The NVIDIA DSX AI Factory Platform brings together the design, simulation and operation of AI factories. DSX Ready is its qualification program for specific equipment and solutions. Together, they broaden equipment selection beyond individual specifications to whether compute, power, cooling and software can operate as a coordinated system.
For enterprises planning an AI data center, procurement therefore needs to consider equipment capabilities, system interfaces and on-site delivery together. DSX Ready helps identify qualified products. Whether an entire facility can support the intended workloads still requires site-specific engineering design and acceptance testing.
NVIDIA DSX Connects AI Factory Design and Operations
At its core, DSX considers facility conditions and computing systems together from the beginning of the design process. This helps enterprises identify power, cooling or cluster-configuration bottlenecks earlier, reducing the risk of discovering only after equipment arrives that it cannot operate as intended.
NVIDIA DSX encompasses reference designs, simulation tools, operational software and partner technologies. Its goal is to increase useful AI output within a fixed power and facility budget. Actual energy and cost savings depend on the workloads, hardware combination and implementation.
Enterprises can understand its role across three stages:
| Stage | Relevant DSX Capabilities | What This Means for Project Teams |
|---|---|---|
| Design | Reference designs provide system architectures for the relevant technology generation. | Check compute, networking, storage and facility requirements within a single design. |
| Simulation | DSX Sim supports simulation of physical and logical systems. | Examine configurations and system interactions before deployment, then compare expected and actual behavior after deployment. |
| Operations | DSX OS provides modular software; power management and signal integration support coordination across systems. | Incorporate workloads, equipment health and facility status into operational decisions. |
DSX brings IT systems and mechanical and electrical infrastructure into a common architecture, while DSX OS supports coordination across systems during operation. Project teams need to establish data interfaces and operational responsibilities early so that equipment status can inform workload adjustments, alerts and maintenance decisions.
Rack Specifications, Cluster Architectures and Data Center Qualification Programs Predate DSX
AI infrastructure already has mechanisms for joint design and validation that address rack interfaces, cluster configurations and colocation environments. DSX brings these engineering considerations to the scale of an AI factory, allowing compute, power distribution, cooling, simulation and operations to be designed together.
| Existing Specification or Program | Primary Focus | How It Differs from DSX and DSX Ready |
|---|---|---|
| DGX-Ready Data Center (launched in 2019) | Identifies colocation data centers suitable for DGX deployment, helping enterprises access suitable power and cooling environments. | Focuses on colocation facilities; DSX Ready qualifies specific CDU, BESS and coolant products. |
| DGX SuperPOD Reference Architecture (design documentation available in 2019) | Provides an approach to building clusters with DGX compute, networking, storage and software. | Focuses on scalable AI clusters; DSX further incorporates facilities and operational controls into a common architecture. |
| OCP Open Rack (including ORv3 from 2022) | Industry-developed open specifications for racks, power supplies, busbars and connection interfaces. | Provides a common foundation for hardware interfaces; DSX integrates NVIDIA computing platforms, facility design and software tools. |
| Omniverse DSX Blueprint (launched in 2025) | Uses digital twins to connect AI factory design, simulation and operations. | A design and simulation framework in the evolution of DSX, now incorporated into DSX Sim. |
DSX can therefore be understood as an NVIDIA-led AI factory platform and framework for joint design, while DSX Ready provides product qualification. Open specifications such as OCP can still underpin hardware design; the two work at different levels. For enterprises, the added value lies in simulating interactions across systems before construction and carrying the same equipment data into operations.
DSX Ready Qualifies Specific Products and Solutions
DSX Ready helps buyers identify products that meet the qualification requirements for their category. The scope is explicit: qualification of one product does not mean that every product from its supplier, or an entire data center using that product, has the same qualification.
DSX Ready currently covers cooling distribution units (CDUs), battery energy storage systems (BESS) and PG25 coolant for single-phase cold-plate cooling in secondary loops.
| Category | Role in an AI Factory | What Buyers Still Need to Check |
|---|---|---|
| CDU | Circulates and manages coolant to maintain the temperature, flow and pressure required by liquid-cooled equipment. | The qualified model, test conditions and compatibility with on-site fluid circuits, piping materials and control systems. |
| BESS | Stores and dispatches electrical energy to address load changes, grid events or power-source transfers. | The qualified configuration and operating scenarios, plus integration with power distribution, protection and control strategies. |
| PG25 Coolant | Removes heat from computing components through single-phase cold-plate cooling in a secondary loop. | The qualified formulation, fluid-quality and corrosion-resistance requirements, loop materials and maintenance conditions. |
The CDU qualification guidelines, BESS qualification guidelines and PG25 fluid requirements each specify their own tests and supporting information. Product qualification reduces some uncertainty in equipment selection. On-site teams must still validate how piping, power distribution and control systems behave when connected.
Meeting DSX Ready Requirements: Qualification and Procurement Checks
Equipment suppliers must complete NVIDIA's qualification process for the relevant product category to obtain DSX Ready status. Buyers must confirm that qualification covers the actual model and configuration they intend to purchase. Product qualification and site-level regulatory compliance are separate matters; engineering teams remain responsible for applicable site safety and regulatory requirements.
1. Confirm the category and identify gaps. Check the product against the qualification documents for CDUs, BESS or coolant, then contact NVIDIA to confirm the applicable requirements. For a CDU, for example, suppliers must respond to product requirements individually, list wetted materials and complete the core requirements checklist. Where gaps exist, they must explain the current status and planned improvements.
