Skip to content
Request a quote Log in
🧮 Tools 🗄️ Equipment list 📰 Insights 🎓 Learning Hub 💬 Forum ✉️ Contact us Request a quote

Container Data Center Insulation: Requirements That Decide It

Shopify API condensation, container data center, EN 13501, GR-487, insulation, modular, NFPA 285, thermal bridge, U-value

A bare 40-ft steel shell absorbs 13.5 kW at a 16 K design difference. The insulation math, the frame geometry that triples the U-value, the condensation thresholds, fire class and solar load - all computed.

A shipping-container data hall is a steel box with an R-value of nearly zero. Every square meter of that box is either a heater in summer or a condensation plate in winter, and the insulation layer is the only thing standing between the two. Yet insulation is usually specified the way paint is specified — by thickness and price, from a brochure line item — and the three properties that actually decide whether it works, fire behavior, frame geometry, and vapor side, are the ones the brochure does not lead with.

This article works one example end to end: a 40-ft high-cube container hall, 12.2 × 2.44 × 2.90 m, 144 m² of envelope, 100 kW of IT, 24 °C setpoint, 40 °C design ambient. We compute what the bare shell costs, what each build-up buys back, where the frame destroys the board number, when the frame sweats, and which side the vapor barrier belongs on. The insulation itself turns out to be the easy part.

40-ft container data center module with corrugated steel envelope
Fig. 1 — A 40-ft container module. The corrugated shell is structural; everything thermal happens in the layer behind it, and every fastener, post and door edge in that layer is a decision.

The one-paragraph version: an uninsulated 40-ft shell absorbs 13.5 kW at a 16 K design difference — 13.5 percent of a 100 kW pod's IT load, before the cooling bill for that heat. Fifty millimeters of proper insulation cuts that to under 1 kW. But stuffing the insulation between the container's frame posts instead of putting it outside the frame triples the real U-value, and a bare steel post inside the insulated wall sits at the mold-growth threshold at 50 percent indoor humidity. Specify fire class first, frame geometry second, vapor side third; thickness comes last and is easy.

What the bare shell actually costs

Start with the number vendors skip. Steel is 45–50 W/m·K; a 2 mm skin has a thermal resistance of essentially nothing. Per ISO 6946, a wall with horizontal heat flow carries surface resistances of 0.13 inside and 0.04 outside [1], so the bare shell total is R 0.17 m²K/W and U ≈ 5.85 W/m²K. At the 40 °C design day against a 24 °C hall:

Q = 5.85 × 144 m² × 16 K ≈ 13.5 kW

That is a 13.5 kW heater strapped to a 100 kW pod — 13.5 percent extra cooling load on the hottest afternoon, and roughly 6.8 percent on the pod's annual energy bill once averaged over the year (59,000 kWh at an 8 K mean difference, about $7,700/yr at $0.10/kWh including a 30 percent cooling overhead). Add solar and the afternoons get worse; we will quantify that below.

Now put insulation in. The same calculation with real build-ups:

Envelope build-up (walls + roof) R (m²K/W) U (W/m²K) Design-day gain (kW) Avg. annual cost
Bare steel shell 0.17 5.85 13.51 $7,690
50 mm mineral wool (λ 0.035) 1.60 0.63 1.44 $823
100 mm mineral wool 3.03 0.33 0.76 $434
50 mm PIR (λ 0.022) 2.44 0.41 0.95 $538
75 mm PIR 3.58 0.28 0.65 $367

Table 1 — U per ISO 6946 for the whole 144 m² envelope at Rsi 0.13 / Rse 0.04; annual cost at $0.10/kWh, 8 K mean ΔT, ×1.3 for the cooling that removes the heat. Material cost for the layer itself: roughly $4,000–6,600 per container.

Design-day heat gain bars and frame geometry comparison
Fig. 2 — Left: what each build-up buys back against the bare shell's 13.5 kW. Right: the same 75 mm PIR is worth three times more outside the frame than between the posts.

