Pakistan's National Telecommunication Corporation is developing sovereign artificial-intelligence infrastructure that will host advanced AI models and high-performance computing within state-managed data centres, seeking to reduce government exposure to public clouds while keeping sensitive public-sector data in a locally controlled environment.

NTC plans to provide Graphics Processing Unit as a Service (GPUaaS) and Model as a Service (MaaS) through its secure data centres, expanding the computing infrastructure available to designated government clients. The corporation, which serves as the foundation of the country's e-governance infrastructure, said the initiative is intended to meet the state's growing computational requirements while providing secure, high-speed and local alternatives to public-cloud services.

At the centre of the project is a three-year contract between NTC and DataVault for the deployment of enterprise-grade, NVIDIA-licensed graphics processing hardware. The planned infrastructure will include NVIDIA H200, A100 and B200 series GPUs, providing the computing resources required for AI inference and training. NTC intends to use the hardware to establish a dedicated platform capable of securely hosting and operating large language models within its own managed environment. The models listed for the platform include Llama 3.1 8B, 3.3 70B and 4 Maverick, GPT-OSS 20B, Qwen3 32B and Kimi K2. Text-to-Speech services are also expected to be supported. The arrangement is designed around the principle of sovereign AI: retaining computing capacity, model deployment and sensitive government information within locally managed infrastructure rather than depending on public clouds.

NTC Managing Director Major General (R) Ali Farhan said computing power had become the backbone of modern digital infrastructure. He said acquiring high-performance NVIDIA GPUs and hosting advanced AI models locally represented a step towards protecting Pakistan's digital sovereignty. The self-hosted environment is intended to minimise government clients' exposure to public-cloud infrastructure while providing AI inference and training capabilities. That combination of computing hardware and locally hosted models represents the principal technological shift envisaged by the project.

Instead of relying solely on external cloud platforms for computationally intensive AI workloads, designated government clients would be able to use resources hosted within NTC's state-managed data centres. The GPUaaS component is intended to provide access to high-performance processing capacity, while MaaS would make AI models available through the same sovereign infrastructure.

DataVault will serve as the implementation partner under the three-year agreement and work alongside NTC on deployment and performance. The partners are targeting a monthly service-level availability of 99.9%, supported by round-the-clock technical assistance throughout the year. The project also reflects the growing computational requirements created by large language models, which require substantial processing resources for both training and inference. NTC's planned use of H200, A100 and B200 hardware is intended to provide the processing capacity needed to operate such workloads within its managed environment.

The initiative is particularly focused on the public sector. NTC said its infrastructure would provide designated government clients with AI capabilities while maintaining the required safeguards for sensitive information. The emphasis on localisation means the project is as much about control over data and computing infrastructure as it is about access to artificial intelligence. By hosting GPUs and AI models within its own data centres, NTC intends to provide a state-managed alternative to processing sensitive public-sector workloads through public clouds.

The three-year DataVault contract provides the hardware and implementation framework for that ambition. Its success will depend on delivering the promised enterprise-grade performance, including the targeted 99.9% monthly uptime and continuous technical support. For NTC, the project marks an expansion of its role in Pakistan's e-governance infrastructure into sovereign AI computing. The corporation is seeking to combine locally hosted high-performance GPUs, large language models and AI services in a single environment designed specifically around government requirements and control of sensitive public-sector data.