Data center equipment demand by metro: a manufacturer research method

What are you measuring?

Define the product, geography, order window and decision first. A generator factory needs qualified set orders; a chiller service team may need replacement surveys. Neither can plan from building count or metro electrical load alone.

Keep four totals: observed installed assets, documented future scope, scope expected to reach an order decision within the window, and orders actually awarded. Never add those columns together. An operating generator belongs in the installed base. It enters replacement demand only when a dated condition assessment, funded plan or owner discussion identifies replacement work. A new generator already ordered belongs in awarded scope, even if its concrete pad is still empty.

Set the purchasing unit before collecting data. A generator set is a unit; a switchgear lineup contains several sections; a cooling tower can contain several cells. Define each unit in the workbook. This method uses November 4, 2026 as its evidence cutoff. The typical case uses calendar 2027 as its order window.

Set the metro boundary and research window

Use a named metropolitan statistical area and retain its county list, geographic code and boundary vintage. The Census Bureau's July 2023 delineation workbook, available when checked for this guide, supplies those components. A sales territory can use a different boundary, but label it as your territory and keep the same definition between reporting periods. A city mailing address is not a reproducible metro boundary.

Keep utility territory as a separate field. One metro can require several utility research tracks, while your dealer's service area may stop at a different line. Assign each physical campus once to the comparison geography. Preserve county identifiers so another analyst can rebuild the assignment without interpreting a hand-drawn map.

Separate source retrieval date, document issue date, image capture date and expected order date. Reading a 2020 decision today does not make its equipment scope new. Record missing dates as unknown. Freeze a workbook version for each review, then record additions, removals and changed inputs instead of quietly rewriting the previous total.

Which records turn a campus sighting into evidence?

Build the project list from public planning cases, permit documents, utility material and operator records, then use dated exterior observations to identify questions. Preserve the original document identifier and the exact passage supporting each quantity. An article repeating an announcement can help locate a source, but it should not become an independent second project.

Dominion Energy's data center request page provides a useful utility example: it asks for a site plan, load letter and load ramp schedule, and describes service arriving in stages. Use that distinction in every territory. Requested ultimate capacity, an initial power plan and a confirmed service milestone belong in separate cells. The utility's process is a research lead, not evidence that a particular customer has completed it.

Assign the next check to each candidate. Establish the purchaser when a record names only the landowner. Request the equipment schedule when a filing gives no quantities. Leave revenue unassigned until those gaps are closed.

Two public records, two different demand questions

The following two-row evidence sample is reproducible from primary records. It is deliberately small and does not estimate either metro's total market. The Census workbook places Santa Clara County in metro 41940 and Kern County in metro 12540. The project quantities below describe the cited documents, not a field-verified installed base or a current sales opportunity.

Walsh's August 2020 decision describes 33 backup generators. Check their current status before using that historical scope. The RB Inyokern page, checked November 4, 2026, lists Under Review and proposes 40 generators at 3 MW each. Their summed ratings equal 120 MW, but the page states up to 99 MW of backup generation without reconciling that difference. Retain both source figures, flag the mismatch for docket review and use neither as IT load.

Do not put either row into an open-order forecast from these documents alone. The first research action is to establish current phase, installation and procurement status. The difference between documentary scope and present purchasing need is the point of the sample.

What can you count from above?

Count visible objects using a written classification rule, then retain an unresolved category. For an outdoor generator yard, record apparent enclosure positions and the portions hidden by screens or roofs. Do not turn an enclosure footprint into a generator rating. For cooling equipment, distinguish a visible device from a verified equipment type and record whether you counted complete units, cells or fan positions.

The Centers for Disease Control and Prevention's cooling-tower identification guidance explains why this matters. Towers can be confused with air-conditioning equipment; side-mounted fans can be difficult to see; shadows, hidden equipment and outdated views can defeat a visual search. Its guidance supports combining imagery with other records and ground investigation. It does not establish a sales need, equipment condition or health finding at a particular campus.

Review both accepted candidates and rejected or obscured locations. Checking only easy positive matches cannot reveal what the survey missed. Record the reviewed sample's selection method and error counts. If the source coverage excludes indoor equipment, say so beside the total. Do not apply a convenient accuracy percentage from a different asset class or territory.

