy proportionality, overprovisioning, stranded capacity and software coordination gaps.
- Public GPU cloud pricing: market proxy for tenant-side GPU-hour value.
Individual campus results will vary depending on infrastructure design, workload characteristics, operating practices, commercial arrangements, baseline utilization, and tenant monetization.
Important limitation
The calculator is a scenario model. It does not use private illustrative hyperscaler data, tenant data, Oracle data, or Synestra customer performance data. Actual results depend on facility design, workload mix, utilization, controls maturity, operating model, power architecture, cooling design, tenant behavior and measured baseline performance.