Construction, quality assurance and a site design project

A data center's civil and structural design succeeds only if it is built as designed, on schedule, and in a way that supports the fast-track, phased delivery that owners demand. Large campuses are built in 18 to 30 months, with buildings often designed while earlier phases are under construction, and with mechanical and electrical installation starting as soon as the roof is on. This final lesson covers how civil and structural work is sequenced and delivered, how quality is assured through inspection and testing, how construction safety and environmental controls are managed, how digital tools and embodied carbon accounting support the work, and finally a design project that brings together the course: the civil and structural concept for a data center campus.

Delivery methods and sequencing

Data centers are delivered under several contract models. Design-bid-build separates design and construction, giving clear documents but a slow schedule. Design-build and construction management at risk bring the contractor in early, allowing packages to be released for construction before the whole design is complete, which is the norm for hyperscale projects. Many owners use standardized designs and pre-negotiated framework agreements with contractors and suppliers so that each new site starts from a proven design, adapted to local soils, climate and codes (the adaptations covered in Lessons 1 to 11).

A typical civil and structural sequence for a single-story data hall:

• Site preparation: clearing, erosion and sediment controls (Lesson 1), stripping topsoil, mass earthworks and ground improvement (Lesson 4).
• Underground services: storm and sanitary sewers, water mains, duct banks and fuel lines (Lesson 2), ideally installed before the building pad is finished.
• Foundations: footings, piles or rafts, pads for outdoor equipment (Lesson 3).
• Slab on grade: often placed early so it can serve as a casting bed for tilt-up panels and a working surface for erection (Lesson 8).
• Structure: steel erection or tilt-up panels, roof structure and deck (Lessons 5, 6 and 8), with the lateral system completed before the building is loaded with equipment.
• Envelope: roofing and walls to make the building "dried in" (Lesson 10), the key milestone for starting electrical and mechanical work inside.
• Site works: paving, final grading, landscaping, perimeter security (Lesson 11).

The schedule is driven by long-lead items, such as steel joists and girders, precast elements, roofing materials in a busy season, and, outside the civil scope but affecting it, transformers and generators, and by weather: concrete placement and roofing are difficult in a Canadian winter, requiring heated enclosures, accelerated mixes and insulated curing blankets.

Quality assurance and quality control

Quality control (QC) is the contractor's checking of its own work; quality assurance (QA) is the owner's and designer's independent verification that the work meets the specifications. Building codes require certain inspections and tests, often by independent testing agencies retained by the owner, and data center owners usually add more. Key activities for civil and structural work:

Concrete strength acceptance

Concrete strength is verified with cylinders made from samples of delivered concrete, cured under standard conditions and crushed at 28 days (or the specified age). Codes define acceptance criteria that allow for normal variability: in both CSA A23.1 and ACI 318, strength is acceptable if the average of any three consecutive tests equals or exceeds the specified strength \(f'_c\), and no single test falls more than 3.5 MPa below \(f'_c\) (for \(f'_c\) up to 35 MPa). Because of this, suppliers design mixes for an average strength well above \(f'_c\).

import numpy as np

# Concrete strength acceptance (CSA A23.1 / ACI 318 style): each test is the average of two or three cylinders at 28 days
fc = 35.0
tests = np.array([38.2, 36.9, 40.1, 33.8, 35.2, 34.0, 37.5, 39.8, 30.9, 36.4, 41.0, 38.7])
print(f"specified strength {fc} MPa; {len(tests)} tests, mean {tests.mean():.1f} MPa, standard deviation {tests.std(ddof=1):.1f} MPa")
ok = True
for i in range(2, len(tests)):
    avg3 = tests[i - 2:i + 1].mean()
    if avg3 < fc:
        print(f"  tests {i - 1}-{i + 1}: running average {avg3:.1f} MPa is below {fc} MPa")
        ok = False
low = tests[tests < fc - 3.5]
for v in low:
    print(f"  single test {v} MPa is more than 3.5 MPa below the specified strength: investigate (cores, load tests)")
    ok = False
print("acceptance criteria met" if ok else "criteria not met: investigation required")
# Required average strength for mix design so that few tests fall below the criteria (ACI 318 style, s = 3.5 MPa)
s = 3.5
fcr = max(fc + 1.34 * s, fc + 2.33 * s - 3.5)
print(f"with a standard deviation of {s} MPa, the mix should be designed for an average of about {fcr:.1f} MPa")

