Europe’s AI Factory Buildout: NVIDIA DSX AI Factory Platform Water, Power, Grid and Data | AIdataNewsGlobal General
This AIdataNewsGlobal General briefing condenses NVIDIA’s message about Europe’s AI infrastructure partners: they are adopting the NVIDIA DSX AI Factory Platform to cut cooling water, reduce electricity for equivalent AI production, speed capacity bring-up, and make workloads adaptable to grid reliability.
Platform logic: from chip-room architecture to daily operation
NVIDIA describes the platform as a full-stack NVIDIA DSX blueprint co-created with global infrastructure suppliers. The design intent is to treat accelerator compute, networking, storage, software, energy delivery and heat rejection as one system, with optimization managed from first architecture choices to live operation.
Europe’s AI cloud operators are applying DSX with suppliers in design, construction, energy and cooling to serve enterprises, developers and governments.
Inside the NVIDIA DSX platform, the named layers have distinct jobs:
- DSX MaxLPS for dynamic power allocation.
- DSX Flex for aligning workload power with grid conditions.
- DSX Reference Design for validated infrastructure architectures.
- DSX Sim for simulation.
- DSX OS for infrastructure management.
- DSX Exchange for links between computing and facility systems.
Operators can use these pieces to design, deploy and run AI factories while saving valuable resources.
Heat rejection path
Older tower-based cooling depends on evaporation and continuous replacement water. NVIDIA DSX reference designs aim for dry heat rejection: liquid cools equipment in a sealed loop, while outdoor dry coolers push heat into air without losing water.
NVIDIA Vera Rubin infrastructure supports 100% liquid cooling with coolant temperatures of up to 45 degrees Celsius. Where dry coolers are used, facility cooling water consumption can approach zero. NVIDIA cites a 2-gigawatt-scale result: 54 million liters of water a day saved against conventional tower systems, matching daily use of 180,000 households. Hot-weather mechanical cooling may still be needed under local climate conditions.
Power ceilings and token economics
Electricity availability often sets capacity ceilings. NVIDIA says DSX MaxLPS and DSX Flex are the response. DSX MaxLPS reallocates power within an AI factory’s provisioned budget, aiming for the same number of tokens with up to 30% less power, and potentially more GPU capacity or throughput under the same ceiling.
Partners exploring DSX MaxLPS include CoreWeave, Firmus, Google Cloud, IREN, Lambda, Nebius, Nscale and Sharon AI.
For a proof of concept on NVIDIA HGX B200 systems, Lambda reported 24% more cluster token throughput and 19% more GPU capacity within a fixed power budget.
Grid-facing operation
DSX Flex is framed as a control layer for flexible AI loads. It can adjust power use around utility needs, demand-response events and energy prices, while coordinating grid supply, onsite renewables and storage. The goal is to support grid reliability and enable faster grid connection.
A pilot involving Emerald AI, EPRI, Nebius, the U.K. National Grid and NVIDIA found reductions of up to 40% in the cluster’s power draw on an NVIDIA Blackwell Ultra GPU cluster during simulated grid events, while critical workloads continued.
Nscale, Iceland and Portugal
Nscale is one demonstration site: DSX MaxLPS, DSX Flex and DSX Sim link power optimization, grid-responsive operations and design.
In an Iceland AI factory evaluation, DSX MaxLPS ran across four NVIDIA GB300 NVL72 racks. NVIDIA reports 49% higher inference throughput and 37% more managed GPU capacity within the same 264-kilowatt provisioned power budget, showing how dynamic allocation can release capacity from installed hardware.
Nscale is also exploring DSX Flex to align workload power with grid conditions. For facilities in Portugal and Iceland, DSX Sim digital twins let teams assess power, cooling and infrastructure settings before deployment, reduce risk, catch issues early and shorten the path to productive capacity.
Simulation assets and SimReady
NVIDIA’s newly launched SimReady Central Library includes validated power and cooling equipment assets, including material from Vertiv, and accelerates DSX AI factory digital twin creation.
With DSX Sim, ecosystem partners including PTC help create simulation-ready assets through computer-aided design. Those assets feed SimReady workflows that allow builders to model infrastructure configurations and test power and cooling choices before installation, helping validate designs earlier, lower deployment risk and move from planning to construction faster.
Operations software: NVIDIA DSX OS
NVIDIA is extending the modular open-source NVIDIA DSX OS portfolio with three components in developer preview to help operators turn installed infrastructure into productive capacity.
- NVIDIA Config Manager for automated network configuration in AI factories.
- NVIDIA Rack Management Service for rack-level power, inventory, firmware and switch management.
- NVIDIA Cluster Readiness Engine for pre-production distributed workload runs that validate GPU clusters and isolate node-level failures.
They join NVIDIA Infrastructure Controller, NVIDIA AI Cluster Runtime and NVSentinel to automate lifecycle management, maintain consistent runtimes and improve health validation, fault attribution and recovery.