2. Complete testing and submit evidence. Provide test reports, raw data and the required documentation under the applicable guidelines. CDU testing covers areas such as cooling performance, controls and redundancy. BESS assessment examines responses to load and grid changes, along with monitoring and control data. NVIDIA determines whether the product qualifies after completing its review.
3. Confirm qualification and listing. Once qualified, a product can be listed in NVIDIA Marketplace and use the corresponding qualification designation. The time required varies by product category, scope and readiness, and can be confirmed with NVIDIA. See the DSX Ready application process for details.
Buyers can begin by checking product listings in NVIDIA Marketplace, then verify the model, configuration and qualification scope with the supplier. They should also include on-site power, fluid circuits, control interfaces and testing responsibilities in the project's acceptance criteria, confirming that qualified equipment can meet the agreed operating requirements once integrated at the site.
Companies Participating in the DSX Ecosystem
DSX participants include power and cooling equipment suppliers, engineering software companies and compute operators. Equipment suppliers provide products that meet the requirements, software companies connect design and simulation tools to the platform, and operators use it to plan and validate AI factories.
When DSX Ready launched in September 2026, the initial qualified CDU solutions came from LG Electronics, LiquidStack and Vertiv, while BESS solutions came from Hitachi Energy, LG Energy Solution and Tesla. NVIDIA's program announcement lists suppliers in both categories.
Specific products include the Vertiv CoolChip 2.3 MW CDU and the LG Electronics 2.6 MW CDU. These high-capacity CDUs support liquid cooling across multiple racks, bringing cooling-equipment selection closer to the scale of the cluster.
Beyond product qualification, DSX platform partnerships also extend into design tools and operational workflows:
| Company | Participation | Problem Addressed |
|---|---|---|
| Schneider Electric | Integrates the ETAP electrical simulation platform with DSX. | Simulates power distribution before construction to assess configurations and operating approaches. |
| Cadence and Jacobs | Apply DSX digital twins in simulation platforms and engineering services. | Assess thermal flows, layouts and the effects of design changes in advance. |
| Switch | Uses the Omniverse DSX Blueprint to build EVO AI Factories and operational systems. | Connects equipment-monitoring data and digital twins with power and cooling optimization. |
| CoreWeave | Uses DSX Air to build and test AI factory digital twins. | Rehearses and validates operational workflows before physical equipment is delivered. |
This gives the supply chain a shared basis for design and testing. Alongside equipment specifications, buyers can compare more precisely how far each offering extends into simulation tools, engineering integration and operational support.
Equipment Competition Expands to System Compatibility and Delivery
DSX affects the industry by requiring suppliers to demonstrate more clearly how their products integrate into an AI factory. Specifications such as power and cooling capacity remain important, but cross-equipment control interfaces, test data and integration capabilities also influence customer adoption.
Cross-system integration raises requirements in three areas. First, procurement places greater emphasis on verifiable interfaces and operating conditions. Second, integrators need to coordinate power, cooling and computing. Third, operators need sufficient monitoring data to determine whether a bottleneck originates in GPUs, workloads or facilities.
Opportunities for Taiwan's supply chain span cold plates, connectors, CDUs, energy storage and integration services. Value at each stage comes from equipment performance, interface compatibility and maintenance support after delivery.
Standardization Can Reduce Integration Costs and Change Supplier Selection
A shared architecture can reduce the work of redefining interfaces and test methods for every project, making equipment combinations easier to replicate. Enterprises must still assess the restrictions they may face when replacing equipment, updating software or deploying other computing architectures after adopting a particular platform.
Reference architectures provide a starting point, but sites differ in incoming power conditions, outdoor heat-rejection space, floor loading and construction constraints. Even when two sites use the same qualified equipment, their actual capacity, redundancy and maintenance arrangements may differ.
Evaluation can be organized into three levels: at the product level, confirm the model and qualification scope; at the system level, confirm the combination of power distribution, liquid cooling and controls; at the site level, confirm sustained loads, abnormal scenarios and maintenance conditions. This brings the benefits of standardization into the project rather than leaving them at the product-list stage.
From DSX Ready Equipment to an Operational AI Data Center
When choosing a construction and integration service, enterprises should ask the provider to specify who designs, integrates, performs acceptance testing and handles operations and maintenance. Even when each item of equipment meets its own specifications, someone must remain accountable for cross-system test results.
For example, if flow drops in the liquid-cooling system, the parties should agree during design and acceptance planning who receives the alert, who adjusts computing loads, who responds on site and how service will be restored. When different vendors handle power, cooling and computing, responsibilities for troubleshooting and change management need to be particularly clear.
Konstra AI, a KONST Brand, Helps Enterprises Implement AI Data Center Technical Requirements
Konstra AI, part of KONST Group, provides mechanical and electrical systems, environmental controls and GPU cluster construction for AI data centers, along with equipment operations and maintenance and compute delivery. It helps enterprises apply technical requirements to new-build, retrofit and expansion projects.
For enterprises planning to adopt DSX reference architectures and DSX Ready equipment, Konstra AI can start with site and power planning, assess rack loads, power capacity and cooling requirements, integrate equipment interfaces, and incorporate test items and delivery responsibilities into acceptance requirements. This helps align the completed facility with the technical requirements adopted for the project.
Contact KONST to discuss AI data center technical requirements and construction needs.
