On-prem DC stack Build a DC Offers / Example Rival Snapshot
Example Rival Snapshot

Rival Snapshot — Datacenter on demand

Three ways to buy modular capacity plus managed infrastructure — not a UPS, rack, or cooling SKU bake-off. Armada when the module includes GPUs and a sovereign edge/AI control plane (Galleon through Leviathan, Bridge, Atlas). Vertiv when the job is a prefab hall (OneCore / MegaMod) with field services and Next Predict. Schneider Electric / APC when EcoStruxure modular pods or all-in-one modules should land on an EcoCare / EcoConsult lifecycle. Commercial terms are quote required. No ranking.

Products: Armada Beacon → Cruiser → Triton → Leviathan → Orion, plus Bridge / Atlas · Vertiv SmartMod → MegaMod / CoolChip / HDX → OneCore, plus Next Predict · Schneider all-in-one, pod, power module, hall, plus EcoCare Essential / Advanced / Plus

Disclaimer. Sample for illustration. Public sources only. Not a Gartner reprint. Speed, density, and savings figures are vendor claims, not Cloudbit benchmarks. Pricing is quote-required.

Exec summary

Score this category on what is inside the module, who operates it after handover, and how fast “site-ready” becomes capacity — not on UPS efficiency or cabinet load ratings (those live on the Power, Cooling, and Racks snapshots). Armada is the compute-included path: Galleon classes from suitcase Beacon through Leviathan (1.77 MW, vendor) and Orion (10 MW, 26 Aug 2026 release), with Bridge and Atlas as software, not enclosures. Vertiv is the critical-facilities prefab path: SmartMod at the room, MegaMod at 0.5–1 MW blocks, CoolChip/HDX liquid pods, OneCore at 5–50 MW, then named services (Next Predict) rather than a priced tier SKU. Schneider Electric / APC is the EcoStruxure path: all-in-one, IT pod, power module/skid, and data hall, with EcoCare plan classes (Essential, Advanced, Plus advanced) where a public code exists. No verified Magic Quadrant exists for this offer shape — do not invent one. Model tables: within-vendor ladders and Services.

Comparison

Armada Vertiv Schneider / APC
Who it’s for Remote edge, industrial, and sovereign/neocloud buyers who want GPUs in the module and software control Enterprise, colo, sovereign, and neocloud halls that bring their own IT and want one prefab + services vendor Sites standardizing on EcoStruxure/APC, from an edge closet to an AI pod, plus a plant lifecycle contract
Modular form Galleon: Beacon (no published kW), Cruiser (3 racks), Triton (5 racks), Leviathan (1.77 MW IT, up to 576 GB300, 2.2 MW cooling, 480 V). Orion (26 Aug 2026): 10 MW, up to 2,880 GPUs / 40 racks — not on the four-config Galleon page. Bridge and Atlas are software OneCore (5 Aug 2025): 5–50 MW, 96–944 racks. MegaMod EMEA brochure: 0.5 or 1 MW blocks, reference designs at 500 kW and 1000 kW. CoolChip brochure: reference designs 1.3 MW / 13 racks and 5.2 MW / 72 racks, family up to 10 MW. HDX PR (14 Jan 2026) says compact up to 1.25 MW — not the same figure as the 1.3 MW brochure row. SmartMod is the room below this (10 racks / 100 kW on its product page) Four forms on the hub: all-in-one, IT pod, power module, data hall. Easy EU PR: 5–14 racks, about 27–94 kW. IT Pod page: up to 40+ racks; Jun 2025 PR: pods up to 1 MW and beyond. One public skid code: PFMPS1250NAVX (1250 kW). Hall range page has no kW; a 2020 reference design is 500 kW
Deploy speed Galleon: days to weeks (vendor). Leviathan: 12 weeks from site-ready vs vendor’s 18–24 month stick-build contrast. Buyer still owns land, utility to spec, and permits Months, not “drop and run.” MegaMod blog: ~10 months average vs up to ~20 if land and permits exist. OneCore: factory parallelism; later Vertiv comments cited in trade press claim up to ~50% less site work — not audited here Less site assembly by design. EU Easy: delivery as little as 12 weeks (European standardized offer). Pod brochure: one-day install and up to 30% faster / 15% savings — brochure claims
Managed layer Armada Edge Platform (Atlas fleet view, Commander, AI Assist). Bridge orchestrates, meters, and can monetize GPUs on Armada gear or yours. Marketplace for hardware, connectivity, apps Commissioning, PM, spares, training. Next Predict: condition-based monitoring managed service (NOC, 24/7 engineers) on power and thermal — not a GPU cloud. Unify on OneCore. MegaMod base warranty stated as 12 months from install or 15 from delivery EcoCare classes on se.com: Essential ECRESS, Advanced ECRADV, Plus advanced ECRPADV. Fetched examples: power module 250–750 kW (ECRESSPFP1YSP-01, ECRADVPFP1YSP-01) and 1000–1500 kW (ECRPADVPFP3YSP-02, ECRPADVPFP5YSP-02); all-in-one 200–240 kW (ECRPADVPFI1YSP-04). Prices quote required. Entitlement differences between classes are not on those pages. EcoConsult has no plan code fetched. Vertiv Next Predict is the peer managed-monitoring name — no public tier code
AI vs hall Both: rugged edge and liquid-cooled AI factory. NVIDIA Quantum / Spectrum-X called out on Leviathan. Compute is the product Hall and high-density AI/HPC infrastructure. GPUs are customer scope unless a separate IT partner is named Edge module through AI pod (NVIDIA partnership on the modular hub). Not a GPU-as-a-service stack
Control / sovereignty Explicit: data on site, in-org, or in-country; Triton/Cruiser air-gap; Bridge tenancy; “no hyperscaler dependency.” Probe whether Atlas/AI Assist must phone home OneCore lists sovereign as a deployment type (you own the hall). Ask where Unify / Next Predict telemetry lives Operational control via EcoStruxure + EcoCare, not a residency story. Put air-gap in the RFP if you need it
Commercial Quote required Quote required Quote required (plan codes public; dollars are not list)

