INSIGHTDeploy stage DP-01 of 06

Site and power readiness, stage one

Site and power readiness decides whether a location can accept, energize, connect, and legally operate a factory-built compute unit — before an order is placed. The output is a decision, not a design: proceed, remediate, or disqualify.

PUBLISHED LAST VERIFIED BY JOSEF ELIMELECHREVIEWED PODOS AI ENGINEERING

8
Readiness categories assessed
3
Outcomes — proceed, remediate, disqualify
DP-01
First of six deployment stages

What stage one establishes

SR-01…08

Eight questions, asked once

It answers eight questions: power, electrical boundary, pad, access, network, permitting, environment, and security.

POWER

Power owns the calendar

A modular unit removes construction from the critical path. It does not manufacture electrons.

ACCESS

Delivery is road law

A factory-built unit arrives as cargo, which makes road law part of the site assessment.

PERMIT

The jurisdiction stays put

Factory construction moves the work, not the jurisdiction.

The chain

Where this sits in the six-stage sequence

The stages overlap, but they do not reorder. Everything downstream assumes a site that has already passed stage one.

DP-01 · This stage

Site & power readiness

Confirm power, permits, ground, access, network. You are reading this stage.

DP-02

Configuration engineering

Fix the build specification from a bounded menu. Stage guide

DP-03

Factory build & testing

Assembly, integration, burn-in on the line. Stage guide

DP-04

Transport & placement

Ship as heavy freight, rig, set, connect. Stage guide

DP-05

Commissioning

Energize, verify, load-test on site power. Stage guide

DP-06

Operations & maintenance

Monitor, maintain, grow unit by unit. Stage guide

Stage one, first principle

Power owns the calendar

Once the building is no longer the long pole, the interconnection is — and that constraint is structural, not cyclical. The IEA reports that data-centre electricity use surged in 2025 while grid-connection bottlenecks tightened, driving a scramble for alternatives.[1] Lawrence Berkeley National Laboratory estimates US data centers consumed 4.4% of US electricity in 2023, on a path to 6.7–12% by 2028.[2] Every site assessed today competes against that curve.

So the first question is not how much power the workload needs, but what power exists here and on what date. Three answers qualify: an existing service with headroom, an executed interconnection agreement with a milestone schedule, or a sized and permitted on-site generation plan. Density then decides how much of that capacity a single placement consumes — Uptime Institute's 2025 survey of more than 800 operators shows fleet rack densities rising into the 10–30 kW band, above what most existing electrical rooms were designed against.[3]

The assessment

The eight readiness categories

Each category is assessed independently and carries its own disqualifier. Separating them matters: a site can fail one and still be worth remediating, while a site that fails SR-01 or SR-04 rarely is.

CodeCategoryWhat is establishedDisqualifier
SR-01PowerExisting service capacity, an executed interconnection agreement, or a sized on-site generation plan — with a date attached.No capacity and no dated path to it.
SR-02Electrical boundaryService voltage, metering position, protection coordination, and who owns equipment on each side of the handoff.A handoff nobody will own or energize.
SR-03Pad and structureA level, drained, load-rated surface with a defined anchor pattern; geotechnical data where soils are unknown.Ground where no engineered foundation is permissible.
SR-04Access and riggingRoute survey against federal road limits, turning radii, overhead clearance, and a rigging position at the placement point.No legal route, or nowhere to set the unit down.
SR-05NetworkFiber or wireless backhaul with a defined entrance path and pathway diversity.No carrier presence and no funded build.
SR-06PermittingHow the jurisdiction treats a placed, factory-built unit: in-plant versus on-site inspection, anchorage, electrical and energy-storage approvals.No recognized pathway for off-site construction.
SR-07EnvironmentDesign conditions — cooling dry bulb, coincident wet bulb, extreme dew point — plus altitude, corrosion, seismic and flood exposure.Conditions outside the unit's rated envelope.
SR-08Security and operationsAccess control, monitoring, visitor handling, and who is on site to escort a technician.No way to control or record physical access.

Categories in detail

Ground, road, fiber, and the authority having jurisdiction

For the item-level survey — the questions to ask and how to score the answers — use the data center readiness checklist. This page covers what those answers mean for the deployment sequence.

Access, pad, and the physics of delivery

Federal size and weight standards on the Interstate System cap a single axle at 20,000 lb, a tandem axle at 34,000 lb, and gross vehicle weight at 80,000 lb, while width on the National Network is fixed at 102 inches — no state may set a higher or lower figure.[4] Height is left to the states. Units built to ISO series-1 container geometry inherit standard external dimensions, ratings, and corner-fitting positions, which is what lets a route be planned and a lift be rigged from published data rather than a field measurement.[10]

The route survey runs from the highway to the exact placement point: turning radii at the last intersections, overhead clearance including utility spans, bridge and culvert ratings on the final approach, and a rigging position with the reach and ground bearing to set the unit down. The pad is a civil deliverable — level, drained, load-rated, with anchor points and conduit stub-ups matching the unit's published pattern. Where soils are unknown, the geotechnical work belongs here, not in the week the truck arrives; the move itself is covered in transport and placement.

Network path and entrance

Network readiness is two findings: whether a carrier can serve the site at all, and whether the physical entrance path exists. ANSI/TIA-942 defines the structure behind the second — entrance rooms, pathways, cross-connects, and the distribution areas that connect them — alongside the redundancy topologies that decide whether one backhoe can take the site offline.[8] Diverse entry means two paths with real physical separation, not two strands in one conduit. How much bandwidth the workload needs is a configuration question; the networking and fiber guide covers the fabric side. Stage one only establishes that a path exists, who owns it, and what it costs to light.

