NVIDIA argues nations need “AI factories” and domestic computing to deploy generative and agentic AI

Nations are treating artificial intelligence as a strategic capability that must be built and deployed domestically, NVIDIA says, pointing to a shift from experimentation to large-scale infrastructure. In a new blog post, the chipmaker argues that the rise of generative and “agentic” AI, which can take actions toward goals rather than only generate text or images, is pushing countries to create the computing and data systems required to produce models and run them in public services and industry.

NVIDIA frames the effort as a modern version of national infrastructure investment, aimed at economic growth, data protection and the ability to tailor applications to local citizens and regulations. The company says countries are investing in the ability to design, train and deploy AI using domestic infrastructure, local datasets and homegrown talent, with the goal of reflecting local language, culture and domain needs in model outputs.

A central part of NVIDIA’s description is data infrastructure. The company says national teams are developing foundation models, including large language models, that are trained on local datasets so outputs better match regional dialects, cultural context and specific use cases. NVIDIA also highlights speech AI as a tool that can support the preservation and revitalization of indigenous languages.

Beyond language, the post links large language models to a wide range of tasks, including writing software code, supporting drug discovery, helping protect consumers from financial fraud, and teaching robots physical skills. NVIDIA adds that as AI and accelerated computing become more important for climate work, energy efficiency and cybersecurity, national AI capabilities become part of broader resilience and sustainability strategies.

NVIDIA’s most concrete infrastructure concept is a new category it calls “AI factories.” In the company’s description, AI factories are next-generation data centers that ingest data and output usable intelligence, built around advanced, full-stack accelerated computing platforms for some of the most computationally intensive AI workloads.

The post also describes different national approaches to acquiring or operating AI computing capacity. NVIDIA says some countries are procuring and running AI clouds in collaboration with state-owned telecommunications providers or utilities. Others are sponsoring local cloud partners to provide shared AI computing platforms designed for public-private use, aiming to keep compute, data handling and governance under national control.

In Europe, NVIDIA cites an example using agent technology built by ThinkDeep on NVIDIA’s AI platform for France’s Ministry of Economy and Finance. The company says the system processes millions of documents and data sources, cutting document search time from two days to two minutes, saving 2 million euros for 10,000 employees, and reducing energy use by using more efficient, in-country infrastructure control.

NVIDIA also points to India’s Sarvam platform, which it says is powered by NVIDIA GPUs and built on domestic infrastructure. The post claims Sarvam is delivering multilingual AI models and voice agents optimized for India’s 22 official languages to support government and enterprise services at scale while keeping data, compute and governance within national control. In Latin America, NVIDIA says Widelabs, running on NVIDIA-accelerated infrastructure, is helping modernize legal services for Brazil’s Public Ministry of Rio Grande do Sul by streamlining internal investigations and making justice records easier to find and use for more than 8 million citizens across nearly 500 municipalities.

NVIDIA’s post is broad, and it does not provide the specific list of “five ingredients of a national AI strategy,” only stating that such a framework exists. It also does not quantify how widely “AI factories” are being adopted or detail procurement terms for national compute clouds, meaning observers will need to look for policy documents, vendor disclosures and government procurement notices to verify how each country operationalizes the concept.

What to watch next is whether the “AI factory” framing translates into public funding and standardized procurement across countries, and how quickly national “foundation model” efforts move from training to reliable deployment in live services. NVIDIA’s post also includes a near-term event reference, saying NVIDIA will participate in the AI for Good Summit in Geneva from July 7 to 10, which could offer additional detail on how the company sees these deployments taking shape.

“The AI factory will become the bedrock of modern economies across the world,” NVIDIA founder and CEO Jensen Huang said in a media Q&A included in the post.