Section 02 · The Model

One design, engineered once, deployed in repeatable blocks.

BlackGrid delivers GPU-ready infrastructure: a proprietary immersion-cooled facility design in 2 MW and 5 MW engineered blocks, on a vetted site or on yours. The design is in MEP engineering review, and we describe it the only way worth doing: outcomes, with the numbers attached.

"GPU-ready," precisely
Everything up to your hardware is complete: site, structure, power, cooling, controls, and commissioning. You bring the servers. We never supply, touch, or manage your compute. If you are shopping for GPUs or leased compute, we are not your vendor.
Why immersion

The immersion advantage, as outcomes

Three reasons, in the order they matter. First: modern AI hardware is outgrowing air. Air cooling tops out around 40 kW per rack while AI workloads push far beyond it; this design carries 100 kW per rack with no throttling. Second: the economics. Third: the calendar.

Rack Density
100 kW
Per rack, chiller-free, GPUs at full clock
Design PUE
~1.07
Versus the 1.58 industry average for air-cooled facilities
Energy Savings
$2.2M/yr
Per 5 MW block vs air-cooled, at $0.10/kWh ($894K/yr per 2 MW)
Refrigerants
0
No chillers, no cooling towers; adiabatic heat rejection
Honest read SAME DISCIPLINE AS OUR DESIGN DOCS
N1.At 2 MW, immersion costs about $2.7M more upfront than an air-cooled equivalent, recovered in roughly 3.0 years of energy savings. At 5 MW, installed cost is roughly at parity on day one, and the energy savings are all advantage.
N2.Water use is climate-dependent: well under a million gallons a year for a designed block in a cool climate, several million in a hot desert running wet mode. We state it per site, never as a slogan.
N3.The component ecosystem behind this design (immersion-born hardware, adiabatic heat rejection) is globally deployed with thousands of installations and millions of run-hours. The integrated BlackGrid design itself is in MEP engineering review, and we will say "certified" only when it is.
Modular tiers

From 2 MW edge to hyperscale, and all the way through

Two engineered designs exist: the 2 MW edge block and the 5 MW block. Everything larger is phased repetition of those blocks, which is the point: each phase is a proven quantity, each phase stands on its own economics, and no phase waits on a bespoke design.

TierBasisIndicative installed CapExDeployment windowNotes
2 MW edge Engineered design $28M 6 to 9 months (design target) Under one acre; fits industrial parcels and existing buildings
5 MW block Engineered design $61.4M 6 to 9 months (design target) The core building block with dual-loop redundancy
10 MW Phased repetition $122.8M First block 6 to 9 months; campus phased Two 5 MW blocks, phased
25 MW Phased repetition $307M First block 6 to 9 months; campus phased Five 5 MW blocks, phased
50 MW+ Phased repetition $614M First block 6 to 9 months; campus phased Campus scale by repetition of the designed block

Directional estimates from planning-grade assumptions (Boston basis). Engineered pro formas are prepared per site on request. Installed totals, not equipment-only figures.

Power architecture

A power plant that happens to serve a data center

The design default is an islanded microgrid: grid-forming battery storage with black-start capability, dual-fuel generation, and roughly 14 days of on-site fuel autonomy. The utility connection is normally open, there as insurance. That inverts the industry's biggest bottleneck: time-to-power stops waiting on the utility interconnection queue. Grid-primary operation is available as a variant where the site's power market makes it the better answer.

Design Default
Islanded
Microgrid-first; utility as backup, normally open
Storage
Grid-forming
Battery system with black start and sub-cycle ride-through
Generation
Dual-fuel
Gas primary with liquid-fuel backup
Fuel Autonomy
~14 days
On-site fuel independent of grid conditions
Deal structures

Buy it or finance it

Two leases, in plain language: a space lease for the ground under you, and a build-out you either own outright or finance through BlackGrid. No clever structures; your CFO should be able to read the whole deal in one sitting.

Path 01 · Purchase

Buy the build-out

Own your facility from day one. Direct clients commit 25% down with progress payments through energization. Best when the capacity is core to your business and your capital position supports it.

Path 02 · Finance

Finance through BlackGrid

Keep your capital in hardware. BlackGrid finances the build-out over a roughly ten-year term: one predictable payment alongside the space lease. Best when GPUs, not buildings, are where your money earns.

Deployment sequence

Designed for 6 to 9 months to power-on

The design is finished before the site is chosen, vendors are pre-sourced, and equipment is selected for availability. That is what compresses the calendar: the sequence below runs as a pipeline, not a relay race. Stated as a design target until the first build proves it; we will publish the record when it exists.

01
Design done
The engineered block already exists; no per-project design cycle.
02
Site vetted
Power, permitting, water, and fiber screened before commitment.
03
Procure
Pre-sourced vendors and availability-selected equipment go on order.
04
Build
Modular construction executed by a team with decades of MEP delivery.
05
Energize
Commissioning, then your hardware moves in.