Building Envelope, Architecture & Site Civil Works

By the time you detail the building envelope and the site, the workload archetype (Chapter 1.1), the siting decision (Chapter 3.1), the building typology and hall layout (Chapter 6.1) and the structural basis for the dense liquid-cooled halls (Chapter 6.2) are already fixed. This chapter is where those decisions meet weather, soil, water and the threat model. It is also where many multi-billion-dollar projects pick up the defects that haunt operations: a skin that condenses on its inner face, a slab set one foot below the design flood elevation, a yard that cannot accept the next transformer because no one preserved the heavy-haul turning radius. None of these is a marquee decision. Every one is costly to undo once the hall is running.

The organizing tension differs from the rest of Part 6. A data center is not a building that happens to contain IT. It is a heat engine wrapped in a weather barrier, and its envelope has the opposite job of an office's. An office envelope holds heat in during winter and out in summer for the comfort of people. A data center envelope surrounds a building that is always rejecting tens of megawatts of heat, has almost no human occupancy, and is held at a tightly controlled interior condition (in the liquid-cooled era, often warm and dry) by a mechanical plant that never stops. Design the skin like an office and you get condensation, air leakage and a plant that burns parasitic power fighting the envelope. Design the site like a warehouse and you find out at delivery that the large substation power transformer cannot make the final turn.

Start from the load. A modern AI hall is a high-internal-gain space: the IT load dwarfs anything the climate does through the walls. A 30 MW hall rejects roughly 100 million BTU/hr of heat whether it is January in Stockholm or July in Phoenix. That inverts the usual envelope logic. Insulation in the walls and roof is no longer mainly about keeping conditioned air in, because the building is fighting to get heat out. It does three narrower jobs: limiting solar gain that adds parasitic cooling load, keeping the envelope's inner surface above the interior dew point so it does not condense, and protecting any air-cooled or hybrid spaces from extreme ambient swings during a mechanical failure.

The dominant moisture decision is the air barrier and vapor strategy, and it follows which way water vapor wants to move. In a hot-humid climate, vapor drives inward toward the cooler, drier interior. In a cold climate the interior is the warm, moist side and vapor drives outward. Put the vapor retarder on the wrong side of the assembly and you trap moisture inside the wall, where it condenses, degrades insulation and corrodes fasteners and steel, a failure that stays hidden until someone opens the assembly. Liquid cooling eases one part of this and sharpens another. Direct-to-chip liquid cuts the sensible air-cooling duty, but it does not remove the moisture load: make-up air and infiltration still have to be dried against the hall's dew-point basis. The CDU galleries, pipe racks and any chilled-water surfaces also create cold spots where condensation can form on the inside of the building, whatever the skin does, so check exposed piping against that dew point. The envelope and the mechanical design cannot be drawn by two teams who never speak.

The third envelope job is air-tightness, and it is where most commercial skins fail. Uncontrolled infiltration lets humid outdoor air into a hall the plant is trying to hold dry, which raises the latent load and the dew-point risk, while exfiltration leaks conditioned air out as parasitic loss. A continuous air barrier, verified by an air-leakage test and not just specified on paper, is the highest-leverage envelope detail. The choice is blunt: an air-barrier-led envelope (continuous membrane or fluid-applied barrier, sealed penetrations, tested) or a conventional commercial skin (metal panel or tilt-up with whatever air-sealing the trades happen to achieve). The first costs more per square foot and adds a commissioning step. The second saves that money once and pays it back every year in parasitic cooling load, dew-point incidents and, in the worst cases, corrosion of the structure it was meant to protect.

Write the acceptance test into the spec. For a large or multizone hall, use ASTM E3158 (the current edition is E3158-24) or an equivalent method the adopted code permits. ASTM E779 is not an unrestricted substitute for the code test on a large building. Before procurement, fix the test boundary, the HVAC-opening setup, the reference pressure, the envelope-area denominator and the reporting. Section C402.6.2 of the 2024 IECC caps commercial air leakage at 0.35 cfm/ft² of thermal-envelope area (above and below grade), tested at 75 Pa, for buildings inside that code's own applicability and compliance rules. Then write the owner's enclosure-release rule into the spec: a failed test means diagnose, seal and retest against the chosen target, even where a code exception or an alternate verification route would let the building pass. Plan the test-fan capacity, and any permitted portion testing, before the enclosure closes out.

The architectural plan resolves three programs at once: security zoning, physical hardening and operational flow. They pull against each other, and the building's quality depends largely on how gracefully the architect reconciles them.

