Zahanat AI has announced a partnership with Google Cloud to develop and train large language models (LLMs) in Urdu, Punjabi, and other regional languages of Pakistan. This collaboration marks a significant step toward making artificial intelligence more inclusive and reflective of how people in Pakistan communicate, both in written and spoken form. What makes this initiative particularly transformative is that the full development cycle of these models is now being carried out locally, hosted within DataVault, Pakistan’s first AI-native data center.
Previously, most LLM training for applications in Pakistan was performed on international servers, which introduced high costs, slower processing times, and limited control over data sovereignty. Zahanat AI’s partnership with Google Cloud changes that framework by enabling language model development to take place entirely within Pakistan. This allows organizations in sectors such as education, government, and healthcare to deploy AI tools more efficiently, securely, and at a lower cost. The ability to train LLMs domestically also ensures sensitive data stays within national borders, supporting compliance with local regulations and enhancing trust among users.
The infrastructure supporting this project is powered by DataVault, a dedicated AI cloud facility purpose-built for deep learning and model deployment. By running a sovereign AI cloud, DataVault guarantees that all models, training data, and computational resources remain under national control. This approach not only helps meet Pakistan’s data protection standards but also empowers local development teams to operate with greater independence and scalability.
Instead of investing in costly hardware, Zahanat leverages GPU as a service from DataVault. This model provides access to high-performance computing resources required for LLM training, instantly scalable based on project demands. The facility is designed specifically for AI use cases, offering rapid training times and efficient deployment environments, ideal for localizing and adapting models to Pakistan’s linguistic and cultural context.
The collaboration addresses a critical gap in the global AI landscape: the underrepresentation of regional languages. Global LLMs are predominantly trained in English, leaving regional languages like Urdu and Punjabi under-supported. This often results in models that struggle to understand or respond accurately in these languages. Zahanat is solving this problem by training LLMs on local text corpora, dialects, and conversational patterns, enabling AI tools that genuinely reflect the linguistic diversity of Pakistan. These tools are expected to enhance user experiences in public services, education, and healthcare, offering more intuitive and accessible solutions to everyday challenges.
DataVault also supports this work with an emphasis on security and sustainability. As a green AI data center, it incorporates low-energy systems and is working toward a solar-powered infrastructure. It also delivers robust cybersecurity through real-time monitoring and regulatory compliance, ensuring that hosted data and models are protected at all times. Furthermore, edge AI capabilities allow for real-time application of LLMs in environments such as government portals or educational platforms.
This partnership between Zahanat AI and Google Cloud, backed by the infrastructure of DataVault, represents a major advancement for Pakistan’s AI capabilities. It lays the foundation for building LLMs that not only operate locally but also understand and serve local users across a wide range of digital services.
