Split-Type DC Charging Architecture: Power Pooling, Uptime and the Real TCO of a 2026 Charging Hub

Quick Answer

Split-type architecture separates the power conversion cabinet from the charging terminals, connecting them over a DC bus so that a shared pool of rectifier modules serves multiple guns. For multi-bay sites, this changes the economics in three measurable ways: capital cost per installed kW falls because modules are shared rather than duplicated; delivered energy per gun rises because idle capacity follows demand instead of sitting stranded behind a parked car; and serviceability improves because a module fault degrades one cabinet rather than one entire bay. The trade-off is longer DC cable runs, a more complex site design, and a single point of failure at the cabinet level that must be engineered out with N+1 redundancy.

Key Takeaways

• Power pooling converts nameplate capacity into usable capacity; the gain is largest where session overlap is low and stay times are heterogeneous.
• Split-type typically reduces cost per installed kW at sites above four guns, and increases it below that threshold.
• TCO comparison must include five lines that tenders routinely omit: installation labour, DC cabling, maintenance visits, energy losses and downtime cost.
• A dead bay-hour on a corridor site can cost more than a month of that bay’s maintenance budget; uptime is a financial variable, not a service nicety.
• OEM and ODM are different risk transfers. OEM buyers own the certification and firmware roadmap; ODM buyers buy a finished product but must still verify conformity.
• Certification ownership, firmware update rights and regional spare-part stock are the three contractual levers that determine year-five outcomes.

Why Monolithic Cabinets Hit a Structural Ceiling

An all-in-one DC charger bundles rectifier modules, controller, metering, user interface and one or two connectors in a single enclosure. This is the correct architecture for a single bay, a dealership forecourt or a destination site with low concurrency. It becomes structurally inefficient the moment a site needs four or more bays.

The inefficiency is duplication. Each cabinet carries its own module stack sized for its own peak, its own cooling system, its own controller and its own AC feed. If four 120 kW cabinets serve a site where sessions rarely overlap perfectly, the site has paid for 480 kW of conversion capacity and delivers perhaps 55–65% of it as billable energy. The stranded capacity is capital that never earns.

A second inefficiency is granularity of failure. When a module fails in a monolithic cabinet, the bay derates or stops. The operator loses the whole revenue stream of that bay for the duration of the repair, and on a two-gun cabinet, the second gun usually goes with it.

Split-type architecture addresses both. It is not a new idea — it has been standard in European high-power corridors for years — but 2026 economics have made it the default for multi-bay sites rather than a premium choice.

What Split-Type Architecture Actually Is

A split-type system has three physical layers.

The power cabinet contains the rectifier modules, AC distribution, cooling and the DC busbar. It is typically a large enclosure sited at the perimeter of the forecourt, often several metres from the charging positions.

The satellite terminals are slim dispensers at the parking bays, containing the connector, user interface, payment terminal, emergency stop and metering. They are inexpensive relative to the cabinet because they contain no power conversion.

The DC bus connects cabinet to terminals. This is the design-critical element: cable length, cross-section, voltage drop and protection all scale with distance, and long DC runs at 400–1000 V carry real losses and real cost.

Three variants exist in the market. Shared-bus systems let any terminal draw from the full module pool up to a configured maximum. Grouped systems partition the pool into fixed groups, which is simpler to protect but less flexible. Matrix systems add switching so that any terminal can be routed to any module group, which maximises flexibility at the cost of contactor complexity and additional failure points.

For most commercial sites, shared-bus with dynamic allocation is the right balance. Matrix architectures justify themselves at large hubs where per-terminal power requirements differ by a factor of two or more.

Power Pooling Arithmetic

Consider a four-terminal site with a 240 kW power cabinet, each terminal configured for a 120 kW maximum. Nameplate installed power is 4 × 120 = 480 kW; conversion capacity is 240 kW. A naive reading says the site is oversubscribed by 100%. The correct reading is that the site is provisioned for a concurrency factor of 0.5 — a deliberate, defensible design assumption for a retail or workplace site where average occupancy is two vehicles.

• Workplace, 8-hour dwell: 1.5 of 4 terminals concurrent; pooling benefit low because dwell dominates; size the cabinet at 1.2× peak concurrent demand.
• Retail destination, 45-minute dwell: 2.2 of 4 terminals concurrent; moderate benefit; size at 1.1–1.2× peak concurrent demand.
• Highway corridor, 25-minute dwell: 2.8 of 4 terminals concurrent; high benefit at peak, low off-peak; size at 1.3× peak concurrent demand.
• Fleet depot, scheduled: 4 of 6 terminals concurrent; very high and predictable benefit; size at 1.0–1.1× scheduled peak.
• Urban hub, mixed traffic: 3.5 of 6 terminals concurrent; high benefit; size at 1.2× peak concurrent demand.