specified strength 35.0 MPa; 12 tests, mean 36.9 MPa, standard deviation 3.0 MPa
  tests 4-6: running average 34.3 MPa is below 35.0 MPa
  single test 30.9 MPa is more than 3.5 MPa below the specified strength: investigate (cores, load tests)
criteria not met: investigation required
with a standard deviation of 3.5 MPa, the mix should be designed for an average of about 39.7 MPa

Although the average of all twelve tests exceeds 35 MPa, one run of three tests and one single test fail the criteria, so the concrete represented by those tests must be investigated. Investigation typically starts with checking the testing procedures (poorly made or cured cylinders often cause low results), then cores drilled from the structure, which are accepted under the codes if their average is at least 85% of \(f'_c\) and no core is below 75%, and, rarely, load testing or strengthening. A mix designed for an average of about 40 MPa, rather than 37, would have made such failures much less likely.

Construction safety

Civil and structural construction includes many high-risk activities: deep excavations (Lesson 4), crane lifts of steel, tilt-up panels and modules (Lessons 8 and 9), working at height on roofs and steel, energized utilities nearby, and heavy equipment moving around workers. Provincial occupational health and safety regulations in Canada (and OSHA rules in the US) require hazard assessments, engineered shoring and temporary bracing where specified, lift plans for critical lifts, fall protection, and utility locates before digging. Designers contribute through "prevention through design": choosing details and sequences that reduce risk, such as parapets that double as fall protection, permanent anchors for roof maintenance, connections that can be made safely from aerial lifts, and avoiding designs that require work near energized lines.

Environmental controls during construction

Construction sites must control erosion and sediment (silt fences, sediment ponds, stabilized entrances), dust, noise, and spills of fuel and concrete washout, under provincial environmental regulations and municipal bylaws in Canada and the EPA's Construction General Permit and state rules in the US. Large data center sites often require permits for work near watercourses and wetlands, and for dewatering discharges. Tree protection, wildlife timing windows (for example, avoiding clearing during bird nesting season under the federal Migratory Birds Convention Act) and archaeological monitoring may also apply.

Digital delivery: BIM and digital twins

Building information modeling (BIM) is now standard for data centers. The civil, structural, architectural, mechanical and electrical models are combined to detect clashes before construction (Lessons 2 and 6), to extract quantities, and to plan sequences (4D) and costs (5D). Laser scanning and drone surveys compare as-built conditions with the model during construction, for example confirming slab levels, embed locations and steel positions. At handover, an as-built model with asset data becomes the basis of a digital twin used by operations (DCE-450): it records the location of every buried service, embedded conduit and post-tensioning tendon, which is invaluable when the facility is modified years later.

Embodied carbon accounting

Owners increasingly report the embodied carbon of their buildings and set targets to reduce it. A whole-building life cycle assessment (to standards such as EN 15978 or ISO 21930) totals the emissions from producing materials (the "product stage", modules A1 to A3), transporting and installing them (A4 to A5), and replacing and disposing of them over the building's life. For data centers, the structure, foundations and envelope typically account for most of the upfront embodied carbon of the building shell, while the mechanical and electrical equipment, replaced several times over the building's life, adds a large share over time.

