The following key metrics define the scope and scale of this 200 GW energy system project. These figures establish the baseline parameters from which all three density scenarios are derived.

Three rack-level power density scenarios have been modeled to reflect the spectrum of modern hyperscale compute architectures — from traditional air-cooled deployments to next-generation liquid-cooled AI inference and training clusters. Each scenario produces dramatically different rack counts and physical footprints at the 200 GW scale.

The physical land and building footprint required to house 200 GW of compute capacity is one of the most operationally significant variables in this analysis. The formula assumes 42 racks per 1,000 SF — a standard industry planning metric for raised-floor or slab-on-grade data hall design. At low density (35 kW/rack), the campus requires over 5.7 million square feet — equivalent to roughly 100 city blocks. At high density (150 kW/rack), that footprint compresses to 1.3 million SF.

All three scenarios share a common planning constant: 42 racks per 1,000 SF of data hall space. This ratio reflects standard industry practice for raised-floor data halls with sufficient aisle spacing for hot/cold containment, cable management, and human access. This ratio does not change with rack power density — it reflects physical space planning, not power planning. The power density per rack determines how many watts are drawn from each of those 42 rack positions, which is what drives the total campus power load and, consequently, the energy infrastructure capital cost.

The vanilla powered shell cost — which explicitly excludes energy systems and IT equipment — varies significantly across density scenarios. Higher-density configurations command a premium cost per square foot due to the structural, mechanical, and electrical infrastructure required to support concentrated thermal loads. These per-SF figures represent the base civil and structural construction cost only.

Multiplying the per-SF shell cost by the total campus square footage yields the aggregate vanilla shell construction cost for each scenario. This figure represents the base real estate and structural investment, before any energy or IT systems are layered on top. The wide range — from $15.9B to $47.6B — underscores how critically the density decision drives total capital exposure at this scale.

The following table presents the full calculation chain for each scenario — from power density input through rack count, campus footprint, per-SF shell cost, and total shell cost. All figures reflect vanilla powered shell construction costs only, explicitly excluding energy systems (grid connection, substations, generators, UPS) and IT systems (servers, networking, storage).

The energy system cost is fixed at $1,200,000 per megawatt, applied uniformly across all three scenarios. This figure encompasses grid interconnection, high-voltage transmission infrastructure, substation construction, backup generation (diesel or gas turbine), UPS systems, and power distribution units down to the rack level. Since all three scenarios deliver the same 200 GW total system power, the energy system cost component is identical regardless of density configuration.

The total project cost combines the vanilla shell construction cost with the energy system cost for each scenario. IT systems costs are explicitly excluded. This combined figure represents the full capital commitment required to deliver a ready-for-IT, fully powered data center campus at 200 GW scale.

The chart below illustrates the total project cost for each scenario, decomposed into shell and energy system components. The dominance of the shell cost in Scenario A reflects the compounding effect of a large physical footprint, while Scenario B's optimal balance of footprint and per-SF rate produces the most capital-efficient outcome across all three models.

A critical metric for comparing density scenarios is the fully-loaded capital cost per delivered watt of compute power. This normalizes total project cost against the constant 200 GW output to produce a direct efficiency comparison. Lower cost per watt indicates superior capital utilization. These figures exclude IT systems, which would add substantially to total cost-per-watt but are constant across scenarios.

Understanding the capital efficiency of each scenario requires examining both the absolute dollar commitments and the relative cost drivers. The table below summarizes the key efficiency metrics across all three scenarios, enabling direct executive comparison.

The density decision involves a fundamental tradeoff between campus footprint and per-SF construction cost. Higher-density configurations reduce land and footprint requirements but demand more expensive, specialized construction. This tradeoff is not linear — the optimal outcome emerges at mid-density where both factors are in balance.

Based on the capital cost analysis, Scenario B at 100 kW per rack emerges as the unambiguous winner from a purely financial perspective. At $17.22B total project cost (shell + energy, excluding IT), it delivers 200 GW of compute capacity at a cost-per-watt of approximately $86 — nearly 3× more capital-efficient than the low-density alternative and 2× more efficient than the high-density option.

While the financial analysis clearly favors Scenario B, executives and capital allocators must weigh several risk factors that exist outside the scope of this shell-and-energy cost model. The following considerations should inform final configuration selection and project structuring.

The three density scenarios imply radically different site selection strategies. A 136 million SF campus (Scenario A) requires a land parcel of approximately 12,500 acres — roughly the size of a medium-sized U.S. city. This scale of contiguous land availability is extremely limited and likely restricts viable site locations to rural areas in the central or southeastern United States, limiting access to fiber infrastructure and skilled labor pools.

Scenario B at 47.6 million SF requires approximately 4,400 acres — still substantial, but achievable within the land availability profiles of several major metro exurban markets. Scenario C at 31.7 million SF (approximately 2,900 acres) offers the greatest site selection flexibility and the widest range of viable locations including proximity to renewable energy zones, existing transmission infrastructure, and established data center markets.

The model's output is highly sensitive to the two primary input variables: watts per rack and shell cost per SF. The following sensitivity analysis illustrates how moderate changes in these assumptions would affect total project cost for each scenario. Decision-makers should stress-test these variables against current market data and project-specific site conditions before committing capital.

This analysis evaluates a 200 GW data center energy system across three rack-level power density configurations. The findings deliver a clear, data-driven recommendation for capital allocation and project architecture at hyperscale.