Data centers run on one non-negotiable principle: power never stops. Never. A single second of downtime costs money, damages reputation, and in healthcare or financial services, can have consequences far beyond the balance sheet. Choosing the right diesel generator for your data center backup is one of the most consequential infrastructure decisions you’ll make — and getting the sizing wrong is a mistake that compounds every time the grid fails.
This guide walks you through the complete sizing process for data center backup generators, specifically addressing the considerations that matter when sourcing from Chinese manufacturers.
Understanding Data Center Power Architecture
Before calculating generator size, you need to understand how power flows through a data center. The chain looks like this:
Utility Grid→
ATS→
UPS→
PDU→
Servers
When the grid fails, the Automatic Transfer Switch (ATS) detects the outage and signals the generator to start. The UPS (Uninterruptible Power Supply) bridges the gap — typically 10-30 seconds while the generator ramps up. The generator then supplies power to the UPS, which continues providing clean power to the servers.
Your generator needs to be sized for the total critical IT load plus cooling, plus lighting, plus security systems, plus UPS charging current, plus a safety margin. Let’s break this down step by step.
The Step-by-Step Sizing Process
Step 1: Calculate Your IT Load
IT load is the power consumed by all computing equipment: servers, storage arrays, network switches, and any other devices that process or store data. There are two ways to calculate this:
- Nameplate method: Add up the nameplate ratings of all IT equipment. This overestimates by 20-40% since equipment rarely draws its full nameplate rating.
- Measured method (preferred): Use power monitoring at the PDU level to measure actual power draw. This gives you real numbers.
For a typical 500-rack data center, the IT load might look like:
Servers (500 racks × 5 kW avg): 2,500 kW
Storage arrays: 300 kW
Network switches & spine: 150 kW
Management & monitoring: 50 kW
─────────────────────────────────────────
Total IT Load: 3,000 kW
Step 2: Add Cooling Load
Cooling systems (CRAC/CRAH units, chillers, cooling towers) consume roughly the same amount of power as the IT load. The PUE (Power Usage Effectiveness) ratio tells you this:
PUE LevelDescriptionCooling Load as % of IT
1.0Theoretical perfection (impossible)0%
1.2World-class efficient20%
1.5Very efficient50%
1.8Average80%
2.0Below average100%
2.5+Inefficient150%+
For our example with PUE 1.5:
Total IT Load: 3,000 kW
Cooling Load (50%): 1,500 kW
─────────────────────────────────
Subtotal: 4,500 kW
Step 3: Add Ancillary Loads
Don’t forget the supporting systems:
Lighting & security: 50 kW
Fire suppression: 30 kW
UPS charging current: 300 kW (≈10% of IT load)
Elevators & common areas: 70 kW
─────────────────────────────────
Ancillary total: 450 kW
Step 4: Apply Redundancy Factor
Data centers use redundancy tiers defined by the Uptime Institute:
Tier N — Basic (No Redundancy)
Single generator covers the full load. If it fails, everything goes down. Suitable only for non-critical workloads.
Generator sizing: 100% of total load
Tier N+1 — Redundant Components
One extra generator provides redundancy. If one generator fails, the others handle the full load. Most common for enterprise data centers.
Generator sizing: (Total Load × 1.0) per unit × (N+1) units
Example: 4,950 kW total load ÷ 4 generators = 1,238 kW each. With N+1, you need five 1,250 kW generators.
Tier 2N — Fully Redundant Systems
Two complete, independent power paths. Each path can carry the full load. Highest reliability for enterprise and colocation facilities.
Generator sizing: 100% of total load × 2 separate systems
Tier 2N+1 — Maximum Redundancy
Used by hyperscale operators (Google, AWS, Azure). Multiple redundant systems with an extra capacity unit.
