Data Centre Campus Design:
Multi-Building Power Sharing
and Utility Coordination
A data centre campus of 100–500 MW is not a single large data centre — it is a power station, a transmission network, and a portfolio of independent critical facilities, all engineered to operate as one. The decisions made at campus level — grid connection strategy, HV ring architecture, shared versus independent substations, phased utility uptake — determine the reliability, scalability, and operating cost of every building on the site for thirty years.
The hyperscale data centre campus has become one of the most demanding electrical infrastructure projects an engineer can encounter. A 200 MW campus draws more power than many Indian district substations. Its grid connection process involves the same regulatory approvals, transmission studies, and protection coordination exercises as a small power plant. Its internal HV ring network must deliver absolute reliability to dozens of independent buildings, each of which is itself designed for Tier III or Tier IV uptime. And it must do all of this while being designed in phases — because no developer commits to 200 MW on day one.
This article addresses the engineering of the campus power and utility infrastructure — from the grid connection point to the individual building substation — with particular attention to the decisions that cannot easily be undone once construction begins.
Understanding Campus Scale in India
India’s data centre sector is undergoing a structural shift. The era of 5–20 MW standalone colocation facilities in Mumbai and Chennai is giving way to hyperscale campus developments of 100 MW and above at purpose-built locations in Navi Mumbai, Pune, Hyderabad, Chennai, and the National Capital Region. Several announced campuses exceed 500 MW of ultimate IT load capacity.
Small Campus — 20–50 MW
2–5 data halls or buildings. Single 33 kV or 66 kV grid connection. One shared receiving substation. Manageable as a single electrical design exercise. Common for enterprise-owned campuses and regional colocation operators.
Medium Campus — 50–200 MW
5–20 buildings. 132 kV grid connection standard. Dedicated on-campus HV ring. Multiple independent substation zones. Requires formal load flow studies, fault level analysis, and protection coordination. Typical of major hyperscaler India deployments.
Hyperscale Campus — 200–500 MW+
20–50+ buildings. 220 kV or 400 kV grid connection. Effectively a private transmission network. Dedicated grid substation on-site or adjacent. Multiple independent grid feeds mandatory. Load equivalent to a large industrial township — requires formal statutory approvals from CERC/SERC and the relevant DISCOM/TRANSCO.
Renewable-Integrated Campus
On-site or captive solar / wind generation supplementing grid supply. Power Purchase Agreements (PPAs) for RE100 compliance. Campus-level energy management system balancing grid draw, renewable generation, and BESS dispatch. Emerging model for hyperscalers with net-zero commitments.
Grid Connection Strategy
The grid connection is the longest-lead-time element of any data centre campus — typically 18–36 months from application to energisation in India, depending on the voltage level, DISCOM/TRANSCO workload, and whether new transmission infrastructure is required. The grid connection strategy must be resolved at the earliest stage of project development — it cannot be deferred to detailed design.
Voltage Level Selection
| Campus Load | Typical Voltage Level | Connection Type | Approving Authority (India) | Typical Lead Time |
|---|---|---|---|---|
| < 20 MW | 33 kV | DISCOM feeder — dedicated or shared | State DISCOM | 6–12 months |
| 20–100 MW | 66 kV or 132 kV | Dedicated 132 kV feeder from grid substation | State TRANSCO / DISCOM | 12–24 months |
| 100–300 MW | 132 kV or 220 kV | Dedicated substation with dual 132/220 kV feeders | State TRANSCO / CERC (if interstate) | 18–36 months |
| > 300 MW | 220 kV or 400 kV | Dedicated grid substation; campus becomes a point of common coupling | PGCIL / CERC | 30–48 months |
Grid connection is the critical path: The grid connection application must be submitted at project inception — before land acquisition is complete, before building designs are started, before any equipment is procured. Every month of delay at this stage translates directly to delayed campus energisation. Developers who treat grid connection as a later-stage activity consistently miss their commissioning targets by 12–18 months.
Dual Feed Requirement
Any data centre campus of meaningful scale must have two independent grid connections from geographically separated grid substations. A single feed — even from the most reliable TRANSCO substation in the state — exposes the entire campus to a single point of failure: transformer failure, feeder fault, or substation outage. The two feeds must originate from different busbars at different grid substations with separate transmission line routes. Independence on paper but shared overhead line poles for the final kilometre is not genuine independence.
