The capex premium for 800V DC architecture is not concentrated in any one layer. It accumulates across all six of the power stack. Cooling infrastructure (Layer 5) carries the largest single-layer premium at $90M for 100 MW greenfield. Layer 2 (perimeter conversion) second at $45M. Total transi
800 VDC becomes the default architecture for new AI infrastructure builds above 100 MW by 2028-2029. Installed-base composition reaches 60 percent DC by 2035. 1500 VDC under active discussion for 500 MW+ facilities. Five specific risks + five specific predictions with named catalysts.
Commissioning at 800 VDC runs 40-60 percent longer than AC baseline. Skills gap requires 80-120 hours retraining per electrician at $8-15K. Pipeline capacity (Vertiv Academy, Schneider EU, Eaton EU, IBEW apprenticeship, university programmes) falls short of estimated 2028 demand by roughly 5,00
48V incumbents face 20-40 percent revenue-at-risk. Vertiv/Eaton/Schneider win through content-dollar expansion 2-3x per rack. WBG semi content grows 3-4x. New entrants (Menlo Micro SSCB, Skeleton HSC, Submer immersion) emerge in specialty niches. Layer-specific screening beats sector-level.
OCP fastest (Mt. Diablo published May 2025, 80+ ecosystem partners). IEC leads on protection (IEC 60947-10 for SSCBs published early 2026). NEC and IEEE trail 24-36 months. Full convergence 2028-2029. Interim projects carry 8-15 percent multi-standard compliance premium.
800 VDC facilities present different power-quality profiles to the utility grid than legacy AC. Rectifier topology choice at Stage 1 determines whether the facility passes the interconnection study or bounces. ERCOT and PJM regulatory posture is tightening.
Above 100 kW per rack, 800 VDC and direct-to-chip liquid cooling are not independent decisions. Air cooling caps at 40-60 kW per rack. Vendor consolidation (Vertiv/PurgeRite $1B, Eaton/Boyd $9.5B, Schneider/Motivair $1.7B) happened because customers want both from one vendor.
Retrofitting an AC facility to 800 VDC is technically possible but rarely economic. Greenfield beats retrofit on 10-year TCO by 30-50 percent under typical assumptions. Retrofit only makes sense in three narrow cases; most operators should scope a satellite greenfield instead.
Native 800 VDC UPS collapses 4 conversion stages to 2 and connects the battery directly to the DC bus. End-to-end efficiency climbs 4-6 points. BMS complexity increases sharply. LFP is displacing NMC in data centre UPS applications faster than in the EV market.
Grounding topology at 800 VDC is a Stage-1 architecture decision that interacts with arc-fault protection and insurance. TN-S vs IT-System has different scaling properties at hyperscale. Insulation-resistance monitoring runs into engineering limits above 100 MW connected load.
DC arcs cannot self-extinguish at zero crossings. 800 VDC arc-flash risk is manageable with the right protection design; Schneider Electric August 2026 study shows incident energy comparable to AC when architecture, capacitor placement, and fault-clearing behaviour are specified correctly.
What Changed Issue 01. A one-page monthly recap on AI infrastructure power. What NVIDIA, OCP, hyperscalers, and power equipment vendors did in July 2026, what the market is getting wrong, and what to watch over the next six months.