From site search to operations: the project life cycle

A data center project begins long before anyone pours concrete, and its engineering continues for decades after the ribbon is cut. The earliest decisions, where to build and how to get power there, often matter more than any later design choice, and they depend as much on utilities, municipalities and communities as on engineering. Once the building is running, its design must support years of maintenance, upgrades and changes in the equipment it houses.

This chapter follows a project through its whole life: site selection, securing power, permits and approvals, negotiating with the municipality, design and procurement, construction, commissioning and operations. It closes the course by showing how the pieces from Lessons 1 to 4 come together.

Site selection

A site search usually starts with a long list of regions and narrows to a few candidate parcels. The main criteria are:

• Power: how many megawatts the utility can deliver, how soon, at what price and with what carbon intensity. This is now the deciding factor in many markets.
• Fiber: proximity to long-haul fiber routes and to carrier networks, ideally with diverse routes from two or more carriers, and latency to the users or other data centers it must serve.
• Land and zoning: a parcel large enough for buildings, substation, generators and setbacks; land use rules that permit a data center, or a realistic path to rezoning.
• Water: availability and cost, if evaporative cooling is planned; sewer capacity for cooling tower discharge.
• Natural hazards: flood plains, seismic hazard, tornadoes, hurricanes, wildfire, and proximity to hazards such as airports, rail lines carrying dangerous goods or chemical plants.
• Climate: cooler, drier climates allow more free cooling.
• Taxes and incentives: property tax, sales tax on equipment, special levies and any incentive programs.
• Workforce and community: construction trades and operations staff nearby, and a community that will accept a large industrial building.

Teams often compare sites with a weighted scoring matrix: each criterion receives a weight that reflects its importance, each site a score, and the weighted sum ranks the sites. The matrix organizes the discussion but does not replace judgment: a site that fails one essential criterion (no power for five years, say) should be eliminated before scoring.

Securing power

Getting power is usually the longest step. The developer submits a load request to the utility or system operator, describing the size of the load, its ramp-up over time and the desired in-service date. The utility then studies whether the grid can supply it: load flow, short-circuit and stability studies, and the transmission lines, substations and generation that would be needed. Large requests may need new transmission lines that take years to approve and build. The developer typically pays deposits for studies and contributes to the cost of the connection, and signs an agreement committing to take and pay for the power.

In the United States, the process depends on the utility and the regional market. Many utilities have created large-load tariffs that require long-term contracts, minimum payments and collateral from data centers, so that other customers are not left paying for grid upgrades if a project is canceled. In Texas, Senate Bill 6, signed in June 2025, set new rules for loads of 75 MW or more on the ERCOT grid: interconnection study fees, disclosure of duplicate requests elsewhere in the state, equipment that lets the grid operator curtail non-critical large loads during emergencies, and a voluntary program for loads that can reduce demand or switch to backup generation on request. Developers in constrained markets increasingly plan on-site generation, either to bridge the years before the grid connection or permanently.

Because power drives the schedule, many developers now buy sites that already have power commitments, or work backward from the utility's in-service date. Fiber is usually less difficult, but diverse routes into the site, entering the building at separate points, must be planned early (DCE-210 and DCE-320).

Permits and approvals

In the United States, land use is controlled by cities and counties under state law. A typical sequence is:

• Zoning: confirm whether a data center is a by-right use (approved administratively if it meets the standards) or needs a special exception, conditional use permit or rezoning, which require public hearings and a vote of the planning commission or elected board.
• Site plan approval: grading, stormwater, landscaping, screening, noise, lighting and access, reviewed by county or city staff.
• Building, electrical, mechanical and plumbing permits, with plan review and inspections by the building department and fire marshal.
• State and federal approvals as needed: air permits from the state environmental agency for diesel generators or on-site power plants, a construction stormwater permit under the Clean Water Act (NPDES), wetland permits from the US Army Corps of Engineers if wetlands are filled, water withdrawal permits, and state utility commission approval for a power plant. Federal environmental review under NEPA applies only when there is a federal action, such as federal funding or a federal permit.

Approval is not guaranteed. On March 18, 2025, Loudoun County, Virginia, the heart of the world's largest data center market, ended by-right approval of data centers: new ones now need a special exception with public hearings before the planning commission and the Board of Supervisors. Other counties have adopted moratoriums or stricter standards for noise, setbacks and generator emissions.