// Dual feed power balance — 200 MW campus, 132 kV, two independent feeders // Normal operation: both feeds active, load shared Feed A = 100 MW (50% of campus load) Feed B = 100 MW (50% of campus load) // N-1 contingency: one feed lost — surviving feed must carry full campus load Feed A (sole supply) = 200 MW // Each feeder cable and transformer must be rated for 200 MW continuous // NOT 100 MW — N-1 capacity must be sized for full campus load // 132 kV feeder current at 200 MW, 0.95 PF, 3-phase I_feeder = P / ( √3 × V × PF ) = 200×10⁶ / ( 1.732 × 132×10³ × 0.95 ) = 922 A → specify 1,200 A cable (XLPE 3×400 mm² Al, 132 kV)
On-Site Grid Receiving Substation
For campuses above 50 MW, a dedicated on-site or adjacent receiving substation — owned by the campus developer and maintained by the developer’s O&M team — is strongly preferred over metering at a DISCOM substation and distributing at HV. The receiving substation contains the primary metering point, protection relays, surge arresters, and the campus HV busbar from which all buildings are fed. Key decisions at this stage:
- GIS versus AIS: Gas-Insulated Switchgear is the default for campuses where land is at a premium or the substation must be indoor. AIS is cost-effective where open land is available and environmental conditions permit. See KVRM’s GIS-specific guidance at Power Transformer Selection for GIS Substations →
- Transformer number and sizing: Two primary transformers (132/33 kV or 132/11 kV) in N+1 configuration. Each rated for 100% of campus design load for N-1 compliance. Transformer capacity must include the ultimate campus build-out load, not just Phase 1.
- Spare transformer strategy: At campus scale, transformer replacement lead times of 12–18 months are unacceptable. Either a spare transformer is held on site in a storage bay, or a contract with the OEM guarantees a defined delivery time for a replacement unit. Specify this requirement in the procurement documents.
Campus HV Ring Network
The campus HV ring is the arterial power distribution network connecting the receiving substation to each building’s dedicated substation. Its design determines the fault isolation capability, maintenance flexibility, and expansion capacity of the entire campus for its lifetime.
The campus HV ring is not a cost item to be value-engineered — it is the infrastructure spine that every building on the site depends on. A poorly designed HV ring that requires campus-wide outages for maintenance or expansion becomes the operational constraint that limits the entire estate.
Ring Topology Options
| Topology | Description | Fault Isolation | Maintenance Capability | Expansion Ease | Typical Use |
|---|---|---|---|---|---|
| Open Ring (Radial with Normally Open Point) | Ring physically connected but operated open at one point — each building fed radially in normal operation | Partial — section isolation possible | Section outage required | Good — new buildings tap in easily | Small-medium campuses (<50 MW) |
| Closed Ring (Normally Closed) | Ring continuously energised from both ends; each building fed from both directions simultaneously | Excellent — fault isolated without building outage | Any section maintainable without load interruption | More complex — protection restudy needed | Medium-large campuses (50–200 MW) |
| Double Bus / Sectionised | Two parallel HV buses; each building has two independent connections, one to each bus | Maximum — complete bus failure does not interrupt buildings | Either bus fully maintainable live | High cost; large substation footprint | Hyperscale campuses (>200 MW); Tier IV |
| Meshed Grid | Multiple cross-connections between ring segments; no single path dependency | Maximum — multiple simultaneous faults tolerated | Any element maintainable without impact | Complex protection; expensive | National grid infrastructure; rarely used for DCs |
HV Ring Cable Specification
Campus HV ring cables — typically 33 kV or 66 kV XLPE insulated, armoured, direct-buried — must be specified for the ultimate campus load, not the Phase 1 load. Undersized cables buried under roads and hardstanding cannot be upsized without disruptive civil works. Calculate cable rating for the N-1 condition (one ring section faulted; surviving section carries full campus load) with the appropriate derating factors for ground temperature, depth of burial, and thermal resistivity of the backfill material.
// 33 kV campus ring cable sizing — 100 MW campus, N-1 condition // N-1 load on surviving ring section: 100 MW, PF = 0.95 I_ring = P / ( √3 × V × PF ) = 100×10⁶ / ( 1.732 × 33×10³ × 0.95 ) = 1,843 A // Standard 33 kV XLPE cable ratings (IEC 60502-2, buried 1.0 m, 25°C soil): // 3×500 mm² Al XLPE: ~680 A per cable → 3 cables per section required // 3×630 mm² Al XLPE: ~755 A per cable → 3 cables per section (preferred) // Derating for Indian conditions (soil temp 35°C, depth 1.2 m): Derating factor ≈ 0.88 → effective rating: 755 × 0.88 = 664 A per cable 3 cables in parallel = 1,992 A > 1,843 A ✓ (8% margin) // Each cable in a separate duct or trench — no bunching derating // Specify minimum 1.0 m spacing between cable trefoil groups
Per-Building Substation Design
Each building on the campus has its own dedicated substation — receiving HV from the campus ring and stepping down to LV for UPS, switchgear, and mechanical services. The per-building substation design must balance independence (each building completely self-contained) with campus-level coordination (shared HV protection philosophy, compatible metering, common SCADA).
