Revenue starts before the campus finishes
Unit one can be energized, commissioned and let while units two through ten are still being built. Capital is not stranded waiting for a single completion date at the end of a multi-year programme.
Proposal
AI Campus Proposal
New generation AI PMDC solution: a 200 MW campus delivered as ten independent 20 MW units, engineered to the highest offshore structural class.
0MW
Capacity
20 MW × 10 units
0Pcs
Modules
2-tier loading capacity
0Racks
Density
High-density GPU cluster
Grade 9
Seismic
Highest offshore class
Reading the Figures
What Those Four Numbers Commit You To
Headline figures on a campus proposal are only useful if you can see what follows from them. Each number below constrains the others, and together they set the density the whole facility has to be engineered for.
| Figure | What it fixes | What it constrains downstream |
|---|---|---|
| 200 MW total | The grid connection the campus is negotiated against, and the size of the high-voltage receiving station on site | Utility lead time and the sectionalization of the medium-voltage loop that feeds each unit |
| 10 units at 20 MW | The increment in which capacity is built, energized and sold | Capital phasing, and the fact that no single fault or shutdown can take more than one tenth of the campus |
| 368 modules | The manufacturing volume and the shipping programme, stacked two tiers high | Factory slot booking, transport permits and the crane and foundation work each landing needs |
| 1,024 racks | The IT density the whole design is sized around | Cooling architecture, busway rating and the coolant temperatures the loop has to hold |
~195kW
Average per rack
200,000 kW across 1,024 racks. Derived from the figures above, and inside the 110 to 250 kW per rack band this platform is designed for.
2tiers
Vertical stacking
Two-tier loading roughly halves the land a given capacity occupies, which is what makes 200 MW fit a single campus boundary.
1/10
Maximum blast radius
Because each 20 MW unit has its own power, cooling and IT PODs, the largest single thing that can fail is one unit.
Density is the parameter that drives everything else. At roughly 195 kW per rack, air cooling is not an option and the cooling loop stops being a service and becomes part of the primary structure of the building.
Data Center Masterplan
AI Campus Concept, 200 MW in Ten Units
Ten 20 MW PMDC units on one campus loop road, with dedicated high-voltage access, on-site substations and ancillary buildings.
Below: the same campus as a site plan.
20 MW PMDC unitIndependent power, cooling and IT PODs. Each unit deploys and commissions without interrupting its neighbours.
HV receiving substationOwnership boundary, revenue metering and bus sectionalization feeding the campus MV loop.
MV distribution yardMain transformers and sectionalized MV buses distributing power to each 20 MW unit.
Ancillary buildingsOperations center, offices, spare-parts warehouse and staff facilities supporting 24/7 campus operations.
Delivery Strategy
Why Ten 20 MW Units and Not One 200 MW Build
A campus this size could be designed as a single facility. It is deliberately not. Independent units change the risk profile, the funding profile and the revenue profile at the same time.
One unit is the largest thing that can fail
Each 20 MW unit carries its own power, cooling and IT PODs. A fault, a shutdown or a planned maintenance window is contained inside one tenth of the campus rather than shared across all of it.
The tenth unit is built the same way as the first
A repeated 20 MW block is a manufacturing problem rather than ten separate engineering problems. Drawings, interfaces and acceptance tests are written once and reused, and each landing goes faster than the one before it.
Capacity can follow the grid, not lead it
Utility capacity rarely arrives all at once. Building in 20 MW increments lets the campus energize in step with what the connection can actually deliver, instead of holding a finished facility that cannot be powered.
The unit is the product, the campus is the arrangement. Ten units is the case shown here. The same block scales down to a single 20 MW site or up beyond 200 MW without redesigning what is inside it.
Core Design & Engineering Standards
Built to Outlast Its First Hardware Generation
Accelerator platforms turn over every few years. The shell, the structure and the coating system have to survive several of those cycles, because replacing them means taking capacity out of service. Three standards govern how long the campus stays serviceable.
Structure
Structural Durability, 25+ Year Life
Premium high-durability modular framework structures safeguard AI physical hardware while reducing long-term OPEX.
Why it mattersA 25 year shell outlives many generations of accelerator hardware. The structure is the one part of the facility that is not meant to be replaced.
Where the OPEX goesDurable framing shifts spend away from repeated structural remediation and toward the equipment that actually earns revenue.
Environment
ISO 12944 C5-M Corrosion Protection
Engineered for high humidity and marine environments, with maximum protection against salt spray and extreme atmospheric moisture.
What C5-M meansThe marine sub-class of the highest corrosivity category in ISO 12944. It is the specification written for coastal and offshore exposure, not for inland industrial sites.
How it is built upA heavy-duty system totalling 320 µm or more of dry film: 80 µm primer, 180 µm topcoat and a weathering clear coat over the top.
How it is verifiedDry film thickness is measured on the finished module, not assumed from the paint schedule.
Seismic
Grade 9 High-Standard Seismic Design
Incorporates seismic isolation bearings and shock-absorbent bracings to ensure uninterrupted runtime during seismic events.
Isolation bearingsBearings between the structure and its foundation absorb ground movement instead of transmitting it straight into the racks.
Shock-absorbent bracingBracing is detailed to dissipate energy rather than simply resist it, which is what keeps equipment mounted and piping joints intact.
The target is uptimeThe design case is not survival. It is a facility that is still running after the event, which is a materially harder requirement.