AI is becoming the cornerstone of innovation across industries, driving new levels of creativity and productivity and fundamentally reshaping how we live and work. And it’s enabled by a new type of infrastructure—the AI factory—that manufactures intelligence at scale and creates the foundation for what many consider the next industrial revolution.

AI factories represent a reset of traditional cloud-computing architecture toward an accelerated computing infrastructure that’s designed and optimized for AI workloads. This architectural shift creates opportunities for new players, including specialized AI factory providers sometimes called “neoclouds.” Such players have achieved rapid success by offering high-performance, GPU-centric AI cloud services for model training, fine-tuning, and inferencing to researchers, startups, and enterprises.

The success of neocloud providers has democratized AI, bringing powerful new LLMs and AI models in the hands of millions. It also has shown that the demand for AI factories across all industries from healthcare to automotive is unprecedented and underserved.

This massive need for AI factories opens a new business opportunity for telecom service providers. Recent McKinsey research suggests that the global GPU-as-a-service (GPUaaS) market addressable by telcos could range from $35 billion to $70 billion annually by 2030.

Sovereign AI infrastructure: a national imperative

As artificial intelligence becomes central to economic growth and innovation, nations increasingly recognize the strategic importance of building their own sovereign AI infrastructure. The notion that “compute divide is the new digital divide” resonates strongly with technology leaders and policymakers around the world. Access to AI supercomputing systems is now considered critical national infrastructure, essential for:

• Enabling domestic AI development aligned with national priorities
• Creating new high-skilled jobs and economic opportunities
• Driving productivity gains across industries
• Ensuring a country’s competitiveness in the global AI economy
• Nurturing a national AI innovation ecosystem, including startups, enterprises, and academic and research institutions
• Providing highly productive citizen services at scale in areas such as healthcare, education, and connectivity

Sovereign AI infrastructure enables countries to shape AI development according to their unique cultural, ethical, and economic priorities; reflect local values; and address specific regional challenges. Another key driver is retaining the control, processing, and ownership of local data within geographical boundaries, including compliance to local security policies and regulations.

Telcos: trusted sovereign AI providers

As every nation looks to build its own AI infrastructure, telecom companies have emerged as particularly well-suited providers. This isn’t coincidental but stems from several intrinsic advantages that telcos have:

• Trusted national infrastructure: Telcos already provide the backbone of national digital and connectivity infrastructure that’s secure and complies with regulations. They have established deep trust with governments and businesses, making them natural partners for hosting sensitive national AI infrastructure.
• Access to data centers and fiber connectivity: Telcos already possess extensive data center facilities as part of their core business, giving them a time-to-market and cost-efficiency advantage. Rather than building new facilities from scratch—which takes years and massive capital investment—telcos can rapidly modernize existing data centers for AI, leveraging their existing access to space and power. These data centers are already connected via high-speed fiber networks, another core telco capability.
• Expertise in managing complex systems: Telcos’ experience in managing large-scale, mission-critical networks translates well to operating AI factories, which require similar levels of scale, reliability, security, and performance.

The industry has recognized these advantages, with 18 telcos launching AI factories powered by NVIDIA over the past 18 months. This is not a one-off phenomenon but a major trend. Many more telecom providers are preparing to launch similar initiatives this year, creating a new growth vector for an industry traditionally focused on connectivity.

Telco-built sovereign AI infrastructure examples

Eighteen telco-led AI factories powered by NVIDIA now span five continents: North America, South America, Europe, Asia, and Africa.

