Sovereign AI: Why Governments Are Building Their Own Models

The release of GPT-4 in March 2023 triggered a global conversation about AI capability that has since evolved into a conversation about AI sovereignty. Governments across the world — from the UAE to France, from India to Singapore — have concluded that dependence on a small number of American AI companies for the foundational models that will power their digital economies is a strategic vulnerability. The response has been a wave of national AI initiatives, each with its own model development programme, compute infrastructure investment, and regulatory framework.

The UAE's Technology Innovation Institute released Falcon — a family of open-source language models that briefly topped the Hugging Face leaderboard — as an explicit expression of the country's ambition to be a global AI hub rather than merely a consumer of AI services developed elsewhere. France backed Mistral AI, which has produced a series of highly capable models that have challenged the assumption that frontier AI requires the resources of a US hyperscaler. India's IndiaAI Mission is funding the development of models trained on Indian languages and cultural contexts, with the explicit goal of ensuring that AI serves the needs of India's population rather than being optimised for English-speaking Western users.

For investors, sovereign AI creates demand for a specific type of infrastructure: national compute clusters, often built with government support, that are designed to train and run models for domestic use cases. These facilities are typically built in partnership with private operators and financed through a combination of government grants, sovereign wealth fund investment, and private capital. The return profile is attractive — long-term contracts with creditworthy government counterparties — but the investment requires navigating complex procurement processes and political relationships.

At Lensbridge, we see sovereign AI as one of the most significant drivers of digital infrastructure investment in Asia Pacific over the next decade. The governments of Singapore, Malaysia, Indonesia, India, Japan, and South Korea are all investing in national AI infrastructure, and the private capital that supports those investments will be well-positioned to benefit from the secular growth of AI adoption in the region.