The company is now targeting 800 VDC architectures for AI-driven data centers, citing potential PUE improvement by eliminating intermediate AC-to-DC conversion stages in UPS and PDU chains
Decision Focus
Nexans, in collaboration with Schneider Electric and VINCI Energies, commissioned a direct-current microgrid pilot at its AmpaCity research center in Lyon, France. The system operates on a 700 VDC main bus, integrating office workstations, lighting, a battery energy storage system, and two 30 kW EV fast-charging stations. The signal for data center energy operators is not the installation itself—it is what Nexans is doing with the validated technology. The company is explicitly positioning 800 VDC distribution as the next architecture for facilities where rack densities exceed 100 kW, using this commercial real estate pilot as the engineering validation step before a heavier-duty application.
90-Second Brief
Now, nexans commissioned a 700 VDC DC microgrid pilot in Lyon, developed with Schneider Electric and VINCI Energies. The system serves as a live Current/OS-compliant testbed using Nexans’ DC Series cable range. The company is now targeting 800 VDC architectures for AI-driven data centers, citing potential PUE improvement by eliminating intermediate AC-to-DC conversion stages in UPS and PDU chains. A photovoltaic integration is planned for a second project phase.
What Is Really Happening?
The Lyon installation is a controlled technology demonstrator, not a data center deployment. What matters is the engineering logic it is validating. Today’s AC-distributed data centers perform multiple conversion steps: utility AC to rectified DC inside the UPS, back to AC for distribution, then to DC again at the server. Each step carries conversion losses. At moderate rack densities those losses are manageable, but at densities above 100 kW per rack—which AI compute clusters increasingly require—cumulative losses become a meaningful fraction of total site power draw.
The 800 VDC architecture addresses this by distributing at high-voltage DC directly from the grid interface point to the rack, eliminating intermediate conversion stages. Nexans claims this reduces copper requirements, lightens cabling infrastructure, and limits distribution losses. These are directionally sound engineering principles, though independent large-scale validation in data center environments has not yet been publicly established.
What is structurally new at AmpaCity is the pairing of 800 VDC distribution with High-Temperature Superconducting cable technology. HTS cables operate with near-zero electrical resistance, which theoretically removes resistive losses at the cable level itself. The voltage architecture addresses conversion losses; HTS addresses distribution losses. Running both in the same demonstrator is a deliberate signal about where Nexans believes high-density power infrastructure is heading and where it intends to hold market position.
The Current/OS compliance angle adds a further strategic dimension. Current/OS is a developing interoperability framework for DC building infrastructure. If 800 VDC data center deployments eventually standardize around a comparable framework—as has occurred in other power electronics transitions—cable vendors who have already built compliant product lines carry an early specification advantage. Nexans is positioning for that lock-in before the standard is set.
Why It Matters for Global Heads of Data Center Energy
The immediate relevance is not deployment readiness—it is procurement timeline awareness. DC power architecture decisions carry infrastructure-level consequences extending well beyond cable selection. UPS topology, transformer configuration, PDU design, and switchgear protection coordination all change when transitioning from AC distribution to high-voltage DC. Facilities committed today on AC assumptions may face significant retrofit exposure if 800 VDC becomes the dominant architecture for AI compute within a five-to-seven-year window.
For operators currently in design or pre-construction phases for GPU-dense facilities, the DC versus AC question at the facility level is becoming less academic. The efficiency argument is directional: fewer conversion stages should yield lower PUE, which translates directly into energy spend and Scope 2 emissions accounting. The magnitude of the gain depends on conversion efficiency assumptions that have not yet been established at data center scale from independent sources, so treating vendor PUE improvement claims as a firm planning number would be premature.
The battery storage integration in the AmpaCity pilot also points toward a relevant operating model. A DC-native microgrid with integrated BESS and on-site generation creates more direct coupling between storage and load. Fewer conversion steps means storage capacity is utilized more efficiently during peak demand events or grid stress periods—a consideration that connects directly to demand response participation strategy and peak capacity cost management at the portfolio level.
Forward View
Three fronts are worth tracking if 800 VDC gains traction in the sector. First, whether hyperscaler facility specifications begin treating DC distribution topology as a design requirement rather than an option—Google and Microsoft infrastructure standards carry market-making weight, and a specification shift there would accelerate the supply chain and reduce procurement risk for everyone following. Second, whether Current/OS or an equivalent DC interoperability standard achieves formal adoption through IEC or IEEE processes, which would compress vendor alignment timelines and create clearer procurement criteria. Third, whether HTS cable technology reaches commercial data center deployment at a cost point that justifies the cryogenic cooling overhead it introduces. The combined 800 VDC plus HTS architecture is the high-upside scenario; the realistic near-term question is whether 800 VDC alone clears the business case threshold.
What Is Still Uncertain
The Lyon pilot is a research-center installation. The scale gap matters: a 700 VDC microgrid serving workstations and EV chargers operates in a fundamentally different fault-management and load-variability environment than a 100 MW hyperscale campus. Efficiency claims in the source material—reduced copper use, lower distribution losses, improved PUE—have not been independently verified at data center load levels in any publicly available study as of this writing.
The regulatory and safety framework for 800 VDC in data center environments is also not yet mature. High-voltage DC introduces arc-fault and protection coordination challenges that differ meaningfully from AC systems. Standards bodies are still developing guidance, which creates engineering uncertainty and potential liability exposure for early movers operating outside a clear compliance envelope.
HTS integration carries additional unknowns. Cryogenic cooling systems introduce their own energy overhead, and the net efficiency gain after accounting for that load has not been publicly quantified in a data center application. Until commercial deployments produce auditable performance data, the combined architecture remains a demonstrator-stage proposition.
One Question for Your Team
Does your current design specification for high-density AI compute lock you into AC distribution assumptions that would require full infrastructure replacement if 800 VDC becomes the sector standard within this decade—and has that retrofit cost been modeled as a line item in your long-range capital plan?
Sources
- Nexans — Direct Current (DC) Microgrid: AmpaCity – Nexans (Link)
