NVIDIA launches DSX AI Factory platform in Europe with $10B investment

Expanding European AI Infrastructure

NVIDIA announced the launch of the DSX AI Factory Platform during the GTC Berlin conference on October 21, 2026. This initiative involves major European technology partners building and operating specialized AI factories designed to provide more GPU capacity. The goal is to create high-performance environments that can handle the heavy demands of modern artificial intelligence.

To support this expansion, NVIDIA pledged nearly $10 billion in financial commitments to the European technology ecosystem. This investment aims to bolster the region's ability to produce and deploy large-scale AI models. The company highlighted Europe's potential in industrial precision manufacturing as a key driver for these new factories.

Several prominent companies, including CoreWeave, Google Cloud, and Nebius, have already planned to use this solution. These partners will work together to establish the infrastructure necessary for continuous AI operations. The focus remains on creating environments that can produce tokens—the basic units of text or data generated by AI—at a high and consistent rate.

}

Accelerating Deployment and Scale

NVIDIA is also focusing on the physical speed of building data centers through the DSX POD. This container-based reference design for Vera-Rubin clusters allows for integration and validation before the units are even shipped. This modular approach can shrink the construction time for a data center from six months down to just 14 days.

The global delivery of these modular systems is scheduled to begin in early 2027 with partners like Dell Technologies and HPE. This move aims to provide a faster way for companies to scale their hardware. By standardizing the container designs, NVIDIA and its partners can deploy large-scale clusters more reliably across different regions.

Several large-scale projects in Europe are already moving forward to utilize this new capacity. Deutsche Telekom is working with vCluster Labs to increase its GPU capacity from 500 to over 10,000 units for German industrial customers. Additionally, a 1.2-gigawatt campus in Portugal is being designed to host more than 66,000 NVIDIA-Rubin GPUs, using a digital twin for pilot testing.

}

Software Tools and Performance

NVIDIA released DSX OS as an open-source preview version for developers on October 21, 2026. This software includes essential components for configuration, rack management, and cluster readiness to help manage complex AI environments. Additionally, the SimReady library was made available on the Hugging Face developer platform to support infrastructure testing.

Real-world performance tests show significant gains in efficiency across different hardware setups. For instance, the company Lambda demonstrated a feasibility study on HGX-B200 systems where the DSX MaxLPS feature provided 24 percent more cluster-token throughput and 19 percent more GPU capacity within a fixed power budget. These results highlight how the platform helps maximize hardware output without increasing energy costs.

Other partners are also seeing measurable improvements by combining different platform features. The company Phaidra achieved roughly 10 percent more computing power without changing the electricity consumption by utilizing the DSX Flex, Max-Q, and Exchange components. These varied results suggest that the platform can be tailored to meet specific hardware and power goals for different industrial needs.

Cooling and Infrastructure Support

To address the physical demands of high-performance computing, Trane is providing seven thermal reference designs. This includes three specific cooling solutions capable of handling 250 megawatt capacities. These designs are part of a broader effort to manage the intense heat generated by large-scale AI clusters while maintaining reliable operations.

The initiative is supported by a wide range of infrastructure partners who are providing reference designs and hardware support. These collaborations ensure that the physical components of the AI factories, from cooling to power management, are compatible with the new DSX platform. This multi-partner approach is intended to create a standardized environment for deploying large-scale GPU clusters across Europe.

These technical advancements come as companies navigate the evolving regulatory landscape in the region. While the DSX platform provides the hardware and software foundation, businesses are also looking at resources to understand the EU AI Act. This includes staying informed on relevant deadlines, obligations, and risk classes to ensure that new AI deployments remain compliant with regional laws.