800 VDC safety in AI data centers: Understanding the real arc flash risk
- Tony Landry, Stuart Sheehan, Wendy Torell, Maria A. Torres Arango
- 20 Aug 2026
- 3 min
What you need to know
As AI data centers adopt 800 VDC to support higher rack power densities, concerns about arc flash risk are growing. Schneider Electric’s analysis of two representative 800 VDC architectures found that incident energy remained below the commonly referenced 1.2 cal/cm² threshold in both cases. The study shows that arc flash exposure is driven more by system architecture, energy storage placement, converter behavior, and protection design than by the 800 VDC voltage level itself.
As AI workloads push rack power densities higher, 800 VDC has emerged as the leading approach for powering next-generation AI data centers. The higher operating voltage naturally raises a question for facility and safety teams: does moving to 800 VDC also mean higher arc flash risk?
Our new white paper, “DC Arc Flash Analysis: A Practical Study on 800 VDC AI Data Centers,” examines this question directly. Using simulation-based case studies of two representative 800 VDC architectures, we modeled fault behavior and calculated arc flash incident energy to see how these systems actually perform – and what drives the result.
Standards still evolving
Arc flash standards and calculation methods are still catching up to 800 VDC. Most existing methods (NFPA 70E, IEEE 1584) were developed for AC systems and don’t fully capture the time-dependent, converter-limited fault behavior of 800 VDC architectures.
Conservative methods overestimate risk
Simplified DC arc flash methods (e.g., the NFPA 70E maximum power method) are a useful conservative starting point, but in many cases they overestimate incident energy compared to higher-fidelity, waveform-based analysis.
Low incident energy observed
In both case studies, calculated incident energy stayed below the 1.2 cal/cm² threshold commonly used to determine when arc-rated PPE is required.
Protection matters more than voltage
Centralized 800 VDC architectures showed higher incident energy than rack-level designs, but adding current-limiting converter behavior and standard protection devices brought incident energy back down to levels typical of well-protected data center equipment.
Architecture drives outcomes
System topology has a greater effect on arc flash exposure than the voltage level itself.
Capacitor placement is critical
Capacitor discharge can dominate the first milliseconds of a fault when capacitance is tied directly to the distribution bus. Capacitor placement and protection strategy are therefore key design levers, not incidental details.
As rack power densities increase, data center operators are rethinking how electricity moves through their facilities. That search for efficiency has elevated 800 VDC from an engineering concept to a serious industry direction.
Safety analysis methods were largely developed for conventional AC systems and do not fully reflect the behavior of modern converter-based 800 VDC architectures.
If we take the example of a traditional electrical fault like water flowing through a river channel. The flow is relatively predictable. A converter-based DC system behaves more like a reservoir with gates that can open, close, or restrict flow depending on conditions. The current can change rapidly, and those changes are influenced by the system's electronics and controls.
To understand how 800 VDC behaves in practice, the whitepaper examined two distinct approaches.
The first places power conversion equipment close to the IT rack through a sidecar architecture. The second centralizes power conversion and distributes power across a broader facility.
At first glance, both systems operate at the same voltage. One might assume their risks would be similar. The results showed otherwise.
The rack-level architecture produced remarkably low incident energy levels.
Researchers intentionally used conservative assumptions. They excluded protective devices from the calculations and assumed conditions designed to represent challenging scenarios.
Even under those assumptions, calculated incident energy remained far below the commonly referenced 1.2 cal/cm² benchmark.
Why?
Distributed energy sources limit the amount of stored energy available to feed a fault.
The whitepaper points to an important distinction: in this architecture, risk is driven primarily by short bursts of stored energy rather than large, sustained flows of current. That makes the fault behave very differently from what engineers might expect in traditional electrical systems.
The second case study examined a much larger centralized 800 VDC architecture serving multiple high-density racks from common power infrastructure.
This system includes:
- Centralized power conversion
- Battery energy storage
- Distribution switchboards
- Busway distribution systems
Naturally, a larger system creates more pathways through which energy can flow.
The analysis showed higher incident energy than the rack-level design. However, the increase was not a sign of uncontrolled risk. Instead, it highlighted the importance of thoughtful engineering.
One design feature stood out in particular: reverse current blocking.
This functionality prevents energy stored in one rack from feeding a fault elsewhere in the system. Imagine closing a series of doors during an emergency so a problem remains contained instead of spreading throughout a building. The electrical principle is similar.
The result is lower fault current, reduced arc flash exposure, and simpler protection coordination.
When these protective measures were incorporated, incident energy remained below key safety thresholds.
The two case studies operate at the same voltage but produce different fault behavior and different incident energy, because the outcome depends on design choices, not on the number “800” itself:
- Where power conversion occurs in the system
- Where energy storage (capacitors, BESS) is located relative to the distribution bus
- Whether current can flow backward into a fault, and how that’s controlled
- How converters respond to a fault (limit, ride through, or trip)
- How quickly protection devices clear the fault
These factors determine how much energy reaches an arc and for how long – which is what ultimately drives incident energy and PPE requirements, not the DC voltage on its own.
The white paper frames this as a preliminary, direction-setting analysis rather than a substitute for system-specific assessment. It recommends that organizations evaluating or deploying 800 VDC:
- Involve safety stakeholders at project inception, before architectural decisions are locked in.
- Treat arc flash risk as a design input converter topology, capacitor placement, and reverse-blocking functionality should be specified deliberately, not left to default component choices.
- Request fault-behavior data (output capacitance, current-limiting response, isolation timing) from equipment vendors as part of procurement.
- Use simplified methods for initial screening, then apply higher-accuracy simulation-based analysis covering multiple fault locations and conditions to validate protection schemes and PPE requirements.
- Document the analysis methodology clearly to support review by authorities having jurisdiction (AHJs), who often have limited prior exposure to 800 VDC systems.
Schneider Electric is also conducting controlled laboratory testing of rack-level and centralized 800 VDC architectures to validate these simulation results against measured data. Those findings will be published separately.
The debate around 800 VDC is often framed around voltage. This whitepaper suggests the real conversation should be about architecture. Read the whitepaper DC Arc Flash Analysis: A Practical Study on 800 VDC AI Data Centers to know more.
Latest in Data Center Research & Strategy
Exploring DCIM and EcoStruxure IT solutions
Generative AI inferencing ramp-up
The price of invisible downtime
800VDC: Powering the future of AI data centers
Need help?
Product Selector
Quickly and easily find the right products and accessories for your applications.
Get a Quote
Start your sales inquiry online and an expert will connect with you.
Where to buy?
Easily find the nearest Schneider Electric distributor in your location.
Help Center
Find support resources for all your needs, in one place.