Generator sizing: Full redundancy plus extra capacity
Step 5: Apply Safety Margin
Add 15-25% headroom for future growth and load uncertainty:
IT Load: 3,000 kW
Cooling (PUE 1.5): 1,500 kW
Ancillary: 450 kW
─────────────────────────────────
Total connected load: 4,950 kW
Safety margin (20%): 990 kW
─────────────────────────────────
Total generator capacity: 5,940 kW
For N+1 with 1,500 kW generators:
6 × 1,500 kW = 9,000 kW (can lose one and still have 7,500 kW)
UPS-Generator Compatibility: A Critical Consideration
This is where data center generator sizing gets tricky. Most modern UPS systems use active power factor correction (PFC) with input power factors of 0.95-0.99. However, the UPS’s charging current and the type of UPS topology affect generator sizing significantly:
UPS TypeGenerator Sizing ImpactRecommended Derating
Double-conversion (online)Large charging currents and harmonic distortion stress the generatorSize generator at 1.3-1.5× UPS input rating
Line-interactiveModerate impact; less charging currentSize at 1.2-1.3× UPS input rating
Delta-conversionMinimal harmonic impact; most generator-friendlySize at 1.1-1.2× UPS input rating
Eco-mode (bypass during generator)Lowest generator impactSize at 1.0-1.1× load
Tesla Power engineers work directly with your UPS vendor to ensure generator-UPS compatibility. We’ve sized generators for data centers using Eaton, Schneider, Vertiv, and Huawei UPS systems — we know the quirks of each.
Fueling System Design for Data Centers
Data centers need extended runtime because outages can last hours or even days. Here’s how to plan your fuel system:
- Minimum runtime requirement: Most data centers design for 24-72 hours of autonomous operation
- Base tank — built into the generator, typically 8-12 hours of runtime
- Day tank — gravity-fed buffer tank for reliable fuel delivery
- Bulk storage tank — above-ground or underground tank for 24-72 hour reserve
- Automatic fuel transfer — pump system that transfers fuel from bulk storage to day tank
- Fuel polishing system — essential for fuel stored long-term; removes water and particulates
Key requirement: Your fuel system must be able to replenish faster than the generator consumes fuel. If your generators burn 1,200 liters/hour total, your transfer pumps need capacity of at least 2,000 liters/hour to maintain the base tank level while accounting for pump cycling and maintenance.
What to Specify When Ordering from China
Data center generators have specific requirements that differ from general-purpose units. When ordering from a Chinese diesel generator manufacturer, include these specifications:
- Ultra-low noise — target 60-65 dB(A) at 7 meters for rooftop or campus installations
- Remote monitoring — Modbus/TCPIP communication with your BMS (Building Management System)
- Parallel operation capability — critical for N+1 and 2N redundancy architectures
- Fast start time — 10-15 seconds to rated voltage and frequency (data centers can’t wait)
- Grade 1 fuel compatibility — specify EN590 or ASTM D975 diesel fuel standards
- Load acceptance testing — step-load test at 0-50-100% to verify the generator handles sudden load application
- Emissions compliance — Tier 4 Final or local equivalent (data centers often face strict environmental regulations)
Tesla Power Data Center Generator Solutions
Tesla Power has purpose-built generator solutions for data center applications:
ParameterData Center Specification
Power Range500 kW – 3,000 kW per unit
EngineCummins QSK series (Tier 4 Final) or equivalent
AlternatorStamford P-series (brushless, Class H insulation)
Control SystemComAp InteliGen NT with parallel operation and BMS integration
Start Time<10 seconds to rated voltage/frequency
Noise Level60-65 dB(A) at 7 meters (ultra-silent canopy)
Parallel OperationUp to 32 units synchronized (isochronous load sharing)
CommunicationModbus RTU/TCP, SNMP, Profibus, OPC-UA
Fuel SystemMulti-tank with auto-transfer and polishing ready
CertificationsISO 9001, ISO 14001, ISO 8528, CE, TUV, SGS
Our Data Center Service Package
- Dedicated data center applications engineer
- Free sizing analysis and redundancy architecture review
- UPS compatibility testing and coordination
- Factory witness testing with client participation
- Commissioning support with step-load testing
- Annual maintenance contracts with guaranteed SLA response times
- Remote monitoring and diagnostics
- 24/7 emergency support with 4-hour response commitment