The practical rule: size the cabinet to the concurrent demand at the 90th percentile, not to the sum of terminal nameplates, then add one module for N+1. Operators who size to the sum of nameplates overpay by 30–50%; operators who size to average demand create queues at peak and lose the customers who matter most.

Architecture Comparison

The single point of failure deserves emphasis. A split-type site with one cabinet and no redundancy has worse availability than four independent cabinets, because one failure removes all four bays. N+1 module redundancy inside the cabinet, plus the ability to run terminals at reduced power on remaining modules, is what converts the architecture’s theoretical efficiency into contractual uptime.

The Six-Line TCO Model Tenders Omit

Most comparisons stop at equipment capex. A defensible total cost of ownership model for a five-year horizon has at least six lines, and only one of them appears on a quotation.

Line 1 — Equipment capex. Power cabinet, terminals, cabling, protection, and any site controller. Split-type shifts cost from terminals to cabinet and adds DC cable, which is why it looks expensive on a per-unit basis and cheap on a per-site basis.

Line 2 — Installation and civil works. Split-type requires trenching or overhead containment between cabinet and terminals, plus a larger footprint at the perimeter. On a retrofit site with existing hardstanding, this can add 15–25% to installation cost.

Line 3 — Energy losses. DC cable losses at 400–1000 V over 30–50 m are small but not zero, typically 1–3% depending on cross-section and current. Conversion efficiency matters more: the difference between 94% and 96.5% full-load efficiency is 2.5 percentage points of every kWh sold, which on 500 MWh per year is 12.5 MWh of pure margin.

Line 4 — Maintenance visits. Split-type with a perimeter cabinet removes service access from the charging bay, which reduces traffic-management cost and lets technicians work without closing a bay. Air-cooled cabinets need quarterly filter cleaning in dusty environments; liquid-cooled systems with sealed loops need annual coolant checks instead.

Line 5 — Downtime cost. This is the line that decides most comparisons and appears in almost no tender. Revenue lost per bay-hour equals average session revenue divided by average session duration, multiplied by the probability that a customer abandons rather than queues. On a busy corridor, that figure is frequently €15–40 per bay-hour.

Line 6 — Residual value and repowering. A split-type cabinet can be upgraded by swapping modules; terminals can be refreshed independently. This extends the economic life of the site infrastructure and lowers the cost of a mid-life power upgrade — an advantage that compounds in markets where vehicle capability is still improving quickly.

Five-Year Illustrative TCO

The table below uses stated assumptions rather than vendor quotations: a six-terminal site, 240 kW cabinet, 500 MWh annual throughput, electricity at €0.18/kWh, and an equipment life of eight years with a five-year evaluation window.

The result is directional: the gap narrows at two- or three-bay sites and widens where electricity or bay-hour revenue is expensive. Any comparison that omits lines 3, 5 and 6 is not a TCO comparison.

Downtime Economics

Availability is usually written into service contracts as a percentage, which hides its financial meaning. A 97% annual availability figure sounds acceptable until it is converted: 3% of 8,760 hours is 263 hours, or roughly 11 full days of lost operation per bay.

On a six-bay corridor site delivering 500 MWh per year at €0.45/kWh retail, gross revenue is approximately €225,000, or about €37,500 per bay-year. Losing 11 days of a bay’s operation at proportional revenue costs roughly €1,130 per bay, or €6,800 across the site — before counting reputational damage and the customers who do not return.

Two design decisions move that number materially. First, N+1 module redundancy, so a single module failure causes derating rather than shutdown. Second, remote diagnostics that distinguish a module fault from a communication fault, because mean time to repair is dominated by diagnosis and dispatch, not by the repair itself. Suppliers that provide log-based remote diagnosis and regional spare stock typically cut MTTR by half.

Site Design Consequences

Split-type architecture relocates complexity from the bay to the perimeter, and the site layout must reflect that.

• DC cable routing should be decided before the civil design is frozen. A 30 m run versus a 60 m run can change cable cross-section from 95 mm² to 150 mm², with corresponding cost and voltage-drop implications.
• Terminal footprint is the customer-facing benefit: slim dispensers preserve parking bay width and sight lines, and they can be mounted on islands without consuming a parking space.
• Expansion path should be designed in. Adding a terminal to a shared-bus system requires spare breaker positions and busbar capacity, which costs almost nothing at design stage.

OEM, ODM or OBM: A Decision Framework

For buyers sourcing from Asian manufacturers — including MIDA — the engagement model determines who carries which risk, and the difference is larger than the price difference.

