The Global Expansion of AI Factories: How NVIDIA’s Cloud Ecosystem is Reshaping Infrastructure

The global demand for artificial intelligence is shifting from experimental development to large-scale industrial production. As enterprises, startups, and nations race to deploy agentic AI and high-volume inference, the backbone of this transformation has become the “AI factory”—a purpose-built infrastructure designed to turn raw data into actionable intelligence at speed.

NVIDIA is currently accelerating the buildout of this infrastructure through its AI Cloud ecosystem. By co-designing full-stack computing, networking, and software platforms, NVIDIA is enabling a growing network of global partners to provide the capacity needed for the next generation of AI applications.

Meeting the Surging Demand for AI Infrastructure

Modern AI workloads—ranging from frontier model training to real-time physical AI and autonomous agents—require more than just raw processing power. They demand high-throughput, energy-efficient, and reliable environments. NVIDIA’s AI Cloud partners are addressing this by deploying regional capacity that brings AI factories closer to where data and developers reside.

This ecosystem is expanding across six continents. Recent additions, such as Cassava in Africa and Claro in South America, highlight the push to provide sovereign and regional AI capacity. This is critical for governments and regulated industries that must adhere to local compliance requirements while still accessing the performance of state-of-the-art accelerated computing.

Engineered for Efficiency: The Role of DSX

A primary challenge in scaling AI infrastructure is the “cost per token”—the total cost of ownership relative to the output of an AI system. To optimize this, NVIDIA has introduced its DSX platform, which provides validated reference designs and automated operational tools to help cloud providers build factories faster and more efficiently.

The DSX platform includes several specialized components:

• DSX Sim: Allows teams to model and validate AI factory performance before physical deployment.
• DSX Flex: Enables dynamic workload adaptation based on power grid conditions.
• DSX MaxLPS: Helps maximize compute capacity within a fixed power budget.
• DSX OS: Automates lifecycle management to ensure consistent uptime at scale.

Innovation on the Front Lines

Several key partners are leveraging this full-stack approach to push the boundaries of what AI factories can achieve:

• Firmus Technologies: Focused on the Asia-Pacific region, Firmus is utilizing the NVIDIA DSX platform and liquid-cooled, modular infrastructure to build energy-efficient AI factories in Australia, and Singapore. Their “HyperCube” design is specifically engineered to minimize the cost per token while scaling to gigawatt levels.
• CoreWeave: As an early adopter of NVIDIA’s latest networking and silicon technologies, including Spectrum-X Ethernet Photonics and the Vera CPU, CoreWeave is positioning itself to support the next wave of physical AI and robotics. By integrating foundation models like NVIDIA Cosmos 3, they are helping AI labs generate synthetic data and accelerate robotics development.
• Nebius: Nebius is developing an open Physical AI Workbench that integrates simulation and synthetic data tools. This allows robotics and autonomous systems teams to move from simulation to production-ready applications without the traditional friction of wiring disparate infrastructure components together.

Key Takeaways

• Global Reach: The AI Cloud ecosystem now spans six continents, supporting both private sector innovation and national AI initiatives.
• Economic Optimization: Success in the AI era is measured by “cost per token,” driving a shift toward highly optimized, energy-efficient factory designs.
• Full-Stack Integration: By combining accelerated computing, advanced networking, and orchestration software like DSX, partners can reduce deployment risks and improve resiliency.
• The Agentic Era: Infrastructure is evolving to support “agentic AI”—systems capable of performing complex, multi-step tasks—which requires reliable, low-latency access to compute resources.

Looking Ahead

The transition toward agentic and physical AI marks a new chapter in the digital landscape. As these technologies move from the lab to the real world, the role of the AI factory will only become more central. By prioritizing modularity, power efficiency, and integrated software management, the NVIDIA AI Cloud ecosystem is setting the stage for a future where AI capability is limited not by infrastructure access, but by the ingenuity of the developers and organizations building the next generation of intelligent tools.