• In Japan, SoftBank Corp. and KDDI have announced significant investments to build large-scale AI data centers and collaborate with partners to develop homegrown Japanese-language large language models and enterprise AI solutions.
• In Southeast Asia, Indonesia’s Indosat Ooredoo Hutchison has established the country’s first sovereign AI factory, providing citizen services, healthcare applications and Bahasa-based open-source LLM to enterprises. Singapore’s Singtel has launched RE:AI, the country’s first sovereign AI factory that will offer AI cloud services to enterprises, government agencies, researchers and academia across the region. It aims to support Singapore’s National AI Strategy 2.0 by democratizing AI. YTL AI Cloud is a telco-led sovereign AI factory that supports Malaysia’s digital economy ambitions. FPT has launched AI factories in Vietnam and Japan, providing NVIDIA GPU-accelerated computing platforms and over 20 ready-to-use AI applications that enable enterprises and startups to accelerate sovereign AI development and adoption. In India, Tata Communications is offering AI cloud infrastructures for sectors such as manufacturing, healthcare, retail, and financial services.
• Qatar-based Ooredoo is investing in AI factory buildouts across the Middle East and nearby countries. Its goal is to provide governments, enterprises, and startups in Algeria, Kuwait, the Maldives, Oman, Tunisia, and Qatar with GPUaaS and localized AI ecosystem development.
• In Europe, Iliad Group unit Scaleway has expanded its AI cloud capacity in its home country of France with high-end GPUs to support startups and enterprises. Telenor launched a sovereign, secure and sustainable AI factory in the Nordics that hosts both internal and external tenants. Starting with a small cluster, Telenor is now preparing to scale in its new data center. Swisscom is building the Swiss AI platform, which is investing in Switzerland’s first NVIDIA SuperPOD. In Italy, Fastweb has launched its NeXXt AI factory, which powers the MIIA (Italian AI Model) and offers AI services to the education, finance, and industrial sectors.
• In Latin America, Telconet, Ecuador’s leading telecom operator, has deployed an AI factory to improve urban safety with AI, support implementation of various uses, and bring the latest-generation GPUs to the region.

The momentum behind telco-powered sovereign AI factories continues to build, with several recent announcements bringing the movement to North America, Africa, and Central Asia.

TELUS: Canada’s flagship sovereign AI factory

TELUS has become the first North American service provider to join NVIDIA’s Cloud Partner program, aiming to bring a comprehensive sovereign AI factory to Canada. This initiative will empower Canadian businesses, startups, and researchers with access to cutting-edge AI infrastructure, helping them develop smarter AI products, streamline operations, and stay competitive, all while maintaining data sovereignty.

TELUS’s AI factory will be deployed in its sustainable data centers that are powered by 99% renewable energy, with the first cluster expected to be available this summer in Quebec. Working closely with NVIDIA and Hewlett Packard Enterprise, TELUS is implementing a fully compliant NCP Reference Architecture and software stack to ensure optimal performance and security.

Cassava Technologies: Pan-African AI infrastructure

Cassava Technologies is pioneering the first pan-African NVIDIA Cloud Partner (NCP) initiative, beginning with a substantial deployment of HGX systems in South Africa that are scheduled to go live by the third quarter of this year. The company is also expanding and modernizing its data centers to extend AI factory capabilities to Egypt, Kenya, Morocco, and Nigeria in future phases. As a technology conglomerate with core telecommunications operations, Cassava is uniquely positioned to interconnect AI factories with business infrastructure across Africa. The company is collaborating with African governments to drive national AI development initiatives, including:

• Infusing AI use cases in the African enterprise market
• Inception programs for regional startups
• Upskilling through NVIDIA Deep Learning Institute (DLI)
• Creation of African LLMs to empower localized services
• Development of use cases that address regional challenges in agriculture, healthcare, education, and mobile payments

Kazakhtelecom: Central Asia’s AI catalyst

Kazakhstan has launched a broad national AI initiative that encompasses the development of a national LLM and AI applications for critical industries. To support the formation of local infrastructure, Kazakhtelecom is launching a commercial AI factory powered by NVIDIA GPUs and NVIDIA AI Enterprise. It’s scheduled to launch this summer. The platform will be delivered as a cloud service, ensuring accessible and secure AI infrastructure for public institutions, private businesses, startups, and universities across Kazakhstan. This will empower key sectors with advanced AI models and applications, including airlines, banking, construction, mining, and oil & gas.

Diverse uses drive telco AI factory growth

As telecom-powered AI factories mature, a few major categories of users and use cases are emerging. Most providers span several of these areas as they scale:

Users	Example AI workloads driving demand
Nations	• Development of national LLMs that reflect local languages and cultural contexts AI use-cases for critical sector across healthcare, agriculture, and education • AI-powered citizen services that streamline government interactions and judicial processes • Support for AI education across schools, universities, research institutes, and other organizations • Support for AI innovation sandboxes, tools, and software to promote an open ecosystem
Enterprises	• Enable enterprises to build their own specific models • Integrations of agentic AI and reasoning capabilities into enterprise workflows • Healthcare, manufacturing, financial services, and retail verticals are prominent tenants that bring a wide variety of AI workloads
Model builders	• Startups creating new AI models • Public or private institutions doing fundamental research
Internal telcos	• Customer service, network operations (BSS/OSS) workloads, accelerated with generative AI/agentic AI, such as digital humans, field operations AI agents, etc. • Authorized LLMs for employee use, internal chatbots, and enterprise IT workloads Centralized network workloads

After their initial capacity buildouts, many telco AI factory providers are already investing in expansions because of growing demand. This proven growth vector has led numerous telecom companies to create dedicated subsidiaries or complementary brands that position them as AI service providers in addition to their key role as connectivity providers. The journey from first dialog to first deployment is often within one year, a reflection of how fast these telcos are moving to capitalize on the AI opportunity. Most of them are also NVIDIA Cloud Partners, which provides a unique advantage from time-to-value perspective, not just time-to-market.