Two facts fall out of Table 1. First, the step from nothing to any real insulation is worth 90–95 percent of the benefit; the step from 50 mm to 100 mm is worth a few hundred dollars a year. Second, the payback on the insulation package itself is under one year in hot climates, which is why the interesting engineering is not the thickness but everything that can silently void the number in the second column.

The code baseline, converted to units you can buy

ASHRAE 90.1 prescribes envelope values by climate zone. For a climate-zone-1 nonresidential building, the metal-building roof maximum is U-0.041 and the metal-building wall maximum is U-0.094 in IP units [2][3] — that is 0.23 and 0.53 W/m²K in SI. A container hall is a conditioned space, not a semiheated one, so those are the numbers an AHJ or a lease auditor will hold you to:

  • Roof at 0.23 W/m²K: about 100 mm PIR or 130 mm mineral wool, after film resistances. Note the roof matters more than the wall per square meter because it takes the sun.
  • Walls at 0.53 W/m²K: about 50 mm PIR or 75 mm mineral wool. Both fit comfortably in a container wall cavity, which is typically 60–100 mm deep between the corrugated outer skin and an inner liner.

European procurement will see the same targets expressed as U ≤ 0.25–0.35 W/m²K for the roof in national building codes. Either way, the datasheet U of the insulation board is not the assembly U — the next section is why.

The frame is the spec, not the board

A container is not a stud wall. It is a corrugated steel skin welded to posts, rails and corner castings, and those members are continuous steel paths from the outer skin to wherever the inner liner stops. Watch what happens to the board U-value under the two ways to build the wall:

  • Design A — insulation between the posts (the cheapest, most common build): suppose the posts and liner contact points occupy 10 percent of the wall area as unbroken steel. The post path alone has R ≈ 0.17 — basically the air films. The area-weighted result for the 75 mm PIR wall is Ueff = 0.84 W/m²K, three times the board's 0.28, and the design-day gain climbs back to 1.9 kW. You paid for 75 mm and got the performance of about 12 mm.
  • Design B — continuous insulation outside the frame, frame on the warm side, liner over it: the board number survives essentially intact (0.28 vs 0.28), because there is no metallic path that bypasses the insulation layer.

This is the single most consequential sentence in a container insulation spec: insulation shall be continuous across the structural frame; the frame shall not bridge from outer skin to inner liner. It is also the detail that separates a purpose-built container data center from a converted shipping container with fiberglass pushed into the corrugations.

Where the money hides: doors, hinged access panels, cable entry plates, and the floor edge. Each is a frame within the frame. A door leaf without a thermal break is a 2 m² post; a cable entry plate bolted skin-to-liner is a point bridge with its own dew-point microclimate. Count them, ask for their surface temperatures in the U-value calculation, and do not accept an assembly U that only models the field of the wall.

Condensation: the failure mode that actually destroys containers

Heat gain is a cost problem. Condensation is a reliability problem — wet insulation loses its R-value permanently, wet steel corrodes from the inside where your coating cannot reach, and damp mineral wool in a warm cavity grows mold. The physics is settled and codified: EN ISO 13788 judges a surface safe if its temperature keeps surface relative humidity below 80 percent, quantified by the temperature factor fRsi [4][5].

Run the winter case. Interior 22 °C, outdoor −10 °C:

Point in the envelope fRsi Surface temp at −10 °C Verdict at 22 °C / 50% RH (dew point 11.1 °C, mold limit ≈14.6 °C)
Plain insulated wall (100 mm MW) 0.96 20.6 °C Dry at any RH the hall will ever see
Bare steel post (Design A) 0.76 14.4 °C At the mold threshold at 50% RH; below it at 60% RH

Table 2 — The post is only 6 K above the hall's dew point in the winter case. Data halls hold low RH by design, which is the only reason Design A survives at all — and one humidification fault or door-left-open week turns every post into a drip line.