Security zoning is a concentric model of public, semi-public, restricted and critical zones. Each boundary has to be a real physical and procedural threshold, not a line on a badge-reader schedule. The classic layering runs inward from the site perimeter (fence, standoff, vehicle control) through a staffed entry and visitor screening, then the back-of-house (loading, storage, gray space), and finally the white space. That innermost zone has the tightest access control, no exterior glazing and often a man-trap at the threshold. The architectural consequence is that circulation must give a delivery driver, a contractor, a network engineer and a security guard each a path that never crosses a more-restricted zone than their clearance allows. Get the adjacencies wrong and you are forever escorting people through space they should never have entered, or worse, you have a loading dock that opens onto a data hall.

Hardening is a threat-model decision: how far up the threat ladder do you design? At the low end sits standard commercial construction with a perimeter fence and CCTV. In the middle are forced-entry/ballistic-resistant (FE/BR) envelopes at critical thresholds, vehicle-barrier standoff (bollards, berms, ha-has) sized to keep a vehicle-borne threat outside a blast standoff distance, and blast-resistant glazing or no glazing at all on the critical core. At the high end, for sovereign, defense and hyperscale-sensitive workloads, come full blast hardening of the critical envelope, progressive-collapse-resistant structure and EMP/TEMPEST considerations. Each step up multiplies cost and constrains the plan, because standoff eats developable land and blast walls eat usable area. Set the tier in the design basis against a written threat model, not during value engineering, when the site no longer has room to add standoff.

A data center is a large, electrically dense, mission-critical object, exactly the kind of structure lightning finds. The lightning protection system (LPS) is the physical apparatus on the envelope and structure: air terminations (rods, mesh, catenary) that present the preferred strike point, down-conductors that carry the current to ground, and the structural-earthing interface that dissipates it. This chapter owns the physical LPS on the building. The electrical bonding, earthing regime, ground grid and surge-protective-device coordination are canonical in Chapter 4.11. Design the two as one system: an LPS that terminates into a poorly bonded ground grid, or down-conductors not bonded to the building steel and the signal-reference grid, creates exactly the ground-potential-rise transients that the SPD coordination in 4.11 exists to suppress.

The governing standards fork by jurisdiction. The international world uses IEC 62305, whose Part 2 (updated 2024) requires a quantitative, probability-based risk assessment that compares site-specific risk with tolerable risk and evaluates protection measures, including the selected Lightning Protection Level (LPL I–IV). The US world uses NFPA 780, a prescriptive installation framework that specifies component geometry without a calculation-driven protection level. Both lay out air terminations with the rolling-sphere method: roll a sphere across the building, and anything it touches needs protection. They set the radius differently. IEC 62305-3 ties it to the level: 20 m for LPL I, 30 m for II, 45 m for III and 60 m for IV. NFPA 780 uses one sphere for an ordinary structure, at most 150 ft, which the standard prints as 45 m (the exact conversion is 45.7 m), and a designer may pick a smaller one to be more conservative. That radius lands close to IEC's LPL III sphere, but a matching sphere does not make an NFPA layout an LPL III system. No risk assessment chose the NFPA layout. An IEC level sets more than the sphere: it also fixes the design lightning current, the mesh size and the spacing of down-conductors. NFPA has no design current at all, and it prescribes spacing and conductor sizes under rules of its own. NFPA does not prohibit denser geometry either, so an owner or insurer who wants LPL I or II density has to spell it out.

On the IEC route the level is an output of the risk assessment plus the owner's and insurer's requirements, not a sector default. Run the assessment and record the level. If it lands at LPL I, the 20 m sphere means a dense mesh of air terminals and frequent down-conductors, and that geometry must be reconciled with the roof-mounted mechanical equipment, any PV and the air-handling penetrations before the roof is detailed.

The high-leverage decision is whether to use the structural steel as the down-conductor and earthing path (a natural-component LPS, where the building's own frame and rebar carry the current to a perimeter ground ring) or to run a dedicated external conductor system. The natural-component approach is cheaper, more robust and cleaner to look at. It also demands electrical continuity through the structure, designed and verified at the structural stage, bonding rebar and steel connections that a structural engineer would not otherwise care about. Decide late and you retrofit bonding straps onto a frame never detailed for continuity. Decide early, with the structural basis (Chapter 6.2) and the earthing basis (Chapter 4.11), and the building's own steel becomes the cheapest Faraday cage available.

The site civil works are the least visible and most schedule-determining scope on the project. They also fix how the campus handles water and weather, and the roads, pads and foundations that carry the heaviest equipment it will ever receive.

