Delta launches 800 VDC power and liquid‑cooling infrastructure for NVIDIA DSX AI factories

Delta Electronics on September 17, 2026, announced integrated energy-to-compute power and cooling infrastructure intended to speed deployment of AI factories based on NVIDIA DSX, pairing 800 VDC power delivery the company rates at up to 98% conversion efficiency with liquid-cooling systems scaling to 3 MW.

In its announcement from Fremont, California, the company described its portfolio as the physical infrastructure bridge between energy availability and compute productivity, connecting onsite energy, facility power, row- and rack-level power and cooling, and chip-level solutions. The release describes the NVIDIA DSX AI Factory Platform as bringing together reference designs, open software, accelerated computing systems, facilities infrastructure and partner technologies in a codesigned platform for AI factory design, deployment and operations.

“AI factory economics begin with a simple question: how much intelligence can be produced from every available megawatt?” said Austin Tseng, president of Delta Electronics Americas. Tseng said the answer depends on more than facility power alone, and that power quality, GPU load transients and heat must be engineered together all the way into the rack. Delta and NVIDIA connect those layers, he said, allowing AI factory operators to reach time-to-first-token sooner and optimize token output at scale.

Vladimir Troy, vice president of AI infrastructure at NVIDIA, said AI factories must be designed as complete systems, with power, cooling and compute working together. He said Delta’s expertise in 800 VDC power delivery and liquid cooling will help customers deploy AI factories based on NVIDIA DSX faster and get more AI performance from every megawatt.

Power Delivery From the Facility to the Chip

At the facility level, Delta integrates energy storage systems, solid-state transformers and medium-voltage infrastructure to store, convert and distribute the power required by high-density AI workloads. Across the IT space, the company combines 800 VDC power delivery with row-to-chip liquid cooling, aligning power flow and heat removal with the workload swings of high-density GPU computing. For AI factories based on NVIDIA DSX, Delta says this system-level approach helps close integration gaps between facility and IT infrastructure.

Delta says its 800 VDC power architecture reduces conversion stages between facility power and compute while responding rapidly to GPU load transients. Its high-density In-Row systems deliver up to 800 kW in a single power rack, with tailored battery and capacitance backup technologies supporting transient stability. At the board level, Delta’s 800 VDC DC-DC power distribution technology steps down directly to 50 VDC or 12 VDC at a peak efficiency the company puts at up to 98.5%, which it says reduces electrical losses, heat generation, copper requirements and stranded capacity while reclaiming white space for compute.

On NVIDIA’s 800 VDC architecture page, NVIDIA describes 800 VDC as its architecture for next-generation power distribution, one that decreases conversion and routing volumes in the compute space and significantly reduces current, copper use and cable bulk compared with rack-level 54 VDC systems and facility-level 480 VAC systems. NVIDIA lists Delta among the data center electrical ecosystem partners it is working with to accelerate adoption of 800 VDC.

Liquid Cooling and Prefabricated Delivery

Delta’s thermal management portfolio spans 2.4 MW and 3 MW liquid-to-liquid cooling distribution units, In-Rack cooling, and thermal solutions for next-generation AI racks. The company says that by coordinating electrical and thermal infrastructure at the facility, row, rack and component levels, its power and cooling systems can meet the density and operating requirements of AI factories based on NVIDIA DSX.

For deployment, Delta’s Prefabricated AI Modular Data Center Solution integrates 800 VDC In-Row Power and 3 MW of liquid-cooling capacity into infrastructure blocks assembled and tested in the factory. Delta says moving more integration and validation offsite reduces onsite complexity, streamlines commissioning and supports phased capacity expansion as demand grows. The company says this prefabricated delivery approach connects infrastructure design with execution, making constrained power more usable, accelerating the path to productive GPU capacity and supporting reliable token output at scale.

NVIDIA’s DSX Platform

According to NVIDIA’s DSX platform page, DSX unifies design, simulation, operations and ecosystem technologies to help build AI factories optimized for lowest token cost, spanning chips and systems, infrastructure software, facilities and partner technologies. Its component technologies include DSX Reference Design, which provides generation-specific validated architectures covering compute, networking, storage and facilities infrastructure; DSX Sim, simulation technologies and partner tools for identifying bottlenecks and validating cross-system tradeoffs; and DSX OS, open-source modular software for lifecycle management, health automation, resiliency and multi-tenant operations.

NVIDIA says DSX MaxLPS, its power and efficiency software, dynamically manages power at the GPU, rack and workload levels and can help customers provision up to 40% more GPUs within the same megawatt budget. DSX Flex bridges AI factories and the power grid, receiving grid signals such as load shedding, demand response and pricing events and adapting workloads in response, while DSX Exchange provides a common communication layer for compute, energy, power and cooling plant signals across IT, operational technology and operations systems. NVIDIA describes DSX as a common design standard serving cloud partners, sovereign clouds, land, power and shell providers, power and cooling equipment manufacturers, independent software vendors, OEMs, system integrators, and architectural, engineering and construction firms.