Winter surface temperature of wall versus frame post against dew point bands
Fig. 3 — Surface temperature of the plain wall and the bare post as a function of outdoor temperature, against the dew point of 22 °C air at three humidity levels. The wall never crosses; the post lives one dehumidification fault away.

Hot-humid climates run the same physics in reverse, and this is where the vapor barrier side is decided. In Singapore or coastal China the exterior air carries a dew point of 26–28 °C for months while the hall sits at 24 °C: the vapor drive is from outside in, all year, so the low-permeability layer (foil-faced PIR's own facer, or a dedicated membrane) belongs on the outside of the insulation, and the inside face must stay vapor-open so the assembly can dry inward. In heating-dominated climates the drive reverses and so does the barrier. The classic container failure is sealing both faces — a foil board butted against a sealed inner liner — which turns the cavity into a humidified trap the first time the temperature gradient flips [6].

One more thing diffusion math will not save you from: air leakage beats diffusion by orders of magnitude. A gap at a cable entry or an unsealed liner joint transports humid air — and its moisture — into the cavity far faster than vapor diffusion through the board face ever could. The barrier is only as continuous as the sealing tape, and every penetration is a hole in it. This pairs with the envelope airtightness targets in our container deployment analysis and the dew-point arithmetic in our dew point guide.

Fire class: pick the material by what it does in a fire, not by λ

The insulation layer is the largest combustible mass in the container after the cable plant. Under EN 13501-1 the materials divide cleanly [7][8]:

Material Euroclass λ (W/m·K) Thickness for the wall target (U 0.53) Notes
Stone mineral wool A1 / A2-s1,d0 0.035–0.040 ≈75–80 mm Non-combustible, ~1,000 °C fiber tolerance, best acoustics
Glass mineral wool A2-s1,d0 0.035–0.044 ≈75–85 mm Non-combustible core, binder is the small organic fraction
Phenolic foam B-s1,d0 0.019–0.022 ≈45–50 mm Best λ-to-fire ratio, priciest
PIR board / panel core B–E (system-dependent) 0.020–0.023 ≈45–50 mm Classified as a sandwich panel system, not a bare board
EPS / XPS E 0.033–0.036 ≈70–75 mm Never exposed; needs full protective linings

Table 3 — Reaction-to-fire per EN 13501-1 [7][8]. Thickness figures are the insulation layer alone; assemblies add films and liners that help slightly.

Two nuances matter for containers specifically. First, the product that actually ships is a sandwich panel — steel facing, foam or wool core — and EN 13501-1 classifies the system: a mineral-wool-core panel with steel facings is routinely A2-s1,d0, while the same PIR core ranges B to E depending on thickness and facing [7]. Ask for the panel certificate, not the foam datasheet. Second, in North America the relevant demonstration for combustible insulation in an exterior wall is the NFPA 285 assembly test — a two-story mock-up with a window flashover fire, pass criteria of no vertical flame spread beyond 10 ft above the opening, no lateral spread beyond 5 ft, and temperature limits in the cavity [9]. IBC triggers NFPA 285 above 40 ft of building height, which a container never reaches — but AHJs, insurers and hyperscale lease auditors ask for the report anyway, and FM Global-insured data centers will look for FM 4880-classed panel systems [10]. The practical hierarchy for a container hall: mineral-wool-core sandwich panel (A2-s1,d0) as the default; PIR only where depth is genuinely constrained and the panel carries a system fire certificate; EPS never.

Fibrous-shedding deserves one line as well: mineral wool inside an air-cooled hall must be faced so fibers cannot migrate into the airflow path and onto server heat sinks — a mat-faced or foil-faced board with sealed joints, not raw wool behind a perforated liner.