Grading and stormwater are coupled and, on large AI campuses, often binding. Roofs, parking, generator yards, switchyards and heavy-haul roads replace permeable ground with impervious cover. More impervious cover means more runoff, but the runoff coefficient does not simply equal the covered-area percentage. The site has to hold back its design runoff and release it at the controlled rate the governing rules set. Municipal, state and watershed criteria size the detention, water-quality treatment, channel protection and flood routing. Derive the volume from the design storm you select, the catchment and the permitted discharge; one campus’s storage does not size another’s. The civil fork is surface detention ponds versus underground detention. Surface ponds are cheap but consume developable acres the campus would rather use for the next building or a BESS yard. Underground systems preserve the footprint but cost far more per cubic foot. Stormwater is therefore a site-plan variable, not an afterthought: detention requirements reshape the buildable area and how many halls the parcel can hold. Whether the project needs NPDES construction coverage depends on the discharge, the operator, the acreage or common plan, program eligibility and waiver status. A covered project needs a stormwater pollution-prevention plan, and that permit sits on the critical path alongside the air and power permits (Chapter 3.9).

Utilities, roads and heavy-haul access are where civil work meets the long-lead equipment supply chain. The heaviest, most schedule-critical object the site will receive is the load-serving HV/substation power transformer, often 150-400+ tonnes. Wood Mackenzie’s May 2026 assessment puts US substation-transformer lead times above 160 weeks (Chapter 2.3). Road geometry, turning radii, bridge and culvert load ratings and pad access must be designed for that transformer from day one. Get the heavy-haul route wrong and a transformer that took two-plus years to build sits at a weigh station while you re-engineer a turn. The same logic extends to prefabrication: if the project buys factory-built power and cooling modules (Chapter 6.4), the roads, laydown yards and crane pads must accept the largest module the transport plan specifies. Civil access constrains the procurement strategy; it is not something to accommodate after procurement is settled.

Follow the water, not the roof-area percentage. Rain inflow is 1,800 m² × 0.040 L/s per m² × 1.0 = 72 L/s. With both outlets available, their combined 90 L/s capacity at 50 mm exceeds the inflow, and the assumed curves give an equilibrium depth of 0.050 m × (72/90)², about 32 mm. One blocked outlet leaves only 45 L/s at the penetration level, and less at every lower head. Subtracting a constant 45 L/s would overstate the time available.

Carry storage and head together: 300 m² × 1,000 L/m³ × dh/dt = 72 − 45√(h/0.050 m) L/s. With u = √(h/0.050 m), integrate from empty storage to u = 1: t = 2 × 300 × 1,000 × 0.050 × [−u/45 − (72/45²) ln(1 − 45u/72)]₀¹ seconds. The penetration is reached in about 380 s, or 6.3 minutes, inside the ten-minute screen. Stop the roof-only calculation there, because water can then enter the cable route toward the electrical room. Starting with stored water shortens that time.

Reject the existing path. Select the independent emergency discharge, route it away from the door, intakes and heavy-haul corridor, and check its supports and ponding load. With both primaries blocked, qₑ = 72 L/s gives h = 0.010 + 0.040 × (72/90)^(2/3), about 44 mm, below the 50 mm penetration in this model. The one-primary crossover is rainfall of 45 L/s ÷ 1,800 m² = 0.025 L/s per m². Below it, the surviving outlet reaches equilibrium below the penetration; at equality it approaches that level asymptotically. It is not a no-storage threshold, because gravity discharge needs head.

Passing the drainage check does not accept the door. Water-test the installed door and the adjacent service penetration, with the pressure, spray rate, duration and inspected inner plane written into the spec. ASTM E1105-15(2023) is a standard field test of installed windows, skylights, doors and curtain walls for water penetration; it may also be applied to their joints with adjacent construction, and for the penetration itself the designer picks a suitable method. Water past the inner plane fails: repair the flashing, seals and water path, then repeat the same exposure. Release the enclosure only on a passing project test record.

The yard is where the campus master plan either holds together or fights itself for 20 years. It is a packing problem with hard constraints: the substation and switchyard (Chapter 4.3), the backup generators and their fuel storage (Chapter 4.9), the chillers, cooling towers and dry-coolers (Chapter 5.7), the BESS, and increasingly behind-the-meter gas generation each demand area, setbacks and adjacencies that conflict.

Fire, code, acoustics and emissions drive the setbacks, and several are owned in detail by other chapters. Fuel storage carries fire-code separation distances and spill-containment requirements (Chapter 4.9, Chapter 6.5). BESS has become the most consequential new setback driver: lithium-ion thermal-runaway propagation has pushed codes (NFPA 855 and equivalents) toward larger separation distances and dedicated deflagration and exhaust provisions. The BESS yard’s distance from the building must follow the applicable code, the listed configuration and accepted fire and gas test data (Chapter 6.5). Generators and their acoustic enclosures and emissions stacks must be placed for exhaust dispersion and community-noise compliance (Chapter 6.8, Chapter 3.11). Evaporative cooling towers need separation from air intakes (Legionella and recirculation) and from property lines for plume and drift. The yard layout is the physical reconciliation of half a dozen other chapters' constraints into one buildable plan.

