GMI Cloud is among the cloud partners building with [A1] [A2] the newest components of NVIDIA DSX OS, part of the NVIDIA DSX AI Factory Platform[A3] that NVIDIA highlighted at NVIDIA GTC Berlin on Oct. 21, 2026.
What NVIDIA announced
At GTC Berlin, NVIDIA said leading European ecosystem partners are adopting the NVIDIA DSX AI Factory Platform. NVIDIA cloud partners across Europe are building and running DSX-based facilities, widening access to accelerated computing for enterprises, developers and public-sector organizations there.
DSX combines accelerated computing, software, power, cooling and digital twins. The aim is to cut the resources it takes to deliver useful capacity, bring new capacity online sooner and let workloads adjust their power draw to help keep the grid stable.
The efficiency numbers
● Power: NVIDIA DSX MaxLPS software can help operators produce the same token output with as much as 30% less power.
● Grid: NVIDIA DSX Flex software lets an AI factory act as a flexible power load, supporting grid reliability and speeding up grid connection.
● Water: In suitable climates, DSX cooling designs can do away with evaporative cooling. Compared with cooling-tower systems, that saves roughly 2.6 billion gallons of water a year for each gigawatt.
The building blocks of DSX
● DSX MaxLPS — dynamic power software and 45 degree Celsius liquid cooling
● DSX Flex — workload power management based on grid conditions
● DSX Reference Design — validated infrastructure architectures
● DSX Sim — high fidelity digital twins and simulation
● DSX OS — infrastructure and platform software
● DSX Exchange — link between computing systems and facility systems
Together, these help operators plan, build and run AI factories that use fewer resources.
What this means at GMI Cloud
GMI Cloud operates its own AI factories, including the Taiwan AI Factory, across 8+ data centers in five countries. We are exploring the latest DSX OS components to automate infrastructure lifecycle management across that footprint and to make our platform more resilient and reliable.
Brought into our AI factories, these components would let us:
● Bring new capacity online faster. With NVIDIA Config Manager automating network setup and NVIDIA Rack Management Service handling power, inventory, firmware and switches at the rack level, we could cut the manual work between installing hardware and putting it to use.
● Validate clusters before customer workloads arrive. NVIDIA Cluster Readiness Engine would let us check GPU clusters with distributed workloads before production work lands, and trace any failure to a specific node.
● Keep environments consistent and recoverable. Together with NVIDIA Infrastructure Controller, NVIDIA AI Cluster Runtime and NVSentinel, these tools would help us keep runtime environments consistent and strengthen health checks, fault tracing and recovery.
NVIDIA describes DSX OS as open source software that automates the infrastructure lifecycle and strengthens health automation and resiliency, so operators can run AI factories at scale. Those are the same goals behind our exploration: more automation in how we manage infrastructure across our AI factories, and a more resilient, reliable platform for the teams building on GMI Cloud.
Build AI Without Limits
GMI Cloud helps you architect, deploy, optimize, and scale your AI strategies