Solar: the paint is doing insulation's job on a bare box

Telcordia GR-487, the telecom outdoor-enclosure standard, tests enclosures under 753 W/m² of simulated sun with the electronics running [11]. Field guidance from the same enclosure industry puts the absorbed-solar effect in plain numbers: on a still day a dark enclosure skin runs 30–40 K above air temperature, a white one 5–10 K [11][12]. In sol-air terms, with an exterior film coefficient of 20 W/m²K:

Finish Solar absorption α Sol-air add-on Roof gain, bare shell Roof gain, 100 mm MW
White 0.14 +5.3 K 940 W 52 W
Light grey (RAL 7035) 0.50 +18.8 K 3,360 W 185 W
Medium grey (ANSI 61) 0.75 +28.3 K 5,040 W 278 W
Dark grey / black 0.95 +35.8 K 6,390 W 352 W

Table 4 — GR-487 irradiance (70 W/ft²) on the 29.7 m² roof, conduction only. On a bare box the finish choice is worth ±5 kW; once the roof is insulated it is worth ±300 W.

The reading: insulation is what makes the paint decision cheap. On an insulated container you can paint it any color the site aesthetics want; on a bare one, dark paint is a permanent 5 kW load. This is also where the insulation layer and the anti-corrosion system meet — the solar-facing surface finish, its absorptivity drift with age, and its repaint cycle are covered in our exterior coating maintenance guide and the ISO 12944 coating selection guide.

Where insulation must not go, and the details that carry the spec

  • Never across an airflow path. If the hall breathes through a floor plenum, a ceiling plenum, or a wall of escape louvers on the cooling side, the insulation stops at that boundary with a sealed, rigid termination. Loose board edges in a plenum shed and rattle.
  • Floor. The container floor sits on a plinth; ground-coupled and rain-splash wetted. XPS or high-density EPS below the floor deck (closed cell, tolerates wet), not mineral wool, and a sealed deck edge.
  • Doors and hatches. Thermal break in the leaf, compression seals all around, and an insulated reveal so the door frame is not a post in disguise.
  • Cable entries and pipe penetrations. Air seal plus fire-stop in the same detail — the vapor barrier and the fire barrier fail at the same holes.
  • Roof walkways and attachments. Every screw through the outer skin into the structure is a small Design-A bridge; use standoff fixings or pad the count.
  • Battery or power rooms inside the same box: those compartments have their own temperature bands; insulate and condition them separately rather than hoping the hall wall does the job.

Verification: what to demand before the container ships

  1. U-value calculation per ISO 6946, combined method, with the frame fraction and every bridge (door, entry plate, fixings) modeled explicitly — not the board datasheet [1].
  2. fRsi report per EN ISO 13788 for the worst junction, checked against your interior design RH [4][5].
  3. Fire certificates: EN 13501-1 class for the panel system (or FM 4880 where the insurer asks), plus NFPA 285 test report if the assembly contains foam [7][9][10].
  4. Thermography at FAT with the hall at least 10 K above ambient: scan the interior liner; any post-shaped hot streak is Design A announcing itself. Same camera pass after the first winter on site.
  5. Surface temperature spot checks at door frames and entry plates against the hall dew point, logged over 24 h.
  6. Moisture: cavity inspection ports or at minimum a humidity logger inside the wall cavity for the first summer, plus the vapor-barrier continuity record (seam map, tape batch, penetration count).

Ten mistakes that show up in year two

  1. Specifying the board U-value and accepting the assembly U-value with an unbroken frame.
  2. Insulation between the posts because it was cheaper per square meter — it is the most expensive square meter in the box.
  3. Vapor barrier on the wrong side for the climate, or on both sides.
  4. Unfaced mineral wool shedding fibers into the cooling airflow.
  5. EPS anywhere it can see daylight, a heat gun, or a fire marshal.
  6. No thermal break in the door leaf; a 2 m² steel post with hinges.
  7. Cable entry plates bolted skin-to-liner with no gasket, no break, no fire-stop.
  8. Dark roof paint chosen for looks on a lightly insulated box (Table 4 says what it costs).
  9. Floor insulated with wet-tolerant-unfriendly materials, or not at all, then wondered about in the rainy season.
  10. No thermography at FAT — the frame bridge is invisible until the first winter, which is exactly one warranty cycle too late.

