Coordinate workload placement and power across GPU, rack, and workload levels to deliver more AI output within a fixed power envelope.
NVIDIA DSX AI Factory Platform
The NVIDIA DSX™ AI Factory Platform unifies design, simulation, operations, and ecosystem technologies to help build AI factories optimized for lowest token cost.
Overview
A Platform for Co-Designing Efficient, Resilient AI Factories With Lowest Token Cost
NVIDIA DSX defines how AI factories are designed, built, and optimized across the full stack—from the grid, facilities, and chips to infrastructure software and partner technologies—to produce more useful AI output from available power and infrastructure.
NVIDIA DSX combines open software libraries, workflow guides, and reference designs with NVIDIA compute platforms and co-designed OEM infrastructure to enable a broad ecosystem of software and hardware solutions.
Benefits
How Does NVIDIA DSX Improve AI Factory Performance and Tokens per Watt
NVIDIA DSX AI factories are co-designed as unified products—optimized for lowest token cost and energy efficiency.
Reduce Time to First Token
Accelerate bring-up with reference designs, pre-deployment bottleneck detection, simulation validation, and pre-tested software stacks.
Maximize Performance per Watt
Improve energy efficiency and performance per watt by aligning infrastructure, power, cooling, and system behavior through one architecture.
Increase Operational Resiliency
Operate AI factories with greater stability under real production conditions, including hardware faults, grid events, operational changes, and continuous large-scale workloads.
Secure AI at AI Factory Scale
Deliver end-to-end integrated accelerated security to support AI factories with NVIDIA Confidential Computing and NVIDIA DOCA™ microservices.
Discover DSX at GTC Berlin 2026
Join us October 20–22 to explore how NVIDIA DSX helps infrastructure builders design, simulate, and operate efficient, resilient AI factories optimized for lowest token cost.
Technologies
DSX Platform Technologies for AI Factories
NVIDIA DSX is a platform that spans multiple technologies that partners can build with and build on—designed to maximize AI tokens per watt across the AI factory.
NVIDIA DSX MaxLPS
Maximize tokens per watt with dynamic power management.
NVIDIA 800 VDC Architecture
Scale AI compute performance.
NVIDIA DSX Flex
Leverage grid-responsive power orchestration for AI factories.
NVIDIA DSX Sim
Simulate AI factory design and operations.
NVIDIA DSX OS
Efficiently operate and scale AI factories.
NVIDIA DSX Exchange
Secure data exchange across IT, OT, and AI factory operations.
Solutions
Operate at Scale With Validated Infrastructure
Build and scale AI factories with trusted solutions qualified for applicable NVIDIA DSX reference design requirements.
- Quote 1
- Quote 2
- Quote 3
- Quote 4
Resources
NVIDIA DSX Resources
Learn more about NVIDIA DSX through blogs, GTC sessions, documentation, and videos.
Next Steps
Start Building With NVIDIA DSX
Design and optimize AI factories with NVIDIA reference architectures, simulation tools, and ecosystem software.
NVIDIA DSX Ready
Explore the trusted program for AI factory builders, offering products and solutions that meet category-specific NVIDIA DSX requirements.
FAQ
How NVIDIA DSX Improves Efficiency and Performance per Watt
NVIDIA DSX is NVIDIA’s AI factory-scale platform. It brings together design, simulation, operations, and ecosystem technologies to help partners build AI factories optimized for lowest token cost. DSX spans chips and systems, infrastructure software, facilities, and partner technologies, enabling AI factories to produce more tokens per watt.
It includes the following technologies: DSX Reference Design, DSX Sim, DSX MaxLPS, DSX Flex, DSX Exchange, and DSX OS.
An AI factory is purpose-built for AI workloads, with an emphasis on inference performance, energy efficiency, and lowest token cost. NVIDIA DSX AI factories are co-designed as unified systems across chips, systems, networking, infrastructure software, power, cooling, and operations, unlike traditional data centers that are often assembled from siloed components.
Traditional data centers primarily store, process, retrieve, and serve information. AI factories generate intelligence, turning data and energy into tokens, predictions, simulations, and AI-powered outputs.
The NVIDIA Omniverse™ DSX Blueprint is an open, comprehensive framework within DSX Sim for designing and operating large-scale AI factories. It is generally available on build.nvidia.com and fully compatible with the NVIDIA Vera Rubin DSX AI Factory reference design.
The blueprint provides the recipe and building blocks for developers to build physically accurate digital twins of their AI factories, simulate physical and logical behavior, and optimize performance before construction or deployment begins. It includes a complete AI factory digital twin, NVIDIA SimReady assets, hardware configs, prebuilt web UI, sample scripts, workflow guides, and NVIDIA CUDA-X™ and Omniverse core technologies.
NVIDIA DSX serves a broad, multi-discipline ecosystem: NVIDIA Cloud Partners (NCPs), sovereign clouds, and Land/Power/Shell (LPS) providers; manufacturers of power and cooling equipment; Independent Software Vendors (ISVs) for design, simulation, and facilities operations; Original Equipment Manufacturers (OEMs) and system integrators; architectural, engineering, and construction firms; and infrastructure ISVs providing IaaS and PaaS solutions.
DSX provides a common design standard so all parties—from Coolant Distribution Unit (CDU) manufacturers to ISVs—can contribute to AI factory builds without bespoke integrations.
NVIDIA DSX helps lower token cost by enabling AI factories to produce more tokens from the same power, infrastructure, and operations budget. Instead of treating compute, cooling, power, and operations as separate systems, DSX coordinates them as one integrated AI factory.
DSX MaxLPS helps recover stranded power so customers can provision up to 40% more GPUs within the same megawatt budget. DSX OS improves uptime through automated fleet-scale operations. DSX reference designs accelerate deployment, while DSX Exchange and DSX Sim help unify facility, power, cooling, and compute signals to improve throughput, efficiency, and tokens per watt.
Start with the NVIDIA DSX documentation to explore reference designs, software components, and partner integration paths for building AI factories. The docs provide technical guidance for designing, optimizing, and operating DSX-aligned infrastructure.