Raw data, refined elsewhere

The Infrastructure gap threatening Uganda’s AI Independence

Kampala, Uganda | IAN KATUSIIME |  For Uganda, a nation with a young, tech-hungry population and a budding startup ecosystem, the race to innovate is hitting a physical wall: a critical shortage of Tier III data centres. While the world debates the ethics of Large Language Models, Uganda faces a more fundamental question of sovereignty.

Without the local “gray brick” infrastructure to store and process its own data, analysts say the country risks becoming a mere tenant in the global AI economy exporting its raw data to foreign servers and importing finished AI tools at a premium.

Currently there are about four data centers in the country; Raxio Data Centre in Namanve, a few kilometres outside Kampala, the MTN data centre, and another by the Ugandan government.

Michael Niyitegeka, Director of Refractory, a software skilling academy shaping the next generation of AI developers, says Uganda has not exhausted the capacity of the existing data centres.

Speaking to the The Independent in an exclusive interview on how the current shortage of data centers and local compute infrastructure affect the ability of the learners at Refractory to build, test, and deploy real-world AI and data-intensive applications, Niyitegeka said “The utility of these data centres is certainly still touch and go.”

“From a training perspective, we need data that is available for training because most of this data is proprietary.” He says it is critical to have public domain data at a national level.

The good news, Niyitegeka says, is that  the Uganda Communications Commision is setting up a training hub at Nakawa that will serve as a test bed  for local developers.

“The essence of a test bed is to test applications, but in their case, they only have the hardware. The data I need to determine whether my model works isn’t readily available,” he said.

Niyitegeka says there  is a need to train on data that is appropriate to Uganda. “We run an AI cohort, and the data available was primarily European. There was someone who built an AI model for pothole detection, and I looked at it and wondered whether it would actually detect a pothole here.”

He stresses the importance of an environment where data can be collected for purposes of learning and training.

“When I go to health for example, all the radiography images that have been taken are digitized and can be purposed for training. This data can be anonymised and accessed at such and such a point.”

According to Niyitegeka it is now more critical because of new regulations requiring organisations to have in-country data resources in standardised secure environments.

Data access very significant

For Uganda’s AI ambitions to get somewhere, ready access to data is what will make a difference and that includes organisations like Refractory training technologists.

“If I am going to train students on building data models, there should be access to good data. Uganda Revenue Authority should make a volume of data available for people interested in tax assessments so that one does not need to scratch to use real data pieces,” he explains.

AI engines are hungry for data and now the Ugandan government has a huge task on its hands. “It can’t be up to Refractory. We need to have a national effort,” Niyitegeka said.

On the collaborative role that the government, private sector, and training institutions play in building AI infrastructure, Niyitegeka says ultimately the buck will have to stop with the government.

“We have to start somewhere. For instance, the Ministry of Health is collecting data, but is it ready for this transition? It’s a combination of advocacy and hands-on work,” he said.

He adds that the leadership role of the government is in building the data infrastructure that can quickly model the agriculture sector as an example but also trickle down to a specific level.

“If you go to Mbarara, what data can you find at the district? As we train and skill, these other aspects need to be in sync.”

He says some of the trainees can explore opportunities outside Uganda because of the data they have been exposed to.

‘Gov’t is the main player’

A lot of work remains to be done but he parts with one reminder: “The government has to rally the key players. Once the opportunities are visible, the MTNs of this world will unlock the data centres. It doesn’t take long to dedicate a rack in their data centre for training. We can use their data centres but you have  a need.”

Uganda has a limited number of data centers and that comes with operational and cybersecurity risks associated with a thin local data center ecosystem, particularly as data traffic, fintech, and AI workloads continue to grow. Naturally it falls to telecoms to plug that gap.

“We draw on the excellent expertise of our African and Global reach to deploy world class solutions for cyber security,” David Birungi, the Public Relations Lead at Airtel Uganda told The Independent in an interview.

“At an enterprise level, we train our human resource across the group to ensure we keep ahead of the industry and are in sync with global developments,” he adds.