Eight questions for the container supplier

  1. What is the calculated assembly U-value per ISO 6946, and which bridges did you model?
  2. Is the insulation continuous across the frame posts, or between them? Show the section drawing.
  3. What fRsi does the worst junction achieve, and what interior RH does that protect?
  4. Which vapor barrier arrangement is installed, and for which climate direction was it chosen?
  5. What is the EN 13501-1 class of the panel system (not the foam), and is an NFPA 285 or FM 4880 report available?
  6. How are the door leaf, cable entry plate, and roof fixings thermally broken?
  7. What sol-air temperature or solar load was the cooling design sized for, and at what finish absorptivity?
  8. Can you support a thermography acceptance at FAT, and re-inspect after the first winter?

Insulation is the cheapest capacity you will ever add to a container data center — a few thousand dollars that remove a 13 kW parasitic load and a corrosion mechanism in the same stroke. The requirements that make it work are written in the details: where the frame sits, which side the vapor comes from, and what the layer does in a fire. Get those three right and the thickness takes care of itself. For a build-out where the envelope is engineered as a system, see our modular container guide or talk to us about a specification.

References

  1. DataDrivenAEC. How to Calculate U-Value: Formula, Targets and Code Limits — ISO 6946:2017 method, surface resistances Rsi 0.13 / Rse 0.04 by heat-flow direction, worked examples. datadrivenaec.com
  2. ASHRAE. Standard 90.1-2019 Envelope Requirements — Table 5.5 climate-zone values for metal building roofs and walls (U-0.041 roof / U-0.094 wall, Climate Zone 1 nonresidential, IP units). State of Oregon Building Codes Division training summary. oregon.gov (PDF)
  3. MBMA / Thermal Bridging Solutions. Energy Code Compliance: A Guide for Metal Building Contractors — prescriptive metal-building insulation values by 90.1 edition and climate zone, continuous-insulation definitions. thermalbridgingsolutions.com (PDF)
  4. Build InSite. Condensation in Buildings — EN ISO 13788 fRsi method, the 80 percent critical surface humidity criterion and the 0.75 dwelling benchmark. buildinsite.io
  5. IIETA. Hygrothermal Analysis of CLT-Based Retrofit Wall Assemblies According to the EN 13788 Standard — worked fRsi/fRsi,min computation from critical saturation pressure with the 80 percent critical surface humidity criterion. iieta.org
  6. ISOCELL. Surface Condensation and the fRsi Check — dew point vs. surface temperature profiles, sd-value drying strategy ("tighten the inside or open the outside, never both"), five-to-one vapor-resistance ratio. isocell.ie
  7. BOKKA. Reaction-to-Fire Euroclasses A1/A2/B/C/D/E/F in Design Practice — EN 13501-1 classifications for mineral wool (A1/A2-s1,d0), PIR (B–E system-dependent), EPS (E), sandwich-panel system classification. bokka.pl
  8. SpecForge. Insulation Board — Euroclasses by material family, λ ranges, thickness-for-target arithmetic, facer and moisture behavior. sourcebyspec.com
  9. RMAX / IBC context. NFPA 285 Requirements: Understanding Exterior Wall Assembly Fire Testing — two-story assembly test, window-flashover exposure, vertical/lateral propagation and cavity temperature criteria, IBC triggers above 40 ft. rmax.com
  10. Phoenixx SmartBuild. PIR Panel vs Mineral Wool Panel — Fire Safety & Insulation Compared — FM 4880/4881 approvals, Euroclass comparisons, λ and weight tables for data-center applications. phoenixxsmartbuild.com
  11. Hoffman / Pentair. Heat Dissipation in Electrical Enclosures & Solar Heat Gain — Telcordia GR-487 solar test at 70 W/ft², absorption by finish color, solar-load and temperature-rise charts. Hoffman application note (PDF)
  12. CYTECH. How To Perform Thermal Testing for Outdoor Telecom Cabinets — GR-487/GR-3108 and ETSI EN 300 019 thermal classes, solar-load test setup, internal temperature targets. cytechorg.com