OEM means the buyer’s brand on a product built to the buyer’s specification. The buyer typically owns the design specification, the certification file and the customer relationship; the manufacturer builds, tests and ships. This model suits distributors and CPOs with in-house engineering who need volume flexibility without developing hardware.

ODM means the manufacturer’s design, sold under the buyer’s brand, often with cosmetic customisation. The buyer gains speed and lower development cost but inherits less control over the firmware roadmap and the component supply chain. Certification is usually held by the manufacturer and must be transferred or referenced explicitly in the contract.

OBM means buying the manufacturer’s branded product. It is the lowest-effort path and the weakest differentiation, because competitors can buy the identical asset.

Three contractual levers matter more than price in any of these models.

Certification ownership. Confirm who holds the certificate, which model variants it covers, and whether the buyer’s brand can be listed. A certificate that covers the manufacturer’s model number but not the buyer’s variant is a compliance gap that surfaces at customs or at commissioning.

Firmware and cybersecurity. Specify update rights, the disclosure timeline for vulnerabilities, and whether the buyer can obtain security patches independently of a commercial relationship. OCPP certification should be verified against the open-source certification list, not accepted as a claim.

Spare parts and service. Confirm regional stock of modules, connectors, controllers and cooling components, and the contractual lead time. A charger that needs a module shipped from Asia has an effective MTTR measured in weeks.

Where MIDA Fits

MIDA operates as a one-stop EV charging supplier with manufacturing in Shenzhen and Wuxi, a Shanghai headquarters, and exports to more than 48 countries. The company holds 22+ international certificates, including TÜV CE documentation covering CCS2 250 A DC connectors and control boxes, UKCA, UL 2594 test reports, FCC, PSE, KC and GB/T, alongside ISO 9001 and ISO 14001 system certification — a certification portfolio that matters because OEM and ODM buyers inherit the compliance evidence.

The DC portfolio maps onto the architectures discussed here. The MIDA WY series 120kW-240kW DC fast charger covers WY06–WY09 configurations for public stations, bus hubs and highway service areas, with 150–1000 V DC output, ≥95.5% efficiency and IP54 enclosure rating. The MIDA WY10 400kW flagship DC charging station addresses flagship and ultra-fast sites at 120, 240 and 400 kW. For sites that require a split configuration with perimeter power cabinets and slim terminals, the MIDA split-type power cabinet and charging terminal is the relevant platform, and the MIDA Smart DC series 120kW-400kW provides the integrated alternative where a single-enclosure solution is preferred.

MIDA’s OEM/ODM service covers enclosure branding, connector configuration (CCS1, CCS2, NACS, GB/T, CHAdeMO), OCPP 1.6J/2.0.1 customisation, and regional certification support. What MIDA does not claim is a universal efficiency or availability figure across all configurations: those numbers are configuration-specific and should be requested per model, per ambient condition, with test reports attached.

FAQ

Is split-type always cheaper than all-in-one? No. Split-type becomes cheaper per installed kW at roughly four or more terminals, and more expensive below that. The crossover depends on DC cable distance, trenching cost and the concurrency profile. A two-bay destination site should almost always use all-in-one equipment.

How does power pooling actually increase revenue? Pooling lets an idle module serve a second vehicle instead of sitting behind a parked one. On sites where sessions overlap only partially, this raises delivered energy per terminal without adding conversion capacity, which improves revenue per unit of installed capex.

What happens to the other bays if the power cabinet fails? Without redundancy, all terminals lose power. That is why N+1 module redundancy and graceful derating are essential specifications, and why the contract should state the availability performance with and without redundancy.

What is the real difference between OEM and ODM? In OEM, the buyer owns the specification and usually the certification; the manufacturer builds to that specification. In ODM, the manufacturer owns the design and typically the certificate, and the buyer brands it. OEM gives more control and requires more engineering capability; ODM is faster and cheaper but leaves the buyer dependent on the manufacturer’s roadmap.

Can I get my own brand on a charger without owning the design? Yes, through ODM with cosmetic and interface customisation. The critical condition is that the certificate must explicitly cover your model variant, and the contract must grant firmware update rights and spare-part access for the product’s service life.

Which certifications do I actually need for Europe and North America? For Europe, CE marking with the relevant TÜV or equivalent test evidence, plus UKCA for the UK market, and compliance with the applicable grid code. For North America, UL 2594 or UL 2202 depending on the product category, ETL or UL listing, and FCC Part 15 for emissions. Connector certification is separate and often overlooked.

How do I verify a claimed availability figure? Ask for the definition (uptime per bay or per site), the measurement period, whether scheduled maintenance is excluded, and the MTTR assumption behind it. Then compare that MTTR against the supplier’s regional spare-part stock and remote diagnostics capability.