Factors That Affect Data Center Generator Sizing
- Climate zone — hot climates require more cooling power, directly increasing generator load
- Altitude — derate 3% per 300m above sea level
- Future growth plans — design for 20-30% headroom above current projected load
- Free cooling capability — economizer modes reduce cooling load, potentially allowing smaller generators
- Liquid cooling vs. air cooling — affects cooling system power requirements
- High-density computing — GPU clusters and AI workloads draw 2-5× traditional server power
- Regulatory requirements — some jurisdictions mandate specific backup power durations
Industry Keywords for Your Research
- Data center diesel generator — backup power for data center facilities
- Diesel generator sizing — calculating required generator capacity
- Backup power data center — power continuity for critical IT infrastructure
- Chinese diesel generator — sourcing generators from Chinese manufacturers
- Data center power backup — comprehensive backup power solutions
- Diesel generator redundancy — N+1 and 2N generator configurations
- Data center PUE calculation — power usage effectiveness and generator sizing
Frequently Asked Questions
Q1: How many generators does a data center need?
It depends on your redundancy tier. A Tier III (N+1) data center typically needs enough generators so that losing any one unit still leaves sufficient capacity. For example, a 5,000 kW load might use six 1,000 kW generators (N+1 = any five can carry the full load). A Tier IV (2N) facility would need two completely separate generator systems, each capable of handling the full load. Tesla Power can design the optimal configuration based on your specific requirements.
Q2: How fast must a data center generator start?
The UPS bridges the gap during generator start-up, so the critical metric is that the generator reaches rated voltage and frequency before the UPS battery depletes. Most data centers design for a 10-15 second start time. The UPS battery typically provides 15-30 minutes of runtime, so even a slow-starting generator (30 seconds) has plenty of margin. However, Tesla Power data center generators are configured for <10 second start times to provide maximum safety margin.
Q3: What fuel storage do data centers require?
Most enterprise data centers design for 24-72 hours of fuel storage. Colocation facilities often mandate 48 hours minimum per carrier requirements. Hyperscale data centers (Google, AWS) may store 72+ hours. Calculate based on total generator consumption at 75% load. For example, six 1,000 kW generators at 75% load burn approximately 900 liters/hour — so 72 hours requires about 65,000 liters of stored diesel.
Q4: Can I use natural gas generators instead of diesel for my data center?
Natural gas generators have advantages (cleaner emissions, no fuel storage) but significant drawbacks for data centers: slower start time (30-60 seconds), lower power density (larger units needed for same output), and dependence on gas pipeline infrastructure (which can also fail during emergencies). Most data centers use diesel for these reasons, sometimes with natural gas as a secondary option. Tesla Power specializes in diesel generators but can advise on dual-fuel configurations if natural gas backup is also desired.
Q5: How often should data center generators be tested?
At minimum: weekly no-load start tests (5-10 minutes) and monthly load bank tests at 75%+ rated load (2-4 hours). Annual comprehensive testing including step-load testing and full-system failover simulation. Many Tier IV data centers perform quarterly failover tests where they actually disconnect from the utility grid to verify the complete backup chain works end-to-end. Tesla Power provides load bank testing services and can help design your testing protocol.
Sizing a diesel generator for your data center is a high-stakes calculation that rewards precision and punishes guesswork. Every kW matters — oversized generators waste fuel and capital, while undersized units fail precisely when you need them most.
Take the time to measure your actual loads, plan for growth, and design your redundancy architecture carefully. And when you’re ready to specify, source, and commission your backup power system, Tesla Power has the engineering expertise and manufacturing capability to deliver generators that meet the exacting standards data centers demand. Let’s talk about your requirements.