A data-center-grade motor control center in 2026 is a 380/220 V, 1000 A low-voltage lineup with 25 kA short-circuit withstand, IP54 enclosure, and 8 or more soft-starter branches feeding chilled-water pumps, CRAH/CRAC fans, and condenser water motors [S5]. Freedom 2100 and Model 6 units are engineered to NEMA ICS 18, which sets the seismic, wiring, and withdrawability rules that hyperscale operators expect in a 24/7 facility [S2][S3].
Data centers are now the second-fastest growing segment of the global MCC market, with the total category tracked at USD 7.01 billion in 2026 and a 7.93% CAGR toward USD 12.91 billion by 2034 [S1]. Inside a 5-50 MW hall, the MCC is no longer a passive bucket lineup: it is the local intelligence layer for cooling and the most common service access point for electricians during a live white-space window.
What the MCC actually has to do in a data center
Inside a hyperscale hall the MCC feeds the mechanical plant that decides PUE: chilled-water primary/secondary pumps, condenser-water pumps, cooling tower fans, CRAH/CRAC blowers, humidifier reheat, and the small UPS-input ventilation set. Eaton frames the product correctly: an MCC is a common bus, vertical sections, wireways, and modular motor units built to NEMA ICS 18, designed to be safe to work on while adjacent sections stay live [S2]. Schneider's Model 6 documentation treats the MCC as a long-life asset, with explicit instructions on shipping, moisture mitigation, and arc-rated upstream over-current device coordination [S3].
The cooling-loop topology drives the MCC one-line. A typical 1N+1 chiller plant expects 6-8 soft-started pump feeders, 4-12 VFD-driven fan feeders, and 2-4 direct-on-line utility feeders on a single lineup. The 1000 A reference design documented by EEP packs 8 soft-starter branches plus motorized feeders into a single Prisma enclosure, with thermostat-controlled cooling for the soft-starter heat load [S5]. For the broader switching and protection hierarchy, the same project also has to select a feeder distribution cabinet on the upstream side.
Selection criteria that actually move PUE and uptime
Four technical numbers separate a data-center MCC from a general-purpose industrial lineup, and each one is verifiable in vendor documentation [S1][S2][S3][S5].
Bus rating and fault duty. A 1000 A horizontal bus with 25 kA short-circuit withstand covers most 380/220 V 50 Hz data-center service entrances feeding a 1-2 MW chiller skid; larger campuses with on-site medium-voltage transformation should be spec'd at 65 kA or higher to match the upstream breaker [S5].
Enclosure and ingress. IP54 (dust-protected and splash-proof) is the current indoor data-center norm, because mechanical rooms share airspace with humidifiers and occasional condensate; NEMA 12 is acceptable in clean electrical rooms, NEMA 4 is reserved for outdoor walk-in cooling yards [S5].
Starter mix. Soft starters dominate chilled-water pumps to limit inrush on diesel-generator backfeed; VFDs dominate tower fans and CRAH blowers for part-load efficiency; direct-on-line contactors are restricted to small humidifier and exhaust loads. Reference designs show 8 dedicated soft-starter branches per lineup, each with bypass contactor and feeder metering [S5].
Intelligence layer. Modern data-center MCCs ship with networkable overloads, branch metering, and gateway modules so that motor status feeds the process control layer and the BMS, rather than living on local pilot lights. Schneider's Model 6 line and Eaton's Freedom 2100 both document Ethernet/IP and Modbus TCP options for the intelligent buckets [S2][S3].
Comparison of the main data-center MCC platforms
Across the 2026 vendor set, four platforms dominate the data-center specification. The table below lines them up against the criteria that actually matter for a cooling-loop lineup; the same comparison logic is used in the MCC buyer's guide for industrial builds [S1].
Eaton Freedom 2100. NEMA ICS 18 construction, arc-resistant option, full intelligent-bucket family, and a long spare-parts window. Suits operators who already run Freedom and want bucket interchangeability across multiple data-center sites [S2].
Schneider Square D Model 6. 100 ms arc-rated design with documented upstream OCPD coordination, 1000 A/25 kA demonstrated in the published schematics, and a heavy emphasis on moisture and contamination control during shipping and storage [S3][S5].
Allen-Bradley CENTERLINE 2100. Common in U.S. greenfield builds, strong integration with Rockwell Logix controllers and EtherNet/IP, and a deep installed base in U.S. hyperscale [S6].
ABB/Siemens and channel-rebuilt custom lineups. Specified when project engineering demands Prisma-style enclosures, mixed-vintage buckets, or a panel-builder with regional engineering depth. Elcontech-style panel builders also integrate Eaton, Allen-Bradley, GE, and Siemens components into a single UL 845 lineup for retrofit and brownfield work [S6].