NVIDIA NCP: the blueprint for telecom AI factories

At the core of these telco-powered AI factories is NVIDIA’s Cloud Partner program, which delivers a comprehensive reference architecture that includes:

Full-stack technology platform for optimized performance and efficiency

• NVIDIA AI enterprise software suite that provides enterprise-grade AI tools and frameworks
• Pre-trained models, application blueprints, and NIM microservices with access to over 900 SDKs and AI models to accelerate development
• Streamlined workflows for AI training and inference at scale
• Comprehensive security features for multi-tenant environments

The power of NVIDIA’s ecosystem to support demand generation

• Large partner ecosystem including global system integrators, independent software vendors, application providers, and data platforms
• Access to 6 million developers across 250 countries
• Connection to 27,000 NVIDIA Inception program startups
• Marketing support to build awareness and go-to-market partnership opportunities

Education and enablement for upskilling

• NVIDIA DLI training programs to build domestic AI talent
• Specialized curriculums for different stakeholders, from students to developers
• Hackathons to stimulate local innovation

This comprehensive approach enables telcos to quickly establish themselves as AI cloud providers while ensuring interoperability, performance, and cost efficiency.

What’s next: Scale up with AI factories, scale out with AI-RAN

While much of the initial focus for AI factories has been on training large models, the future points toward two key developments that align with telecom providers’ strategic advantages:

• The inference opportunity – As foundation models reach widespread deployment, the demand for inference infrastructure is exploding. Telecom companies, with their distributed edge data centers, are perfectly positioned to create inference engines closer to end users, reducing latency and improving performance for AI applications.
• Convergence of RAN and AI – AI-RAN has emerged as the next evolution for wireless networks. It enables AI and 5G workloads to run on a common accelerated computing infrastructure, while also improving wireless networks performance and efficiency with AI. This means the radio access network can now double up as an AI inference hub.

This presents the next big opportunity for telcos, and something only they are positioned to do—build an AI grid at national scale. An AI grid leverages centralized AI factories for compute-intensive AI workloads such as training and reasoning, along with distributed AI-RAN data centers for edge AI workloads, including inferencing. NVIDIA provides a cohesive reference architecture for AI factories and AI-RAN, in addition to capabilities like NVIDIA Cloud Functions, a serverless API that can seamlessly orchestrate demand and supply across the AI factory and AI-RAN infrastructures.

With the unprecedented demand for AI, telcos can stay ahead of the curve by building an NVIDIA-powered AI grid that can scale-up and scale-out, bringing AI as close as possible to the end users. SoftBank Corp. and IOH are trailblazing the path to AI grids by building AI factories and AI-RAN, and creating new AI applications in collaboration with NVIDIA and partners. Collectively, they are building a RAN-Ready AI infrastructure to create, distribute and consume intelligence across their nations.

By leveraging NVIDIA’s NCP reference architecture and ecosystem, telecom providers are becoming connectivity and AI service providers, creating a virtuous cycle. As more domestic AI applications emerge, demand for AI infrastructure grows; as infrastructure expands, more developers and enterprises can participate in the AI economy. Through this cycle, telecom companies are not just building data centers—they’re building the foundations for inclusive, sovereign AI economies and turning the flywheel for the intelligence revolution in every country.

Explore NVIDIA-accelerated sovereign AI factories and learn more from industry leaders:

• GTC Paris Panel: How Telcos are Powering Sovereign AI for Europe (Orange, Swisscom, Telenor, NVIDIA)
• GTC Panel: Accelerating Sovereign AI Factories: Insights from Telco Case Studies (SoftBank Corp., Telenor, TELUS)
• GTC Panel: How Indonesia Delivered a Telco-led Sovereign AI Platform for 270M Users (Accenture, Hippocratic AI, IOH)
• https://developer.nvidia.com/blog/telcos-across-five-continents-are-building-nvidia-powered-sovereign-ai-infrastructure/?linkId=100000367809641