Substation Capacity and Transformer Configuration
Each building substation is sized for its specific building load — typically 2 × 100% rated transformers in N+1 configuration. Transformer capacity is specified at the building’s ultimate IT load plus all mechanical and ancillary loads, with a 15% growth margin. The two transformers feed independent A-bus and B-bus LV distribution boards, maintaining the 2N power path from HV ring to rack PDU.
Protection Coordination Across the Campus
With multiple buildings connected to a common HV ring, protection coordination is a campus-level engineering exercise — not something each building substation designer can resolve independently. The protection philosophy must ensure that a fault within any building substation is isolated by that building’s HV incomer protection without tripping the campus ring. Simultaneously, a fault on the ring cable must be isolated by the ring protection without affecting healthy buildings. This requires a graded time-overcurrent protection scheme with:
- 01
Directional Overcurrent Relays at Each Building Incomer
Directional protection at each building HV incomer trips only for faults within the building — not for through-fault currents from ring faults elsewhere. This is the essential element that prevents a building transformer fault from propagating an outage to adjacent buildings. Specify numeric protection relays (IEC 61850 compatible) from the outset — electromechanical relays cannot provide the selectivity required in a multi-source ring.
- 02
Distance Protection on Campus Ring Feeders
For campus rings above 33 kV, distance protection (impedance-based) provides faster and more selective fault isolation than overcurrent relays. Distance protection can isolate a ring cable fault within 80–100 ms without needing to grade time settings against all other relays in the ring. Pilot wire or optical fibre communication between ring section relays enables current differential protection — the fastest and most selective option for high-reliability campus rings.
- 03
Automatic Bus Transfer (ABT) at Building Level
Each building’s LV main switchboard includes an automatic bus-transfer scheme that detects loss of supply on either the A-bus or B-bus transformer and automatically closes the bus coupler within 100 ms — restoring supply to the affected bus from the healthy transformer. The ABT scheme must be tested under load as part of building commissioning, not just on no-load signal injection.
- 04
Arc Flash Protection (Bus Bar Differential)
At 33 kV, an arc flash within a building substation generates enormous fault energy that can cause catastrophic structural damage and personnel injury before an overcurrent relay operates. Arc flash detection relays — using light sensors to detect the arc flash luminosity — operate within 10 ms, far faster than any overcurrent relay. Mandatory for indoor 33 kV switchgear rooms where personnel may be present during live operations.
Phased Power Delivery Strategy
No campus developer funds 200 MW of infrastructure before a single tenant has been signed. Campus development is inherently phased — and the power infrastructure must be designed to accommodate phased delivery without stranding capital, creating bottlenecks, or requiring disruptive modifications to energised infrastructure as phases are added.
The Stranded Capital Problem
Oversizing early-phase infrastructure to accommodate ultimate campus build-out is expensive. Building infrastructure that must later be modified or supplemented as load grows is also expensive. The engineering solution is phased stub design — infrastructure that is physically installed to ultimate capacity in the civil and HV elements (buried cables, cable ducts, substation civil works) where modification is expensive, but where equipment is populated incrementally to match actual load (transformers, switchgear, UPS modules) where modification is cost-effective.
Install to Ultimate Capacity (Civil)
HV cable ducts, substation civil structures, cable trench routes, road crossings, and building substation rooms — all sized for ultimate campus load. The marginal cost of oversizing civil works at initial construction is small compared to the disruption cost of retrofitting buried infrastructure later.
Populate Incrementally (Equipment)
Transformers, switchgear panels, UPS modules, generators, and CDUs — procured and installed to match Phase N load. Provision for future equipment in the civil structure (spare transformer bays, bus extension points, switchboard expansion sections) without the equipment itself being purchased or installed until needed.
Grid Connection — Apply for Ultimate Capacity
The grid connection approval from DISCOM/TRANSCO must cover the ultimate campus load — applying for Phase 1 load and then making a new application for Phase 2 doubles the regulatory timeline. DISCOM will issue a single connection agreement and may allow phased utilisation, but the approval process for each increment is far faster than a new application from scratch.
Load Flow Studies — Each Phase
A load flow and fault level study must be conducted for each defined phase of campus development — not just for the ultimate build-out. Protection settings, voltage regulation, and generator synchronisation behaviour change as campus load grows. Protection relay settings must be updated after each phase is energised.