Key public markers: Leviathan page (12 weeks from site-ready, 1.77 MW) · OneCore PR 5 Aug 2025 · MegaMod 0.5 MW blocks · EcoStruxure modular hub + Jun 2025 pod brochure · EcoCare / EcoConsult product ranges. Do not treat vendor percentages as bid-level promises.

Within-vendor ladders

Same public figures as Services, compressed. Fetched 3 Oct 2026. Not a SKU price list. Conflicting vendor numbers are shown side by side.

Armada

ClassPublished envelopeLayer
BeaconSuitcase. No kW or rack count publishedEdge hardware
Cruiser20 ft, 3 racks. Air-gap option. No kWEdge hardware
Triton40 ft, 5 racks. Air-gap option. No kWEdge hardware
Leviathan1.77 MW IT; up to 576 GB300 or 512 HGX B300/B200; 2.2 MW cooling; 480 V / 60 Hz; no onsite water; 4-unit pod stated at 5.35 MW. Vendor: 12 weeks from site-readyAI system. Bridge sold separately unless the quote includes it
Orion26 Aug 2026 release: 10 MW/unit, up to 2,880 GPUs, 40 racks, six-module core, months from deposit. Not on the four-config Galleon pageAI system on Bridge
Bridge / Atlas / Commander / MarketplaceNo appliance SKUs. Bridge: turnkey on Galleon, or on your GPUs. Atlas is fleet insightsSoftware

Vertiv

LinePublished classWatch-out
SmartModProduct page: up to 10 racks and 100 kW or less. Max-page blurb also says 2–14 racksDo not average
SmartMod MaxUp to 200 kW. DX ≤26 racks, CW ≤24. UPS 100/150/200 kWRoom, not a hall
TimberModMass-timber variant of SmartMod and Power ModuleNo kW of its own
MegaMod / PlusEMEA brochure: 0.5 or 1 MW blocks, up to 2 MW or more. Reference designs 500 kW (96 or 120 racks) and 1000 kW (192 or 240 racks). EXL S1 UPSUS product page geo-blocked this pass. Blog: ~10 months average, not weeks
MegaMod CoolChipBrochure R11/25: up to 10 MW, 2–180 racks. Designs: 1.3 MW / 13 racks and 5.2 MW / 72 racksScalability cell in the table conflicts with the four-block design write-up
MegaMod HDXBrochure: 2–144 racks, same 1.3 / 5.2 MW designs. PR 14 Jan 2026: compact up to 1.25 MW / 13 racks; combo up to 10 MW / 144 racks; 50 to >100 kW/rack1.25 MW (PR) ≠ 1.3 MW (brochure)
OneCore5–50 MW, 96–944 racks, −20 to 55 °C, 11–35 kV and 400–480 V. Unify. 5 Aug 2025Slab-up hall. GPUs not included
Next Predict and ServicesNamed practices (project, digital, thermal, UPS, liquid cooling, and others). Next Predict = AI condition-based managed monitoringNo public tier SKU or price

