NVIDIA DSX Is the Open-Source Blueprint for the AI Factory Era

NVIDIA used its GTC Taipei keynote on May 31, 2026, to launch the NVIDIA DSX platform — a full-stack playbook for designing, simulating, building, and operating AI factories . The platform unifies open-source software, reference designs, simulation tools, grid-integration, and partner technologies into a common, co-designed architecture that aims to maximize token performance per megawatt .

DSX OS: An Open-Source Operating System for AI Factories

DSX OS is the central software layer of the platform. It is being released as open-source, modular software that handles lifecycle management, runtime consistency, health automation, resiliency, multi-tenant operations, and platform services . The goal is to create a single operating system for the entire AI factory that integrates compute, software, facilities, and partner technologies .

By making DSX OS open source, NVIDIA is betting on a strategy similar to how Linux and Kubernetes became infrastructure standards. The company wants to accelerate ecosystem adoption, reduce fragmentation between different builders, and allow community-driven customization for diverse AI factory configurations .

DSX MaxLPS: Liquid Cooling and Dynamic Power for 40% More GPUs

DSX MaxLPS is a power and efficiency software suite that dynamically manages power at the GPU, rack, and workload levels . It leverages 45-degree Celsius liquid cooling and fine-grained power allocation to recover stranded power capacity .

The headline result: AI factories can run up to 40% more GPUs at peak energy efficiency within the same fixed power budget, delivering the lowest possible cost per token . This recovers power that is normally wasted, allowing operators to provision more compute without requesting additional power from the grid .

DSX Flex: Making AI Factories Grid-Responsive

DSX Flex connects AI factories directly to power-grid services, enabling dynamic workload adaptation to grid signals such as load shedding, demand response, and pricing events . It orchestrates power across utility sources, on-site renewables, and storage, turning AI factories into flexible assets that can help stabilize the electrical grid rather than simply strain it .

The software is already powering a commercial, multi-megawatt pilot with Silicon Valley Power and Emerald AI to demonstrate grid-responsive AI factories that adjust power consumption in real time while protecting workload performance . This pilot aims to safeguard grid reliability and affordability for all customers .

The Broader DSX Ecosystem and Partner Adoption

NVIDIA designed the DSX platform to span the full stack, from silicon and systems to infrastructure software, facilities, and partner technologies . The platform includes:

• DSX Sim, an Omniverse blueprint for designing and validating AI factories in a digital twin before hardware is deployed .
• DSX Air, a SaaS platform for logically simulating AI factory infrastructure that integrates with partner solutions for storage, routing, and security .

NVIDIA is partnering with leading Taiwan system manufacturers — including Asus, Foxconn, Gigabyte, Pegatron, Quanta Cloud Technology (QCT), Wistron, and others — to expand the DSX ecosystem .

Cloud partners are already deploying core DSX components to reduce risk, improve GPU utilization, and bring AI cloud capacity online faster . The cloud partner list includes:

• CoreWeave
• Crusoe
• Firmus
• IREN
• Lambda
• Nebius
• Nscale
• Yotta Data Services

These partners are deploying DSX Sim, DSX MaxLPS, and DSX OS to build out AI factories with what NVIDIA calls "extreme codesign" at the core .

Why DSX Matters: A Strategy Shift for AI Infrastructure

GTC Taipei 2026 made it clear that NVIDIA is no longer positioning itself as just a GPU vendor or even a rack-scale systems provider. As Jensen Huang stated during the keynote, "Every layer, chip, rack, network, power, cooling, grid must be designed together from end to end because compute is revenues" .

DSX represents a methodology for building AI factories that spans three lifecycle stages: design, deployment, and operation . By open-sourcing the operating system layer and providing reference designs for everything from liquid cooling to grid integration, NVIDIA aims to set the industry standard for how next-generation AI infrastructure is built and operated .

The payoff is economic: more tokens from the same power, infrastructure, and operations budget. DSX MaxLPS alone can recover stranded power to provision up to 40% more GPUs, and the platform as a whole is architected to minimize time to first token while maximizing long-term operational efficiency .