NVIDIA Launches Open-Source DSX Platform to Standardise AI Factories

A Blueprint for the AI Factory

NVIDIA has pulled back the curtain on DSX, an open-source platform aimed at standardising how companies design and run the enormous data centres now powering artificial intelligence. Announced at GTC Taipei, the platform reframes the company's role from chip supplier to infrastructure architect. "We're not simply delivering chips; we're equipping every infrastructure builder with a thorough playbook to construct AI factories," said NVIDIA founder and CEO Jensen Huang during the reveal.

The centrepiece is DSX OS, modular open-source software now published on GitHub. It handles lifecycle management, scheduling, health automation, and multi-tenant operations across what NVIDIA calls AI factories. Crucially, it's designed for incremental adoption, letting operators fold it into existing software stacks rather than ripping out and replacing their current tooling.

Cooling, Density, and Energy Efficiency

Alongside the operating system, NVIDIA introduced DSX MaxLPS, a set of technologies pairing 45-degree Celsius liquid cooling with in-rack optimisations. The result lets operators run up to 40% more GPUs at their most energy-efficient operating point within a fixed power budget. As AI workloads strain electrical grids worldwide, squeezing more compute out of every watt has become a central engineering challenge.

Talking to the Power Grid

DSX also reaches beyond the server room. A component called DSX Flex connects AI factories to power-grid services, allowing workloads to flex up or down in response to utility signals. A multi-megawatt pilot with Emerald AI and Silicon Valley Power is already demonstrating grid-responsive AI workload management, hinting at a future where data centres act as cooperative partners to the grid rather than insatiable drains on it.

A Broad Ecosystem

The rollout arrives with heavyweight backing. Dell Technologies, HPE, Lenovo, and Supermicro are building DSX-ready systems, while cloud providers including CoreWeave, Crusoe, and Nebius are deploying parts of the stack. Mirantis announced it is an inaugural NVIDIA AI Cloud-Ready ISV partner, weaving DSX OS components into its k0rdent AI platform alongside a large-scale reference deployment with IREN spanning thousands of GPUs.

Open Source as Strategy

NVIDIA's choice to open-source DSX OS mirrors a broader industry pattern in which infrastructure software becomes shared foundation while vendors compete on services and optimisation. By letting providers adopt components selectively and build differentiated offerings on top, NVIDIA encourages adoption without locking customers into a single proprietary stack, cementing its hardware at the centre of the AI buildout.