Schneider / APC

Form or planPublished classCode
All-in-One EssentialFour standard designs. No kW on the range page—
Easy All-in-One (EU)5–14 racks, about 27–94 kW. As little as 12 weeks delivery (8 Jun 2022 PR)No SKU in the PR
IT PodUp to 40+ racks (range page). Pods up to 1 MW and beyond (11 Jun 2025 PR)—
Power moduleGalaxy VXL enclosure (UPS, switchgear, Li-ion, cooling). No MW list on the range pageBrochure not retrieved (HTTP 403)
Power skid1250 kW, 480 V / 60 Hz, Galaxy VXPFMPS1250NAVX
Data hallRange page: no kW. 16 Jun 2020 reference design: 500 kW, Tier III, 400 VDoc 500KW_DataHall_T3_400_RefDe_EN
EcoCare EssentialPower module 250–750 kW, 1 yrECRESSPFP1YSP-01
EcoCare AdvancedPower module 250–750 kW, 1 yrECRADVPFP1YSP-01
EcoCare Plus advancedPower module 1000–1500 kW, 3 yr and 5 yr; all-in-one 200–240 kW, 1 yrECRPADVPFP3YSP-02, ECRPADVPFP5YSP-02, ECRPADVPFI1YSP-04
EcoConsultAssessment and roadmap. No kWNo code fetched

EcoCare prefixes on these codes only: ECRESS Essential, ECRADV Advanced, ECRPADV Plus; PFP power module, PFI all-in-one. Not a generator for unlisted bands. Reseller prices were not copied.

When to shortlist

Lean When Skip when
Armada GPUs ship inside the module; edge or site-ready AI capacity is measured in weeks; sovereignty, air-gap, or Bridge GaaS is the point The RFP is a mixed 10–50 MW hall, BYO servers, and a global facilities maintainer — Armada is the wrong peer for that SOW
Vertiv Prefab power and thermal at hall scale, one vendor from design through commissioning, Next Predict or Vertiv Services on the plant You need the vendor to supply and resell GPU capacity, or the site is a handful of edge racks
Schneider / APC APC / EcoStruxure is already the standard; you want a pod, all-in-one, or power module plus EcoCare — or EcoConsult before a refresh You want a sovereign GPU cloud, or policy forbids a single-vendor electrical and mechanical stack

Buyer probes

  1. Armada: Which Galleon (Beacon, Cruiser, Triton, Leviathan, or Orion) and what contractual clock starts at site-ready — including permits you still own? Freeze IT kW, GPU model and count, liquid loop, UPS ride-through, and spares. Is Bridge in the price, and can Atlas / AI Assist stay dark on an air-gapped site?
  2. Armada: Who rolls a truck when the CDU, chiller, or UPS fails, in which countries, and what do you keep if you leave Bridge?
  3. Vertiv: OneCore hall, MegaMod / Plus blocks, or MegaMod CoolChip / HDX? Split the schedule into factory, ship, set, commission, and your IT install. Which published “up to 50%” schedule or TCO lines are actually warranted?
  4. Vertiv: Next Predict asset list, regions, and exclusions vs a classic PM contract. Where does telemetry reside? Base MegaMod warranty vs the services term on the quote.
  5. Schneider: All-in-One Essential / Easy, IT Pod, power module, or data hall — and which reference design? Separate factory-ship promises from utility and liquid-cooling readiness. Which region the “12 weeks” or “one-day” claim applies to.
  6. Schneider: EcoCare entitlement at your kW (switchgear, UPS, batteries, cooling, response time). What EcoConsult hands you as a document vs what EcoCare operates. Whether EcoStruxure IT (Data Center Expert / IT Expert / Advisor) is a separate license.

Flags: No MQ and no rank order. No list prices. Vertiv/Schneider do not include GPUs unless the quote says so. Armada “no grid” language is the small Galleon story — Leviathan still wants utility power. “12 weeks,” “one-day,” “50%,” and “25% life or TCO” stay vendor claims until the SOW repeats them.

Sources

  1. Armada ladders (3 Oct 2026): Galleon · Leviathan · Orion, 26 Aug 2026 · Bridge · Atlas. Also armada.ai · Products · Galleon · Leviathan · Bridge
  2. Vertiv: OneCore PR (5 Aug 2025) · MegaMod · MegaMod timeline blog · Next Predict · IT Brief (Vertiv time-to-token claims)
  3. Schneider ladders: IT Pod · PFMPS1250NAVX · ECRESSPFP1YSP-01 · ECRPADVPFP3YSP-02 · ECRPADVPFP5YSP-02 · ECRPADVPFI1YSP-04. Hub: EcoStruxure Modular DC · Pod brochure (Jun 2025) · Easy all-in-one EU PR · EcoCare · ECRADVPFP1YSP-01 · EcoConsult
  4. cloudbit.tv/services.html · component snapshots: Power, Racks, Cooling, DCIM