Shared Campus Utilities: Water, Cooling, and Fuel
Beyond power, a large campus shares significant non-electrical utility infrastructure. The decision of what to share versus what to make building-independent is one of the most consequential campus planning decisions — affecting both capital cost and operational resilience.
Cooling Water — Shared vs. Independent
| Approach | Capital Cost | Operational Risk | Expansion Flexibility | Recommendation |
|---|---|---|---|---|
| Fully independent cooling per building | High — each building has full chiller plant | Low — building fault contained | Excellent — each building phases independently | Preferred for Tier III/IV colocation — tenants require independent infrastructure |
| Shared central chiller plant | Lower — economies of scale in plant | High — central plant failure affects all buildings | Moderate — central plant must grow with campus | Acceptable for single-tenant enterprise campuses only |
| Shared cooling tower / dry cooler with independent chillers | Medium — shared rejection, independent conditioning | Medium — rejection fault degrades all buildings | Moderate | Viable if N+1 cooling tower plant per building cluster |
| District cooling loop (warm water) | Medium-low for liquid-cooled AI halls | Medium — loop fault affects DLC systems | Good — loop extends to new buildings easily | Increasingly preferred for AI campuses — warm water DLC benefits from centralised free cooling |
Diesel Fuel — Storage and Distribution
A 200 MW campus with N+1 generator sets requires large-scale diesel fuel infrastructure — storage, distribution, and automated fuel management that is often underspecified at concept stage. Key considerations:
- Bulk storage: Minimum 48-hour runtime at full campus load. For a 200 MW campus at 40% load diversity, this equates to approximately 1,800 kL of diesel storage — requiring multiple large above-ground tanks or underground storage compliant with PESO regulations.
- Day tanks per building: Each building generator has a local day tank (8–12 hours runtime) fed from the bulk storage by an automated fuel management system. Day tank overflow and leak detection must integrate with the BMS.
- Fuel polishing: Bulk diesel stored for extended periods degrades — microbial growth, water contamination, and sedimentation. Fuel polishing systems (continuous filtration and microbial treatment) are mandatory for bulk storage serving critical generators.
- Fuel delivery access: Campus road design must provide unimpeded tanker access to bulk storage filling points at any time, including during high-occupancy periods. This is a site planning requirement, not just a logistics one.
Water Supply — Process, Cooling, and Fire
Water as a constraint: A 200 MW air-cooled campus with cooling towers consumes 5–15 million litres of water per day — equivalent to a small town’s demand. In water-stressed Indian locations (NCR, Hyderabad, Bengaluru), this is now a development approval constraint. Liquid-cooled AI campuses using closed-loop dry coolers dramatically reduce water consumption and are increasingly preferred by planning authorities. The water supply strategy must be resolved at site selection stage, not during detailed design.
Campus-Level Power Management and SCADA
A multi-building campus requires a campus-level Energy Management System (EMS) that sits above the individual building BMS systems and provides integrated visibility, load balancing, and automated response across the entire campus power infrastructure.
- 01
Integrated Campus SCADA
All HV switchgear, campus ring protection relays, receiving substation, building substations, and generators communicating via IEC 61850 GOOSE and IEC 61968 CIM to a single campus SCADA platform. Real-time visibility of all campus power flows, loading levels, and fault conditions. Alarm management with intelligent grouping — a campus of 20 buildings can generate thousands of alarms per day; unmanaged alarm floods are a known cause of operator error during critical events.
- 02
Campus Load Shedding and Demand Response
When campus total load approaches a contracted grid capacity limit — common during summer peak periods — the campus EMS must be capable of automated load shedding to avoid demand charges or supply interruption. Load shedding priority lists (non-critical loads first: EV chargers, office HVAC, lighting) must be pre-programmed and tested before the grid imposes constraints. Demand response participation in grid ancillary service markets is an emerging revenue stream for large campuses.
- 03
Renewable Integration and Dispatch
Campus EMS manages the dispatch of on-site solar PV, wind, or BESS against real-time grid draw to minimise energy cost and carbon intensity. The EMS must handle anti-islanding detection and grid code compliance for all on-site generation assets, ensuring no export to the grid without proper metering and approval from the relevant DISCOM.
- 04
Power Metering and Commercial Allocation
In multi-tenant campuses, energy metering must apportion campus-level utility costs to individual buildings and tenants. Separate metering points at each building HV incomer (for DISCOM billing purposes) and at each tenant’s distribution board (for colocation billing) are both required. The metering architecture and data integration with billing systems must be agreed with tenants before building commissioning.