The asset-to-demand worksheet

Use one row per campus, phase and equipment package. Keep a parent campus identifier so three buildings sharing one central plant do not become three purchases. Attach multiple evidence records to that row rather than duplicating it every time another source mentions the project. Store generator and chiller packages separately even when the purchaser is the same.

Keep observed, documented and modeled values in separate columns. A blank quantity means unknown; use zero only when evidence supports none. Assign the missing quantity or status to a reviewer before including the row in a total.

How do generators, switchgear and chillers become separate packages?

Move from campus capacity to the relevant equipment schedule. The Department of Energy's July 2024 design guide describes separate cooling and electrical systems, including power distribution and direct liquid cooling. That system map helps you ask for the right records. It does not supply a universal count of generators, switchgear sections or chillers per megawatt.

For generators, identify the backed-up load, usable unit rating, duty and redundancy arrangement. For switchgear, obtain the single-line diagram and lineup schedule. For chillers, establish the cooling architecture and scheduled plant additions or replacements. A liquid-cooling announcement alone does not quantify a chiller purchase. Record the facility heat-rejection approach before estimating that product family.

Power usage effectiveness, or PUE, is an energy ratio in DOE's design guidance. An annual PUE value is not a generator sizing study. Likewise, a campus's ultimate advertised capacity does not establish the first phase's equipment order. Keep the design engineer's scope, manufacturer's application review and actual purchasing package as the controlling inputs.

When does a project belong in the order forecast?

Enter a package in the quantified forecast when you can identify its equipment scope, purchasing route, award status and plausible order window. Keep announcements and unquantified projects on a separate watch list. Construction activity can justify another research call; it cannot supply a missing equipment quantity or tell you whether someone already bought it.

Track order, delivery, energization and commissioning dates separately. DOE's October 26, 2026 equipment-shortage discussion describes continuing transformer supply constraints and work on refurbishment and reuse. That is current context for investigating procurement timing. Obtain a dated supplier response for the exact configuration.

Set a low, base and high scenario for release within the order window. Define the factor as the chance that this identified package becomes an order in that window, conditional on the stated scope and product eligibility. An analyst-selected factor remains an input until supported by your own outcomes. Do not silently multiply it by a second schedule factor that measures the same risk.

A cancelled phase leaves the forecast with a recorded reason. A deferred phase moves to its new window. A package already awarded leaves open demand and enters the award ledger. Those movements should reconcile between reporting periods.

Worked example: one typical metro

Anvilfield research models a typical manufacturer screening 24 unique campuses within one fixed metro boundary. Fourteen have operating equipment but no identified replacement budget; four have announcements without defined packages; three have already awarded their generator packages. The remaining three each have one separate expansion package. The three expansion packages alone enter the example's open-demand calculation.

Each expansion has one 8 MW backed-up load block and identical generators with 2 MW of usable capacity per set at the specified site and duty. The case uses four duty sets plus one reserve set per block, with no sharing between campuses. That gives five sets per package and 15 sets in total. An engineering review has already resolved transient performance, derating and load sequencing for this typical configuration. Use engineering review to select equipment for the actual installation.

All 15 sets meet the manufacturer's product and channel requirements, all remain unawarded, and each complete five-set package has a 60% probability of being ordered during January through December 2027. Treat release as all-or-nothing within each package. The case uses a constant factory selling price of $400,000 per set and a 25% probability that the manufacturer wins each released package.

The price includes the generator set and factory enclosure only. Exclude freight, taxes, fuel storage, switchgear, site installation, commissioning, service contracts, financing and currency changes. Use US dollars, hold price constant through the window, and use no partial awards, cancellations after release, returns or additional discounts. Costs and margin are outside this revenue example.

How sensitive is the result to schedule slippage?

Change the release input before changing the factory plan. Hold all other example inputs constant and test 30%, 60% and 80% release within 2027. The table shows the consequence directly. Compare the three release scenarios to test sensitivity.

Fractional expected quantities are acceptable in a portfolio calculation. Actual orders remain whole five-set packages under the example inputs. The base capture value corresponds to 2.25 expected sets, but nobody can order 0.25 of a generator. Shared utility or customer risks can delay several packages together. Correlation does not change a sum of expected values when the marginal inputs are fixed, but it changes the range of possible outcomes. Do not use the expected value as a guaranteed production floor.