Permitting a unit that was built somewhere else

The ICC/MBI off-site construction standards describe the split: in-plant inspection of what was fabricated off site, third-party inspection agencies, state modular programs, and on-site final inspection of foundation, anchorage, and connections by the local authority having jurisdiction.[5] Reading that pathway early is what keeps a finished unit from waiting on a permit.

Two code families apply regardless of where the unit was built: electrical work at the site boundary and inside the enclosure follows the National Electrical Code[6], and batteries bring stationary energy-storage requirements governing siting, separation, and fire protection.[7] Physical security has its own baseline — the NIST SP 800-53 physical and environmental protection controls cover access authorization, monitoring, visitor records, and emergency shutoff, which a remote or shared site still has to satisfy.[11] The safety and security guide covers how those controls land in a modular enclosure.

Site conditions the enclosure has to survive

Heat rejection is designed against the site, not a national average. ASHRAE publishes per-station climatic design conditions — cooling design dry bulb, coincident wet bulb, extreme annual dew point — and those are the numbers a dry cooler is sized to, with the as-of edition recorded alongside them.[9] Altitude derates air-side equipment; coastal salt and industrial dust drive material choices; seismic and flood exposure drive anchorage and pad elevation. None of these disqualify a site alone, but all change the configuration — which is why they belong here rather than in commissioning. The thermal enclosure guide covers what the envelope is asked to do once those conditions are known.

Early exits

What disqualifies a site early

Disqualifying early is the cheapest work in the deployment. Each condition below ends the assessment, because no downstream stage recovers it.

01

No power and no dated path to it

A queue position without a milestone schedule is a forecast, not capacity.

02

No legal delivery route

If a permitted vehicle cannot reach the placement point, the unit cannot arrive at any price.

03

No rigging position

A reachable site with nowhere to stand a crane is the same failure one step later.

04

No off-site construction pathway

If the authority will not recognize in-plant inspection, the factory advantage is spent on rework.

05

Unbuildable ground

Fill, high water table, or contaminated soils where no engineered foundation is permissible.

06

No network path and no funded build to one

Backhaul that depends on an unfunded carrier extension is an assumption.

07

Design conditions outside the rated envelope

Those sites need a different product, not a harder deployment.

08

No one to control physical access

An unattended site with no access control cannot meet a basic security baseline.

A queue position with no date is a forecast, and should be recorded as one.

Stage one records what exists, not what is hoped for

8

Categories, each with its own disqualifier

Honest limits

When this stage is not the right fit

The assessment assumes a specific problem: placing standardized compute units on land you control or can lease. It is the wrong instrument in several common cases.

  • You are buying capacity, not siting it. Colocation and cloud move every category here onto someone else's balance sheet; the comparison is commercial, not civil.
  • The requirement is one or two racks. Below the point where a dedicated unit is the sensible increment, an existing electrical room usually wins.
  • The workload has no location constraint. If no data, latency, or sovereignty requirement ties compute here, readiness answers a question nobody asked.
  • The site is already a commissioned data center with spare capacity. The real study is then a density and cooling retrofit, not a placement.
  • The schedule is set by something else. If funding or hiring gates the project by a year, an early readiness pass goes stale before it is used.
  • And what stage one cannot do: it does not shorten an interconnection queue, bind a jurisdiction to a verbal interpretation, or replace a stamped civil design.

In the product

How PODOS runs stage one

Because the hardware is standardized, so is the assessment. Each PODOS Pod is designed as a standardized 1-MW building block and designed for 128 GPUs, so the electrical, structural, and access requirements are known before a site is visited. Stage one becomes a comparison between fixed requirements and a specific piece of ground, rather than a design study whose answers move as the building does.

That is also what makes the schedule conditional. PODOS targets a 90-day window from order to commissioning for a standard unit — a window measured from a site that has already passed this stage, with readiness work running in parallel with the factory build. The full sequence is in the six-stage deployment overview, the electrical side in the power architecture guide, and the reason the work splits this way in factory-built versus site-built. Terms are defined in the AI infrastructure glossary.

QUESTIONS

Frequently asked questions

What is site and power readiness?

It is the first stage of a modular data center deployment: an assessment of whether a location can accept, energize, connect, and legally operate a factory-built compute unit. Its output is a decision — proceed, remediate, or disqualify — not a design.

What disqualifies a site fastest?

Power with no credible path. With no existing megawatt-class service, no executed interconnection agreement, and no viable on-site generation plan, work on the other categories cannot change the schedule. The next fastest is access: a placement point heavy road freight cannot reach.

Does a modular unit still need permits?

Yes. Off-site construction moves fabrication into a factory; it does not remove the local authority having jurisdiction. ICC/MBI 1205 describes the split: in-plant inspection of what was built off site, plus on-site final inspection of foundation, anchorage, and connections.

How early should this assessment happen?

Before an order is placed. Every later stage assumes a site that can be energized on a known date, and finding an interconnection queue afterwards converts a schedule problem into a storage problem. The assessment is cheap and reversible; the order is neither.

Bring us a site and a date

Stage one is cheap and reversible. Send the service capacity, the access route, and the jurisdiction, and engineering will tell you which of the eight categories needs work.

Size your deploymentDeployment overview