The recurring decision is compact-but-coupled versus spread-but-resilient. A tight yard saves land and shortens cable and pipe runs, but it concentrates risk: a BESS fire near the generators near the fuel near the substation is a single-event cascade. A spread yard buys fire breaks, maintenance access and concurrent maintainability, at the cost of land, longer runs and a bigger civil scope. The frontier-AI default has shifted toward more separation as BESS and behind-the-meter gas have grown. The right answer is site-specific, and once the pads are poured and the underground utilities run, it is very costly to reverse.

The envelope and site basis is ultimately a bet on the local climate and hazard regime, and one upstream decision carries that bet into nearly every structural and envelope detail: the Risk Category. Under ASCE 7-22 (and analogous provisions elsewhere), most data centers are code-assigned Risk Category II. Many owners still ask their designers for more resilience than that, and NIST's 2023 report Resilience for Critical Facilities (GCR 23-037) finds the usual way is to follow Risk Category IV criteria, the tier that covers hospitals and emergency-response facilities. That choice selects hazard-specific seismic, wind, flood and snow/ice design criteria. For seismic design it raises the importance factor from 1.0 to 1.5, which is not a multiplier applied to every load. Freeze the choice before construction. It sizes the structural steel, the foundation, the envelope's wind-pressure rating and the flood elevation of the slab, and changing it after the hall is live can mean adding steel, reworking foundations and swapping envelope assemblies while the building keeps running.

Buying that margin at design costs little against the whole build. The same NIST report finds a data center's structural system a small share of facility cost, with the MEP systems a much larger one. In Mastt's breakdown of a typical greenfield build, shell and structure take about 13–14% of construction cost, against 15–20% for mechanical, so a 10% premium on shell and structure adds only about 1.3–1.4% to construction cost. Over-build the structure, the slab elevation and the envelope ratings.

Choosing Risk Category III or IV can also bring in tornado design. If your adopted code includes 2024 IBC §1609.5 and ASCE 7-22 Chapter 32, check whether an RC III or IV building in the mapped tornado-prone region must pass that chapter's screens. Find the tornado speed from the site and the building's effective plan area, then apply the tornado-speed and wind-speed/exposure screens in §32.5.2. Where tornado loads apply, the main system, components and cladding are designed for whichever of wind or tornado governs. Tornado speeds use 1,700-year return periods for RC III and 3,000-year for RC IV. RC II buildings face no mandatory tornado design under Chapter 32. If an owner adopts enhanced criteria voluntarily, write tornado performance into the basis, including anchorage of the roof, panels and rooftop equipment. ICC 500 storm shelters are a separate matter from Chapter 32.

The hazard-specific hardening then follows the regional threat. In hurricane and high-wind regions, the envelope needs impact-rated assemblies, missile-resistant glazing or none, and roof systems rated against uplift and the wind speeds the updated ASCE maps now demand. In seismic regions, the structure and any base isolation (ASCE 7 Chapter 17) follow the structural seismic provisions. The anchorage of heavy plumbed racks and yard equipment follows ASCE 7 Chapter 13 for nonstructural components, or the nonbuilding-structure provisions where the item's support configuration puts it there. Rack-civil integration is owned in Chapter 6.7, and the geotechnical and seismic basis in Chapter 3.8. In flood-prone sites, pick floor and equipment elevations for critical spaces from the flood rules your jurisdiction adopted, the site's flood design class and what the owner requires the building to survive. Dry-floodproof the envelope to that line and keep all critical electrical and mechanical equipment above it. Where ASCE 7-22 Supplement 2 or ASCE 24-24 apply, follow their class-specific design flood and minimum elevations, add relative sea-level rise, and honor any stricter community requirement. The guide's screen for a critical facility is the 500-year (0.2%-annual-chance) flood plus freeboard, derived in Chapter 3.8 along with the FEMA NFHL or equivalent mapping to check at diligence. In wildfire, hail and extreme-heat regions, the envelope needs defensible space, ember-resistant detailing, impact-rated roofing and louver/intake protection. These regional deltas are consolidated as a quick-reference crosswalk in Appendix G. They are envelope and site decisions, made early, that the building lives with for two decades.

Freeze the overflow path, the enclosure test boundary and the heavy-haul turns before closing the roof and pouring the yard. If stormwater or the next transformer has to find its own route, electrical rooms and delivery dates are exposed to the cheapest omitted detail.