How should you rank metros?

Compare each product family using the same boundary vintage, order window, scope rules and scenario definitions. Report gross eligible quantity, release-weighted quantity, awarded quantity and unresolved packages separately. Show how much of the candidate inventory received document review and how much remains visually obscured or unresearched. More identified assets can mean better coverage rather than stronger demand.

Rank product families separately. A smaller metro with named buyers and dated packages may justify more sales effort; a larger installed base may justify service research. Do not combine generator sets, switchgear sections and chillers into one equipment count.

Carry the biggest open question beside the rank. Examples include an unknown purchasing route, an unresolved power milestone or an unconfirmed cooling architecture. When one unanswered question controls most of the modeled revenue, fund that verification before buying a larger inventory. Track past forecast rows through award or deferral so future timing inputs can reflect your own results.

What to document

A useful handoff contains the source extracts, the package worksheet, the inclusion and exclusion rules, and the calculations. Preserve earlier revisions. Record who changed a quantity, why it changed and which source now controls. If sources conflict, retain both claims and assign a resolution action instead of averaging incompatible numbers.

Replace notes such as large campus, strong prospect with a package record: campus A, phase 2, five modeled sets pending schedule confirmation; phase 1 excluded; purchaser unknown. Assign the check that must be completed before setting a release probability.

Confirm the legal operator and purchasing entity separately from the parcel owner. Record an authorized business contact's role and the confirmation date. Hand the sales team a package question to resolve, not a promise that the property owner is buying equipment.

Common mistakes

• Counting the campus announcement, utility filing and equipment permit as three separate projects.
• Treating a complete campus buildout as one year's equipment purchases.
• Replacing unknown equipment scope with a universal units-per-megawatt multiplier.
• Calling already awarded equipment open demand because delivery is still months away.
• Applying a modeled replacement rate to every installed unit without condition, budget or owner evidence.
• Combining factory equipment revenue with installation and service revenue while retaining the same price label.
• Reporting weighted opportunity value as confirmed orders, manufacturer market share or profit.

Field checklist

Commission a visible-asset inventory for the next review

Specify metro counties, asset classes, counting units and evidence dates in the inventory request. Define how shared yards and multiple buildings count. Request accepted, rejected and unresolved candidates linked to your existing campus IDs.

PorchRocket's Asset Search offers custom searches and monitoring for assets visible from above. The practical next step is to request the visible inventory that your worksheet lacks: generator enclosures, outdoor cooling equipment or substation changes within the selected boundary. Specify the observation dates and repeat-review interval you need, then confirm what coverage supports that scope.

Keep engineering and commercial verification in your process. An exterior count cannot establish switchgear sections inside a building, equipment brand, remaining useful life or an open purchase order. Review a small initial batch against your classification rules before expanding the territory. Use discrepancies to improve the request and keep the resulting inventory separate from the demand calculation.

References and technical cross-checks

The linked primary references were checked on November 4, 2026. Their dates serve different purposes: the Census workbook fixes geography, the 2024 DOE guide supports equipment relationships, and the utility and project pages checked in 2026 support current research questions. The Walsh decision remains historical evidence. None supplies a complete metro equipment forecast or confirms the typical prices and probabilities.

Use the related Anvilfield guides to develop the follow-up questions: generator sizing and paralleling for unit scope; power distribution and utility substations for package boundaries; cooling system types and chiller startup for plant scope; power density and construction phases for staged buildout. Carry the resulting questions into the equipment schedule and owner discussion. Update this research record when those answers change the quantity or order window.

Methodology

Counts, prices, work rates, schedules and outcomes in typical cases are inputs to the Anvilfield research model. Calculations use the stated scope, formulas, cost inclusions and exclusions. Cited public records and technical documents support their attributed facts. Release cases test sensitivity rather than statistical confidence; prices and release and win probabilities are model inputs. The historical supply-chain source is separate from a November 2026 lead-time quote for a specific equipment package.

Related tools

FAQ

People also ask

• How do you size a data center generator?
• What are the conditions to parallel two generators?
• How does power get from the utility to a server?
• How does a data center get power?
• How are data centers cooled?