Can the MEC-24-NA fit a full-depth 2U GPU server?

The 1200 mm cabinet depth is designed around equipment up to a 950 mm device-depth envelope, including a reference 946.7 mm full-depth 2U server. The exact vendor rail kit, rear cable bend and exhaust clearance must still be validated.

How much power can the Edge AI cabinet support?

The base North American single-phase planning limit is 5.76 kW continuous from a 208/240 V, 30 A input. A 208Y/120 V, 30 A three-phase growth configuration has an 8.6 kW continuous planning limit.

Is this cabinet intended for unrestricted 8-GPU training servers?

No. The MEC-24-NA is intended for compact edge nodes, full-depth 2U GPU servers and limited 4U deployments. Sustained loads above 8.6 kW or liquid-cooled training platforms should move to a larger data-center cabinet and dedicated facility engineering.

Does the cabinet include cooling?

The platform uses front-to-rear airflow with perforated intake, an open server air path and rear exhaust. Replaceable fan modules can be selected for the payload, but room HVAC remains a deployment prerequisite.

Home / Solutions / Edge AI Power Cabinet

North America SME platform · MEC-24-NA

Edge AI Power Cabinet

Name: MEC-24-NA Edge AI Power Cabinet
Brand: ECO

Compact edge AI without data-center complexity. A freestanding, serviceable 24U enclosure for on-premises inference, industrial edge compute and local AI services - designed for both short-depth systems and full-depth 2U GPU servers.

Review my server list Check equipment fit

MEC-24-NA 24U Edge AI power cabinet with server equipment and lower power compartment — Concept rendering. Final rack rails, PDU, thermal kit and certification scope are confirmed for the selected payload.

24U19-inch EIA rail envelope

600 × 1200 mmExternal width × depth

350-950 mmDevice-depth envelope

5.76 kWBase continuous planning load

Design decision

Use one deep, serviceable enclosure rather than a shallow cabinet.

A 1200 mm chassis depth leaves a workable envelope for a short-depth 440 mm edge server and for a full-depth 2U GPU server approaching 947 mm, while preserving rear cable-bend and exhaust space.

The 24U layout can be configured as 20U of compute plus a 4U lower power zone, with 0U vertical PDU channels, adjustable front/rear rails and front-to-rear airflow.

This is an SME edge platform, not a replacement for a high-density data-center row. Final node count is governed by measured input power, heat rejection, rail compatibility, weight distribution and local electrical code.

Universal rack interface

19-inch EIA-310 four-post rails with square holes and 450-1000 mm rail adjustment.

Deep equipment envelope

Supports 350-950 mm device depth with planned service and cable-clearance zones.

Separated power zone

A lower 4U space can house input distribution, bypass or an optional LFP UPS module.

Front-to-rear thermal path

Perforated intake, unobstructed server airflow and a defined rear exhaust zone.

Equipment compatibility

Built around the rack, not one server vendor.

Board-level AI devices mount on ventilated shelves, while rack servers use approved vendor slide rails. Every final configuration still needs a mechanical, thermal and electrical review.

Device class	Reference size	MEC-24-NA fit	Integration rule
Jetson / IGX / industrial AI module	Board or compact enclosure; not a native rack device	Yes	Use a lockable ventilated 1U/2U shelf, retention bracket and cable strain relief.
Short-depth edge server	Reference: 445 W × 87 H × 440 D mm, 2U	Yes, substantial depth reserve	Use four-post rails and a 0U vertical PDU while maintaining front-to-rear airflow.
Full-depth 2U GPU server	Reference: 482 W × 86.8 H × 946.7 D mm	Yes, maximum planned depth class	Validate the exact rail kit; leave rear cable-bend and exhaust clearance.
4U multi-GPU server	Up to 4U and ≤950 mm device depth	Physical fit, limited deployment	Plan no more than 1-2 nodes and select the upgraded electrical and thermal kit.

Reference dimensions are used only to define a practical compatibility envelope. They do not replace vendor rail testing, thermal validation or final production drawings.

Power architecture

Scale the electrical kit with the real AI payload.

A staged design keeps the entry model practical while providing a defined path for higher-power edge deployments.

Configuration	Input and continuous planning budget	Recommended payload
Base - North America single phase	208/240 V, 30 A locking input; 5.76 kW continuous planning limit	One or two compact GPU servers or several lower-power edge nodes; 0U metered PDU; optional external UPS.
Growth - three phase	208Y/120 V, 30 A three-phase input; 8.6 kW continuous planning limit	Two or three full-depth 2U GPU nodes plus network and local storage where room HVAC is verified.
Beyond this platform	More than 8.6 kW sustained, liquid-cooled platforms or unrestricted 8-GPU training servers	Move to a 42U, 750 mm wide data-center cabinet with dedicated cooling and facility engineering.

Freestanding 24U edge computing cabinet with lockable casters and perforated side panels — 24U freestanding format with a separated lower power zone and serviceable access.

Protective bonding

Plan PE bonding points, lockable disconnect access and cable strain relief.

Environmental monitoring

Temperature sensors, high-temperature alarms and a monitored controller.

Listed power components

Use listed PDUs, breakers, inlet assemblies and UPS/battery modules.

Room HVAC prerequisite

Cabinet airflow cannot substitute for correctly sized room heat rejection.

Where it fits

Local AI infrastructure for small and medium businesses.

The cabinet is intended for secure on-premises inference and industrial edge workloads where a full data-center deployment is unnecessary.

Office

Private local inference

House compact GPU servers, switching and storage for controlled on-premises AI services.

Retail

Back-room edge compute

Support video analytics, local data processing and site applications in a serviceable footprint.

Healthcare

Clinic and imaging workflows

Deploy local compute where data handling and response time favor on-site processing.

Industry

Factory-cell AI

Integrate inference servers, network equipment and monitored power near the process.

Design boundaries

Freeze the payload before releasing the cabinet.

Prototype quotation should follow an equipment list, rail-kit review, input-power budget and room thermal check.

Can it fit a full-depth 2U GPU server?

The 1200 mm cabinet depth is designed around an equipment-depth envelope up to 950 mm. The exact slide rail, bezel, rear cable bend and exhaust clearance still need validation.

How much continuous power is available?

The base 208/240 V, 30 A design uses a 5.76 kW continuous planning limit. The 208Y/120 V, 30 A three-phase growth option uses an 8.6 kW planning limit.

Is it suitable for unrestricted 8-GPU training?

No. High-density, liquid-cooled or sustained loads above 8.6 kW should move to a larger data-center cabinet with dedicated facility engineering.

What must be confirmed before production?

Freeze the server list, vendor rail kits, PDU and inlet, branch circuit, UPS choice, total heat load, room HVAC, acoustic requirement, management interface, certification route and local-code review.

Request a payload review

Send the exact server models, not only the U count.

We will check depth, rails, power, airflow, cable service space and the lower power-zone arrangement before quotation.

WhatsApp Kevin Email server list