Negotiating with the municipality and the community

A data center brings a municipality large tax revenues and construction activity but relatively few permanent jobs, and it can bring heavy truck traffic during construction, generator noise and exhaust, transmission lines, water demand and a very large building. Approval often comes through negotiation, recorded in a development agreement. Typical subjects are:

• Off-site infrastructure: road upgrades, water and sewer extensions, electrical substations and who pays for them. Developers commonly fund these improvements as a condition of approval.
• Taxes and incentives: property tax on buildings and, in many states, on the computing equipment itself; sales tax exemptions for equipment that many states offer to attract data centers; and negotiated tax abatements or payment-in-lieu-of-taxes agreements.
• Operating limits: noise limits at the property line, generator testing hours, lighting, visual screening with berms and landscaping, and setbacks from homes.
• Community benefits: local hiring, training programs with colleges, heat reuse, donations to community facilities.

Effective negotiation rests on preparation and honesty. Engineers support it with facts: noise models of generators and cooling equipment, traffic studies, visual simulations, water balances and emissions estimates. Learning what the municipality needs (its tax base, its infrastructure plans, its residents' concerns) and offering design solutions to real problems works better than pressure. DCE-410 and DCE-470 cover the process and negotiation skills in depth.

Design and procurement

Design proceeds in stages: a basis of design that records the requirements (capacity, density, redundancy, efficiency targets, codes), schematic design, design development and construction documents. Electrical, mechanical, structural, civil, fire protection, architectural and telecommunications designs must fit together, and many owners use building information modeling (BIM) to coordinate them in 3D.

Major equipment is ordered early. Large power transformers, medium-voltage switchgear, generators and chillers often have lead times of one to three years, longer than the time to construct the building. Owners with many projects reserve manufacturing capacity in advance. To save time, much equipment now arrives prefabricated: power skids with switchgear and UPS factory-wired and tested, modular electrical rooms, and prefabricated cooling plants.

Construction and commissioning

Construction moves from site work and foundations to the building shell, then to the mechanical and electrical installation, which is most of the cost and most of the work. Quality control on site, inspections by the authorities and the owner's own reviews continue throughout.

Commissioning proves that everything works, alone and together, before IT equipment arrives. It is usually organized in levels:

• Level 1: factory witness testing of major equipment such as switchgear, UPS systems and generators.
• Level 2: inspection of equipment as delivered and installed.
• Level 3: start-up of each piece of equipment by the vendor.
• Level 4: functional performance testing of each system under all its operating modes.
• Level 5: integrated systems testing (IST), in which load banks simulate the IT heat and power load and the team fails the utility, generators, UPS modules, chillers and pumps one at a time to prove that the facility rides through as designed.

Commissioning finds problems that design reviews miss: wrong settings, wiring errors, controls sequences that behave unexpectedly. It is far cheaper to find them with load banks than with customers' servers. Lesson 4 noted that Uptime Institute recommends certifying the constructed facility, not just the design, for this reason.

Operations

Once live, the facility enters its longest phase. A small team of facility engineers and technicians runs it around the clock, supported by vendors. Operations rest on documents and discipline:

• Standard operating procedures (SOPs) for routine tasks, methods of procedure (MOPs) that script every step of maintenance on critical equipment, and emergency operating procedures (EOPs) for failures.
• Preventive maintenance on a schedule, and predictive maintenance using monitoring data such as infrared scans, oil analysis and battery tests.
• Change management: every change to critical systems is reviewed, approved and scheduled.
• Monitoring: building management and electrical power monitoring systems, and data center infrastructure management (DCIM) software that tracks capacity, power, temperature and assets.
• Capacity management: tracking how much power, cooling and space remain, and planning upgrades before they run out.

Over a life of 20 years or more, IT equipment is replaced every few years, rack density rises, and cooling and power systems are upgraded while the facility runs. Designs that leave room for growth, such as space for more cooling equipment, pipe connections for future liquid cooling, structural capacity for heavier racks and a power connection that can be expanded, last longer.

Bringing it together

A single decision touches every discipline. Choosing to build an AI hall at 120 kW per rack sets the electrical design (busway and high-current feeds), the mechanical design (direct-to-chip liquid cooling with CDUs), the structure (heavier floor loads), the fire protection (lithium-ion UPS batteries, liquid near electronics), the network (high-bandwidth fiber between racks), the site (a bigger substation, perhaps on-site generation), the permits (air permits for that generation) and the operations team's skills. The rest of this program covers each of these in depth: power in DCE-310 and DCE-340, cooling in DCE-330, cabling in DCE-210 and DCE-320, site, permits and power in DCE-410 and DCE-420, structures in DCE-430, fire and security in DCE-440, and operations in DCE-450.