Cloud partners exploring the latest NVIDIA DSX OS components include Aleria, Cassava, CoreWeave, Crusoe, Firmus, Firebird, G42, GMI Cloud, IREN, Lambda, Nebius, Nscale, Radiant, Sharon AI and Volta.
Europe’s industrial ecosystem roles
The buildout across Europe’s facilities is supported by industrial software, power and cooling partners through DSX reference designs, lifecycle digital twins and qualified equipment, linking facility infrastructure with accelerated computing.
| Partner | DSX-related role |
|---|---|
| Schneider Electric | Power and cooling reference designs for NVIDIA Vera Rubin; digital twins with AVEVA across infrastructure design, construction and operations. |
| Procore | Construction management platform integrated with DSX Sim to support AI factory builds. |
| Siemens | Uses DSX Sim to simulate power, cooling and acoustics. |
| Dassault Systèmes and Vertiv | Digital twin models of Vertiv’s OneCore infrastructure for simulation and optimization through model-based systems engineering. |
| ABB | SimReady assets with NVIDIA DSX Sim; source-to-rack architecture evaluation before construction; advancement of 800 VDC through Infinitus Direct Current. |
| Trane Technologies and Eaton | Integrated design for NVIDIA Vera Rubin DSX AI Factory Reference Design to improve data center efficiency and lower installation cost; combines Eaton’s Beam platform and power infrastructure with Trane Continuum cooling and coordinated controls in a grid-to-chip architecture. |
| Trane Technologies | Seven thermal management system reference designs, including three 250-megawatt DSX-based cooling reference designs. |
| LiquidStack GigaModular and CoolChip | Trane Technologies’ LiquidStack GigaModular and Vertiv’s 2.3-megawatt CoolChip, both coolant distribution units, qualified under applicable DSX categories through NVIDIA DSX Ready. |
| Hitachi Energy | Battery energy storage system earned NVIDIA DSX Ready BESS qualification; integration between its energy management system and DSX Flex to optimize grid, generation and storage resources with workload demand. |
| GE Vernova, Siemens Energy and Hitachi Energy | Power infrastructure and digital twin collaborations with NVIDIA for energy supply and grid integration, linking energy-system engineering and grid planning to facility requirements. |
Data citation snapshot
| Metric | Value | Scope |
|---|---|---|
| Cooling water saving | 54 million liters of water a day | 2-gigawatt scale, versus tower cooling; equals daily use of 180,000 households |
| Liquid cooling | 100% liquid cooling; up to 45 degrees Celsius | NVIDIA Vera Rubin |
| Power efficiency | up to 30% less power | same number of tokens with DSX MaxLPS |
| Lambda HGX B200 | 24% more cluster token throughput; 19% more GPU capacity | fixed power budget |
| Grid trial | up to 40% in the cluster’s power draw | NVIDIA Blackwell Ultra, simulated grid events |
| Nscale Iceland | 49% higher inference throughput; 37% more managed GPU capacity | 264-kilowatt provisioned power budget, four NVIDIA GB300 NVL72 racks |
| Reference designs | seven thermal management system reference designs; three 250-megawatt DSX-based cooling reference designs | Trane Technologies |
| Equipment qualification | 2.3-megawatt CoolChip | NVIDIA DSX Ready |
FAQ
What is the main infrastructure claim?
NVIDIA’s claim is that Europe’s AI cloud and facility partners can lower cooling water and electricity intensity while using the NVIDIA DSX AI Factory Platform to design, build and operate AI factories.
How does water use fall?
Dry-cooler-based heat rejection avoids evaporative cooling. The platform can support 100% liquid cooling, and NVIDIA Vera Rubin can run with coolant temperatures of up to 45 degrees Celsius. At 2-gigawatt scale, NVIDIA DSX can save 54 million liters of water a day, equal to daily use by 180,000 households.
What does DSX Flex do for grid operators?
It adjusts workload power to utility needs, demand-response events and energy prices, while coordinating grid power, onsite renewables and storage. In one trial, Emerald AI, EPRI, Nebius, the U.K. National Grid and NVIDIA showed reductions of up to 40% in the cluster’s power draw on an NVIDIA Blackwell Ultra GPU cluster during simulated grid events, while critical workloads continued.
Which software components define the platform?
The platform includes DSX MaxLPS, DSX Flex, DSX Reference Design, DSX Sim, DSX OS and DSX Exchange. NVIDIA DSX OS now has three components in developer preview: NVIDIA Config Manager, NVIDIA Rack Management Service and NVIDIA Cluster Readiness Engine.
For AIdataNewsGlobal General readers, the through-line is operational: the NVIDIA DSX platform turns power, cooling, grid coordination and validation into one set of design choices for Europe’s AI factories. The practical tests will be local climate, deployment validation and adoption of qualified equipment through NVIDIA DSX Ready.