Nvidia Extends AI Cloud Ecosystem Worldwide

Nvidia says its AI Cloud ecosystem is rapidly scaling to support the global buildout of "AI factory" infrastructure, as partners add capacity for enterprises, startups, national programmes, AI labs and developers rolling out agentic AI applications. The company describes the offer as purpose-built clouds, co-designed with Nvidia's full-stack accelerated computing, networking and AI software, to serve soaring token demand and bring compute closer to data, users and industry hubs across regions.

"Every company and every country needs AI factory infrastructure to turn data into intelligence," said Jensen Huang, founder and CEO of NVIDIA. "NVIDIA AI Clouds bring full-stack AI factories closer to the regions, industries and developers building the next generation of AI, from model training to real-time inference and AI agents that will transform how people and organizations work."

Nvidia highlights momentum across telcos, sovereign AI builders, vertical infrastructure providers and specialist AI cloud operators addressing frontier models, enterprise AI, telecoms, developer clouds and national AI programmes. Regional expansion is quickening in Southeast Asia, Australia and the Americas, with coverage now on six continents following additions including Cassava in Africa and Claro in South America.

Partners such as CoreWeave, Firmus, IREN and Nscale are scaling out large AI factory footprints to serve model training, fine-tuning, agentic applications and high-volume inference. Across local markets, providers including Firebird, GMI Cloud, Indosat Ooredoo Hutchison, Lambda, Naver Cloud, Sharon AI, Yotta and YTL are supporting emerging AI companies, government initiatives and sectors from financial services and telecoms to manufacturing, education, healthcare and developer ecosystems. Nvidia says regional clouds can aid sovereign controls and compliance for governments and regulated industries, while lowering friction for developers and enterprises that need low-latency access to accelerated infrastructure near their users and data.

Firmus Technologies is widening its AI factory presence in South Australia and Southeast Asia to meet demand for large-scale training, inference and agentic workloads. Through Project Southgate, Firmus is developing sites in Tasmania, Melbourne, South Australia and New South Wales, focusing on renewable energy, advanced liquid cooling and modular builds to accelerate time-to-capacity. The company has also deployed infrastructure in Singapore via ST Telemedia Global Data Centres. Firmus is using Nvidia accelerated computing and reference architectures, with the Nvidia DSX platform to streamline design, deployment and operations. Its liquid-cooled HyperCube, engineered in alignment with DSX, aims to fast-track modular builds and optimise for low cost per token while innovating across cooling and energy.

"AI agents are creating a new class of industrial-scale demand for tokens, and Asia-Pacific needs AI factories that can be built faster, liquid-cooled more efficiently and operated at gigawatt scale," said Tim Rosenfield, co-CEO of Firmus. "Together with NVIDIA, Firmus is building liquid-cooled, AI infrastructure designed to deliver AI tokens as efficiently and rapidly as possible for the region's most important customers."

CoreWeave is expanding its Nvidia AI Cloud platform to power the next wave of agentic and physical AI, as well as frontier model workloads. An early adopter of Nvidia Vera Rubin and the Nvidia Vera CPU, CoreWeave is also among the first to use Nvidia Spectrum-X Ethernet Photonics to provide a networking foundation for million-GPU AI factories. The company is extending its platform for robotics and physical AI, incorporating Nvidia Cosmos 3 — a frontier world foundation model — to generate synthetic data, fine-tune models and accelerate robotics data flywheels. Leading labs including Anthropic run large-scale frontier models on CoreWeave's infrastructure.

"AI factories are becoming the foundation for the agentic era," said Michael Intrator, cofounder, chairman and CEO of CoreWeave. "Together with NVIDIA, CoreWeave is building the full-stack cloud infrastructure that gives AI labs, enterprises and developers the performance, scale and reliability they need to turn frontier models, AI agents and physical AI systems into production applications."

Nebius is broadening its Nvidia-powered AI Cloud with a full-stack platform spanning training, inference and physical AI development. Also an early adopter of Nvidia Vera Rubin, Nebius is integrating infrastructure from silicon through software, including the Nebius AI Cloud, a Token Factory inference layer and a new Physical AI Workbench. The workbench combines technologies such as Nvidia Cosmos 3, Nvidia Isaac Sim and Isaac GR00T into composable workflows that AI agents can assemble, helping robotics and autonomous systems teams move faster from simulation and synthetic data to training and evaluation.

"Developers should be able to build AI systems without spending weeks wiring together infrastructure," said Arkady Volozh, founder and CEO of Nebius. "With NVIDIA, Nebius is creating an AI cloud where AI agents can compose the tools, data and compute needed to accelerate AI workloads — from robotics and life sciences to the enterprise — from experimentation to production."

Since Nvidia introduced its Exemplar Cloud programme last year, six Nvidia Cloud Partners — CoreWeave, Crusoe, Lambda, Nebius, Vultr and YTL — have achieved Exemplar Cloud status. Nvidia says these providers set a higher bar for consistent performance, reliability and efficiency, giving enterprises, developers and AI labs more validated options for scaling training, inference and agentic services.

As AI shifts from pure model development towards reasoning and sustained, high-volume inference, Nvidia argues the key metric is no longer just announced capacity but the economics of token output — driven by utilisation, uptime, asset longevity and the breadth of deployable AI agents. Built on Nvidia's full-stack platforms, AI Clouds are designed to optimise infrastructure around these measures. The company frames cost per token as a true total cost of ownership metric that reflects hardware performance, software optimisation, ecosystem support and real-world usage, and says it delivers the industry's lowest cost per token through delivered token throughput, software stack optimisation and co-design across compute, networking, memory and storage.

To help providers stand up capacity faster and run more efficiently, Nvidia says AI Clouds are adopting the Nvidia DSX platform to design, build and operate AI factories. DSX brings validated reference designs, simulation, software and ecosystem technologies to accelerate time-to-revenue while improving resiliency. DSX Sim lets teams model and validate factories pre-deployment; DSX Flex adapts workloads to grid conditions; DSX MaxLPS targets power-constrained sites by enabling up to 40% more GPUs within a fixed power envelope; and DSX OS automates lifecycle management and operations at scale. According to Nvidia, these capabilities can reduce deployment risk, increase tokens per watt and drive towards the lowest-cost token output.

www.nvidia.com/en-gb/