Representative End User Clients
Representative Automation Clients
Representative Software Clients
The study uses transient simulation and digital twin modeling to provide practical guidance for safely deploying higher-voltage power infrastructure in high-density AI data centers.
Schneider Electric announced the release of a study examining how data center designers and operators can help to assess and manage arc flash risk in emerging 800 VDC power architectures. Based on deployment scenarios informed by hyperscaler design patterns, the study compares two representative architectures and evaluates how system configuration, capacitor placement, fault location, and fault-clearing behavior influence arc flash outcomes.
The research comes as data center operators explore 800 VDC distribution to help support the increasing power requirements of high-density artificial intelligence (AI) infrastructure. The transition is expected to help enable AI factories and large-scale data centers to support IT racks of 400 kW and above, but it also requires a more detailed understanding of electrical fault behavior, protection coordination, and safe work practices.
The analysis found that arc flash risk in 800 VDC systems can be managed and, under many conditions, can be comparable to the risk associated with typical AC distribution systems. It also found that conventional calculation methods may overestimate incident energy in capacitor-dominated DC systems because they do not fully account for transient fault behavior.
Assessing Two 800 VDC Architectures
Schneider Electric evaluated rack-level and facility-level 800 VDC architectures representing two major implementation approaches emerging across the data center industry. The analysis combines standards-based calculations, transient simulations, and system-level modeling to assess fault currents and incident energy over time.
The rack-level configuration uses a sidecar, also referred to as a power rack, located close to the IT equipment. Under the conservative methods and assumptions used in the study, incident energy remained below the referenced personal protective equipment threshold of 1.2 cal/cm², even without dedicated protection devices.
The centralized configuration distributes 800 VDC power from facility-level equipment. This architecture produced slightly higher incident energy under conservative assumptions that excluded overcurrent protection. The results varied according to fault location and the position of reverse-blocking diodes, which affected back-feed, peak current, and the duration of the event.
When fault contributions were limited using standard protection devices, the study found that incident energy declined to levels considered more appropriate for the work environment and generally comparable with common AC architectures.
Factors Affecting Arc Flash Risk
The analysis identified several factors that influence arc flash behavior in 800 VDC systems:
Transient Fault Currents: Capacitor discharge can dominate fault current during the first milliseconds of an arc flash event, making time-dependent analysis important.
System Architecture: Rack-level and centralized designs produce different fault-current paths and incident-energy outcomes.
Capacitor Placement: The location and capacity of stored electrical energy can influence peak current and the duration of an event.
Fault Location: Faults occurring upstream or downstream of reverse-blocking devices can produce different levels of back-feed and incident energy.
Protection Speed: Millisecond-scale detection and interruption can substantially reduce arc duration and incident energy.
Converter Behavior: The response of power conversion equipment and its control logic must be considered when assessing fault contribution.
These findings indicate that arc flash risk cannot be determined by operating voltage alone. The wider electrical architecture, protection design, system topology, and transient response must also be evaluated.
Using Simulation and Digital Twins
Arc flash analysis is established practice in AC data centers, but there is currently no industry-wide framework specifically addressing the hazards associated with converter-fed 800 VDC systems. Schneider Electric’s study examines how existing safety methodologies can be adapted when they are applied in an architecture-aware and time-dependent manner.
The analysis uses ETAP software and electrical digital twins to model system topology, switching behavior, protection coordination, converter response, capacitor discharge, and fault-clearing times. This provides a more detailed representation of system behavior than simplified calculations based on steady-state assumptions.
Physics-based simulation can help engineering teams evaluate alternative system designs, identify high-risk fault locations, coordinate protection devices, and test switching or maintenance scenarios before equipment is installed or energized. Digital twins can also preserve the engineering model throughout the infrastructure lifecycle, allowing it to be updated as system configurations and operating conditions change.
Supporting Higher-Density AI Infrastructure
The study reinforces the importance of integrating electrical safety analysis into the early design of high-density AI data centers. As rack power requirements move toward hundreds of kilowatts and eventually megawatt-scale configurations, power distribution, protection, cooling, and maintenance practices must be considered as parts of a connected infrastructure system.
Designers can help to reduce risk by selecting appropriate capacitor configurations, controlling reverse-current paths, coordinating protection devices, and using high-speed fault interruption. Accurate system models can further support these decisions by showing how the infrastructure is likely to behave during realistic fault scenarios rather than relying only on conservative assumptions.
Schneider Electric has also conducted testing of live-swap power capabilities for 800 VDC systems to support maintenance activities. The company’s findings are documented in the white paper, DC Arc Flash Analysis: A Practical Study on 800 VDC in Data Centers.
Related ARC Insights
The study reflects the growing need to design high-density AI data centers as integrated electrical, thermal, and digital systems, with safety analysis and simulation incorporated from the outset. Related ARC Advisory Group coverage includes:
Schneider Electric and AMD Develop Reference Design for High-Density AI Infrastructure
Schneider Electric and ETAP Introduce Physics-Based Digital Twin for Grid Operations
Follow ARC Advisory Group for the latest trends in AI data center infrastructure, electrical safety, digital twins, simulation, and energy management.