For hyperscale operators the practical ranking on data-center fit is Freedom 2100 and Model 6 first, CENTERLINE 2100 first for U.S. Logix-native sites, custom Prisma/panel-builder lineups first where footprint, seismic, or arc-withstand is non-standard. A side note on the wire-and-cable package: long soft-starter runs from a control cable tray back to the motor must be spec'd for the VFD or soft-starter PWM signature, not as plain THHN.
Soft starters, VFDs, and the PUE trade-off
Soft starters are the default for pumps in data-center chilled-water plants because they cut starting current to 3-5x FLA instead of 6-8x for direct-on-line, which keeps the upstream diesel-gen set in its continuous-rated window and reduces voltage dip on the IT bus [S4][S5]. Solcon Igel's data-center brief recommends soft starters specifically to lower PUE by reducing I²R losses in the chilled-water distribution and limiting mechanical stress on pump bearings during daily on-off cycles [S4].
VFDs go on the fan side: cooling tower fans, CRAH blowers, and condenser-water pumps where flow must track IT load. The 1000 A reference design keeps VFDs and soft-starters in separate compartments, with thermostat-controlled fans on the soft-starter wall to keep the thyristor heatsinks inside their 0-60 °C ambient rating [S5]. Where the MCC must hand motor data up to a PLC control network, both starter families expose run, fault, current, and thermal-state registers, which lets the plant controller stage chillers without a separate RTU.
Who an MCC is for, and where it is the wrong answer
An intelligent LV MCC is the right answer for any data center with more than 2-3 MW of cooling load, multiple chillers, and at least two utility feeders that need selective coordination [S1][S2]. It is also the right answer for retrofit work in a live white-space, because withdrawable buckets cut maintenance window from hours to minutes and let a faulted pump be isolated without opening the IT bus.
It is the wrong answer for a single 200 kW edge site, where a wall-mounted control valve panel and a packaged chiller controller will do the same job for a fraction of the footprint and cost. It is also the wrong answer when the upstream MV-LV transformer and the data-center bus-tie scheme cannot be co-engineered with the MCC supplier, because the arc-flash and selective-coordination story falls apart without a single point of design responsibility [S1][S6].
Failure modes, constraints, and what to watch on the bid day
Three failure modes dominate data-center MCC field calls, and each one is preventable at spec time. First, condensation inside a soft-starter wall during a chilled-water pipe leak; Model 6 documentation flags moisture-avoidance as a commissioning prerequisite and rejects energization on a wet lineup [S3]. Second, arc-flash incident energy above 40 cal/cm² because the upstream breaker was not coordinated with the bucket short-circuit current rating; the published 1000 A reference is rated 25 kA at 380/220 V and assumes a coordinated main breaker [S5]. Third, access control gaps: contractors and OEM field-service engineers need to be able to isolate a single bucket without locking the whole lineup, so spec the MCC with a per-section Kirk-key or castell interlock, not a single lineup-wide padlock.
Constraint to plan around: data-center MCC rooms are space-starved, so prefer a 1000 A lineup in 600 mm-deep sections rather than 800 mm, and verify the soft-starter wall has front-only access if the room has a 1.2 m rear-clearance minimum. Also, plan for the data logger gateway or process control node on the lineup, because every modern intelligent bucket expects an Ethernet port and a 24 VDC control bus. The EEP 1000 A design documents both voltage monitoring and branch metering on the same network, which is the right reference for a 2026 spec [S5].
Standards, sourcing, and what to require in the submittal
Lock the submittal to NEMA ICS 18 for the mechanical/electrical build, UL 845 (North America) or IEC 61439-6 (global) for the assembly, and IEEE 1584 for arc-flash incident-energy calculations [S2][S3]. Require the manufacturer to provide a coordination study, an as-built single-line, a thermal study of the soft-starter wall, and a network map of every intelligent bucket. The buyer's guide rule still applies: the supplier that cannot deliver that documentation pack on bid day is the supplier that will be expensive to live with for the next 15 years [S1][S6].
Verifiable signals to track over the next 6-12 months: whether major hyperscale operators publish a unified data-center MCC spec that fixes bus rating, soft-starter type, and metering protocol across regions, and whether IEC 61439-6 revisions tighten the partial-dissipation testing rules for arc-resistant sections. Watch also for MCC vendors to ship built-in PUE-aware soft-starter firmware that broadcasts kW and kVAR per branch on Ethernet/IP, replacing the separate power meter on the lineup [S4][S5].