Airtel is the second most profitable company in Uganda; given its scale and volumes, some analysts say there should be  more aggressive private-sector investment in hyperscale-grade data centers in Uganda. More voices are calling for such companies to co-invest in expanding the country’s data center capacity.

 “Uganda is at a different point on the ICT journey in stride with global ICT businesses that include the hyperscale-grade data centers. The outlook is promising but there are things we need to set in motion as a country. Both at a policy, finance, and regulatory level,” Birungi says.

He stated that Airtel Uganda is well positioned to support development that does not replicate the limited resources to deliver the services.

“We are open to reciprocal partnerships as we have done with MTN, Starlink and others on infrastructure sharing, who might show interest,” he said.

In October 2025, there were reports of a new data centre in Uganda by Synectic Technologies to boost Uganda’s AI transition. On its website, Synectic Technologies says it is developing “The Aeonian Project” which it calls an off-grid green energy regional Hyperscale Data and High-Performance Computing Centre (DHPC) powered by 100MW of excess pre-transmission electricity from the government owned 600MW Karuma Hydro Power Plant (K-HPP) in Uganda.

Synectic says the DHPC includes a 30MW capacity Artificial Intelligence facility which will be trained to autonomously manage the DHPC amongst other capabilities.

“Data transmission is through leased two (2) core dark fibre optic cables to link the facility with the company head office in Kampala (Uganda) for remote monitoring and to connect to the submarine cable in Kenya for connectivity to the rest of the world,” a company statement says and adds that the alternative arm of the cabling will provide data transmission to and from neighbouring land-locked countries.

The proposed service provision is in Artificial Intelligence, Colocation, Blockchain, Data Storage and On-Demand Cloud Computing. The Center will also provide Disaster Recovery Services (DRaaS) to existing data centres within the region.

National Data Centre

The National Information Technology Authority Uganda (NITA-U) says its National Data Centre offers a comprehensive suite of services tailored to meet the diverse needs of government entities and other stakeholders.

The body that was set up in 2009 to coordinate and regulate IT services in Uganda says through its data centre it is aimed at enhancing operational efficiency, data security, and supporting the delivery of critical public services.

NITA-U says the data centre is designed to meet the highest international standards for data centre operations with redundant power supply through dual diesel generators and 150 Dual KVA modular UPS systems providing uninterrupted power supply.

It also contains advanced cooling systems with precision air conditioning to maintain optimal operating temperature. NITA-U says low-latency links to key cities in East Africa facilitate efficient data exchange and collaboration across the region.

The thin data eco system has been a source of frustration among AI developers in Uganda and on the African continent. At the inaugural Global AI summit for Africa held in Kigali, Rwanda last year, Kate Kallot, CEO of AminiTech, a Nairobi-based AI and data firm, said African innovation was being locked outside of the continent.

Kallot decried compute scarcity in the Global South. “Today it costs us more to import compute in Africa than it is to buy compute in the US. How do you expect us to continue fueling our research and training models that are adapted to our context locally if we have to send all our data outside?” she asked.

100GW by 2030

With the handful of data centres, Uganda could be left behind as the global AI revolution gains momentum. In its 2026 Global Data Centre Outlook, JLL, a US investment management firm, says that nearly 100 GW of new data centers will be added between 2026 and 2030, doubling global capacity.

“The global data center sector will likely expand at a 14% CAGR through 2030, which will require energy innovations to alleviate grid constraints. Hyperscalers will remain a key driver of sector growth, executing a dual strategy of leasing and self-building,” the firm said.

By 2030, the report states that AI could represent half of all workloads with inference becoming the primary driver.

“AI only represented about a quarter of all data center workloads in 2025, with training driving most of the demand adding that a significant shift is anticipated in 2027, when inference workloads could overtake training as the dominant AI requirement.

“The sector is experiencing an infrastructure investment supercycle requiring up to $3 trillion by 2030.”

****

This story was done with support from ​​Collaboration on International ICT Policy for East and Southern Africa (CIPESA)