import numpy as np

# Order-of-magnitude structural quantities and embodied carbon for a 40 MW single-storey data hall building
area = 120 * 200                       # m2 footprint
items = [
    # name, quantity per m2 of footprint, unit, kg CO2e per unit
    ("slab on grade 250 mm", 0.25, "m3 concrete", 300),
    ("foundations and pads", 0.12, "m3 concrete", 300),
    ("tilt-up walls (perimeter)", 0.09, "m3 concrete", 320),
    ("reinforcing steel", 38.0, "kg", 1.2),
    ("structural steel (columns, girders, joists)", 55.0, "kg", 1.4),
    ("steel deck", 11.0, "kg", 2.5),
    ("roof insulation and membrane", 1.0, "m2", 25),
]
total = 0.0
for name, q, unit, ef in items:
    Q = q * area
    co2 = Q * ef / 1000
    total += co2
    print(f"{name:44s} {Q:10,.0f} {unit:12s} {co2:7,.0f} t CO2e")
print(f"total about {total:,.0f} t CO2e, or {total * 1000 / area:.0f} kg CO2e per m2 of footprint")
print(f"per MW of IT capacity: {total / 40:.0f} t CO2e; equal to the operational emissions of 40 MW for about {total / (40 * 8760 * 0.03):.1f} years on a 30 g/kWh grid")

slab on grade 250 mm                              6,000 m3 concrete    1,800 t CO2e
foundations and pads                              2,880 m3 concrete      864 t CO2e
tilt-up walls (perimeter)                         2,160 m3 concrete      691 t CO2e
reinforcing steel                               912,000 kg             1,094 t CO2e
structural steel (columns, girders, joists)   1,320,000 kg             1,848 t CO2e
steel deck                                      264,000 kg               660 t CO2e
roof insulation and membrane                     24,000 m2               600 t CO2e
total about 7,558 t CO2e, or 315 kg CO2e per m2 of footprint
per MW of IT capacity: 189 t CO2e; equal to the operational emissions of 40 MW for about 0.7 years on a 30 g/kWh grid

Concrete and steel dominate. In this rough estimate the building shell embodies about 7,500 metric tons of CO₂e, equal to less than a year of operating emissions on a very clean grid such as those of Quebec, Manitoba or Ontario, but only weeks on a coal-heavy grid. As grids decarbonize and data centers buy clean power, embodied carbon becomes a larger share of the total, which is why owners now target it through efficient structural design (avoiding oversized slabs and foundations), lower-carbon concrete and steel (Lesson 8), and reuse of standardized designs that have been optimized once and repeated. The figures here are indicative only; real assessments use project quantities and product-specific EPDs.

Handover and documentation

At completion, the civil and structural team delivers documentation that operations will rely on for decades: record (as-built) drawings and models; geotechnical reports and settlement monitoring data; design criteria including floor load capacities by area, roof load allowances for future equipment and hanging loads; test reports; warranties for roofing and other systems; and maintenance requirements such as roof inspection, sealant replacement and stormwater pond maintenance (Lesson 1). Floor load capacity plans, posted in each room, help operators avoid overloading floors when equipment densities rise, a real risk as AI racks become heavier.

The design project

The course project develops the civil and structural concept for a two-building data center campus of 2 × 30 MW IT capacity on a real or realistic site in Canada or the US. Working in teams, students produce a concept design report and drawings covering:

• Site analysis: topography, flood and stormwater, soils from a provided geotechnical report, utilities and access; grading concept and pad elevations with cut and fill estimate (Lessons 1 and 4).
• Stormwater management: rational method peak flows, detention sizing and quality controls (Lesson 1).
• Site utilities: fire water supply check, cooling water strategy, duct bank and fiber routes with diversity, fuel storage containment (Lesson 2).
• Foundations and slab: foundation type, footing sizes, settlement and any ground improvement, slab-on-grade design for rack and equipment loads (Lessons 3 and 4).
• Structure: structural system, grid coordinated with rack layout, load criteria including collateral and hanging loads, roof girder and joist sizing, expansion joints, lateral system (Lessons 5, 6 and 8).
• Natural hazards: snow, wind and seismic loads for the site with importance category, drift loads at rooftop equipment, anchorage concept (Lesson 7).
• Envelope: roof and wall assemblies, airtightness target, louver sizing (Lesson 10).
• Resilience and security: flood freeboard, hostile vehicle mitigation layout and barrier ratings, separation of redundant systems (Lesson 11).
• Delivery: prefabrication opportunities, construction sequence and schedule, QA plan, and an embodied carbon estimate with reduction measures (Lessons 9 and 12).