Uganda is becoming more serious about artificial intelligence. But as the country moves towards a national AI strategy, a harder question is emerging: what will it actually cost to control a meaningful part of its AI future?
That question comes as Uganda implements its five-year National Data Strategy, backed by a projected Shs1.663 trillion investment. The strategy is intended to improve how government data is collected, managed, protected and shared, while addressing fragmented systems and duplication across public institutions.
For an AI economy, that foundation matters. Data is the raw material of AI. But data alone does not create AI sovereignty.
Sovereignty means having meaningful control over the infrastructure, data, skills and systems that determine how AI is developed and deployed. That control comes with a cost, and Uganda now has to decide which parts of the AI stack it can realistically afford to control.
The recent debate around the National Data Strategy shows why this will require more than money.
The Uganda Bureau of Statistics has raised concerns that parts of the strategy could overlap with its statutory mandate to coordinate Uganda’s National Statistical System. At the same time, UBOS is implementing its own Fourth Plan for National Statistical Development for 2025/26–2029/30, which includes modernising statistical production and expanding the use of Big Data, Machine Learning and AI.
The disagreement is important because it exposes a larger issue. Uganda cannot build a coherent data and AI economy if institutions develop overlapping systems, standards and responsibilities.
Organisations cannot afford to work in silos. The institutions responsible for statistics, digital infrastructure, planning, research, regulation and technology development hold different pieces of the same ecosystem. The challenge is to make those pieces work together, with clear mandates and common standards, rather than creating parallel systems around the same national resource.
That becomes even more important when the discussion moves from data to computing. A country can store sensitive information locally while still depending on foreign companies for the chips, cloud infrastructure, AI models and expertise needed to process it.
That is localisation. It is not necessarily sovereignty.
Modern AI requires specialised computing infrastructure, particularly GPUs. Uganda therefore has to decide whether it wants to own some of this capacity, lease it, subsidise access or rely mainly on international cloud providers.
Uganda has already taken an initial step through the National AI Research Cloud established by Makerere University and the Research and Education Network for Uganda. It provides high-performance computing and secure storage for researchers, students and innovators, helping reduce dependence on expensive international cloud services.
But research computing is different from the infrastructure required to support large-scale national AI workloads.
Then comes electricity. AI data centres require reliable and affordable power. Uganda’s ambition to position itself as an AI hub will therefore depend as much on energy infrastructure as on computing hardware.
President Yoweri Museveni’s recent discussions with technology investor Mehamood Hosein around positioning Uganda as an AI hub have included energy, computing infrastructure, research and education.
The economics will matter just as much as the ambition. A sovereign AI strategy cannot become a race to announce data centres without asking who will use them and how they will be sustained. Computing infrastructure is expensive and becomes outdated quickly. If there is insufficient demand from researchers, government and businesses, the investment can become a financial burden rather than an economic asset.
Uganda’s digital market provides a starting point. By the end of 2025, the country had 47.1 million active mobile subscriptions, 36.3 million active mobile-money accounts and 18.5 million mobile-internet subscriptions, according to UCC data.
Yet millions of Ugandans remain disconnected, particularly in rural and hard-to-reach areas. That creates a basic test for the country’s AI ambition. If investment stops at GPUs and data centres, Uganda could end up with sophisticated infrastructure serving a relatively small part of the economy.
The objective should instead be to connect infrastructure to actual economic and social needs: agriculture, healthcare, education, financial services, public administration, research and local enterprise.
This is where UBOS’s work becomes particularly relevant. Its plan to integrate Big Data, Machine Learning and AI into statistical production points to a more practical use of national computing and data capacity. If high-quality official statistics can be combined responsibly with other datasets, researchers and businesses could develop systems that better understand Ugandan markets, languages and local conditions.
That is where sovereignty begins to have economic value.
Rwanda provides one African example of a targeted approach. Its AI strategy distinguishes between sovereign computing, commercial compute, data centres and affordable access, with a target of at least 24–48 sovereign GPUs by 2028.
India has pursued the same broader principle at a much larger scale. Its IndiaAI Mission was approved with a budget of 10,371.92 crore rupees over five years, targeting at least 10,000 GPUs alongside programmes for indigenous foundation models, datasets, skills and AI startups.
Uganda cannot copy India’s scale. But it can learn from the structure.
AI sovereignty is an ecosystem, not a server room.
There is also no need for Uganda to pursue technological isolation. It does not have to manufacture every processor or build its own version of ChatGPT. It can use foreign foundation models while developing Ugandan datasets. It can rent international computing capacity while maintaining domestic infrastructure for sensitive workloads. It can partner with global technology companies while protecting strategic data and intellectual property.
It can also support open-source models and adapt them to Ugandan languages and local needs.
The objective should be strategic control, not isolation. That makes the Shs1.663 trillion National Data Strategy an important starting point, but not the end of the conversation.
The investment includes funding for data infrastructure, governance, research, innovation and strategic data use. Its success, however, should ultimately be measured by what those investments enable.
Can Ugandan researchers access the infrastructure? Can businesses build products from it? Can government use data to make better decisions? Can AI systems understand local languages and contexts? Can intellectual property created from Ugandan data remain in Uganda? And can the institutions responsible for this ecosystem work together?
The last question may prove the most important. Uganda’s AI future will not be built by one ministry, one university, one regulator or one technology company. The country’s data institutions, technology agencies, researchers and businesses need to operate as parts of one ecosystem.
That does not mean eliminating institutional mandates. It means making those mandates complementary. Uganda does not have to control every layer of the AI stack. But it should identify the layers it cannot afford to surrender and put a realistic price on controlling them.
The ultimate measure of the National Data Strategy should therefore be more than the amount of data stored or infrastructure installed.
It should be whether that investment produces Ugandan research, Ugandan companies, Ugandan intellectual property and AI systems that solve Ugandan problems.
Otherwise, Uganda risks spending billions building the foundations of an AI economy that creates most of its value somewhere else.
If Uganda wants AI sovereignty, the first decision is not which model to build. It is deciding what sovereignty is worth, what it is willing to pay for, and whether the institutions building it are prepared to do so together.