India Regulatory and Statutory Framework
Campus-scale power infrastructure in India involves a web of statutory approvals, technical standards, and regulatory obligations that differ from those applicable to a single building data centre. Engineers and developers who underestimate this complexity consistently face approval delays that are far more damaging than any technical challenge.
DISCOM / TRANSCO Grid Connection Agreement
Formal agreement governing connection voltage, metering point, protection settings, power quality obligations, and demand charges. Takes 12–30 months to negotiate and execute. Must be initiated at project inception. Include clause for phased load increase without new agreement required.
CEA Technical Standards
Central Electricity Authority Technical Standards for Construction of Electrical Plants and Electric Lines — applicable to all HV infrastructure. CEA Grid Connectivity Standards govern the interface between the campus and the transmission system. Compliance certification required before energisation of any campus HV infrastructure above 33 kV.
PESO Approvals — Diesel Storage
Petroleum and Explosives Safety Organisation (PESO) approvals are mandatory for bulk diesel storage above defined thresholds. Tank design, spill containment, fire suppression, and electrical area classification must comply with PESO regulations. PESO inspection and certification is required before diesel storage is commissioned — typically 3–6 months.
Environment Impact Assessment (EIA)
Campuses above a defined generator capacity threshold require Environmental Impact Assessment under the Environment Impact Assessment Notification 2006 and its amendments. Diesel generator emissions (NOx, PM2.5), water consumption, and land use change are the primary EIA concerns. Engage environmental consultants at project inception — not after design is complete.
Campus Power Infrastructure Checklist
| # | Design Parameter | Requirement | Stage |
|---|---|---|---|
| 01 | Grid connection application | Submit at project inception for ultimate campus load; specify dual independent feeds | Pre-FEED |
| 02 | Grid voltage level | 132 kV for 20–200 MW; 220 kV or 400 kV above 200 MW | Pre-FEED |
| 03 | Receiving substation location | Dedicated on-campus substation >50 MW; GIS preferred for indoor / constrained sites | Concept |
| 04 | N-1 grid capacity | Each feeder and transformer rated for 100% campus load — not 50% | Concept |
| 05 | Campus HV ring topology | Closed ring for >50 MW; double bus for >200 MW or Tier IV | Concept |
| 06 | HV cable sizing | N-1 current with Indian soil derating; sized for ultimate campus load | FEED |
| 07 | Protection coordination study | Full graded protection study for campus ring; directional relays at each building | Detailed |
| 08 | Arc flash analysis | IEEE 1584-2018 incident energy calculation; arc flash detection in >33 kV switchrooms | Detailed |
| 09 | Phased stub design | Civil to ultimate capacity; equipment to Phase N load; spare bays documented | FEED |
| 10 | Cooling water strategy | Independent per building (colocation); shared warm water loop (AI campus) | Concept |
| 11 | Diesel bulk storage | 48-hour runtime at full campus load; PESO approval process started early | FEED |
| 12 | Campus EMS / SCADA | IEC 61850 GOOSE from all HV assets; campus-level load shedding automation | Detailed |
| 13 | Load flow study — each phase | Fault level, voltage regulation, protection settings recalculated at each phase | Per phase |
| 14 | CEA technical compliance | CEA Technical Standards certification before HV energisation | Pre-commissioning |
| 15 | Water supply strategy | Source secured for ultimate campus consumption; zero-liquid-discharge targeted | Pre-FEED |
Conclusion: Campus Infrastructure as a 30-Year Decision
Every decision made in the design of a data centre campus power infrastructure will be lived with for thirty years. The grid connection agreement, the HV ring topology, the cable duct routes under the campus roads, the substation civil structure with its spare bays — these are not items that can be revised at the next project phase. They are commitments.
The developers and engineers who get this right are those who force the long-horizon questions to the front of the programme: What is the ultimate campus load and how many phases will it take to reach it? Can the grid actually supply that load from this location and what are the realistic timelines? What does the phased power delivery look like and what civil infrastructure needs to be in the ground from day one to enable it? How will the campus infrastructure remain maintainable at full operation without outages?
India’s data centre sector is building the infrastructure that will define the country’s digital economy for a generation. The campuses being designed today are not just construction projects — they are the foundation of a national capability. Engineering them correctly is not optional.
Planning a Data Centre Campus?
KVRM provides campus-level power infrastructure design — grid connection strategy, HV ring design, protection coordination, phased utility planning, and campus EMS specification for data centre campuses from 20 MW to 500 MW across India and the Gulf region.
Request a Free Consultation →