African Compute Initiative's inception workshop advances shared vision for AI infrastructure in Africa

The African Compute Initiative (ACI), led by Associate Professor Jonathan Shock, convened its first project workshop in Cape Town on 7–8 May 2026, bringing together researchers, AI practitioners, infrastructure experts, funders and institutional partners to help shape the future of shared compute infrastructure for African AI research.

Co-organised by the Mozilla Foundation and the University of Cape Town (UCT), the workshop marked an important early milestone in the ACI project. Rather than serving as a formal launch, it was designed as a collaborative working session to test assumptions, surface priorities and co-design the technical, governance and ecosystem foundations that will guide the initiative as it develops.

Over two days, participants explored a central question: what kind of compute ecosystem does Africa need to support locally relevant, high-impact AI research? Discussions examined both the supply side of compute infrastructure – including architecture, governance, optimisation, sustainability and storage – and the demand side, focusing on the kinds of research currently constrained by limited access to compute across the continent.

For UCT, the workshop also reflected a broader institutional commitment under Vision 2030: to produce research that responds to Africa's challenges, to expand digitally enabled research and learning environments, and to unleash human potential in pursuit of a fair and just society. In this sense, the workshop was not only about infrastructure planning, but about how African universities and their partners can create the conditions for research that is more equitable, collaborative and responsive to the needs of the continent.

A workshop focused on building the right infrastructure

One of the clearest messages to emerge from the workshop was that the challenge is not simply one of hardware procurement. Participants stressed that meaningful compute access depends on a wider ecosystem of storage, technical support, training, governance, data infrastructure and partnerships.

The workshop therefore focused not only on what kind of compute cluster ACI should build, but on what it would take to make that infrastructure usable, sustainable and valuable to researchers across different institutional and national contexts. Questions of storage, user support, access models, operational capacity and long-term maintenance were treated as central to the design process, rather than secondary concerns.

This practical approach was reflected in discussions about infrastructure design. Participants argued that ACI should prioritise a system that can be deployed efficiently, supported reliably and used productively from the outset, rather than pursuing technical complexity for its own sake.

UCT's High Performance Computing (HPC) cluster served as a practical reference point in discussions, illustrating how a working research system combines processing power, GPU acceleration and shared storage to support scientific workloads across disciplines. For participants, it offered a useful model of what integrated, well-supported compute infrastructure looks like in practice.

Professor Rob Simmonds, systems administrator at ilifu Research Cloud, noted during the workshop:

"The largest depreciation in a cluster comes in the first 18 months… You absolutely have to have the system running as quickly as possible and be reliable. That's much more important than maximum performance."

That perspective captured an important theme of the workshop: that building African AI infrastructure is not only about acquiring advanced equipment, but about creating systems that work in practice – systems that are robust, accessible and responsive to the research environments in which they will operate.

Surfacing the research demand behind the infrastructure

The workshop also created space to examine the demand side of the equation: what kinds of research could be advanced if African researchers had more meaningful access to compute?

Participants described a wide range of projects currently constrained by limited infrastructure, including work in African language technologies, health and biomedical research, climate and weather forecasting, and other public-interest applications. These examples highlighted the extent to which access to compute is shaping not only the pace of research, but also the kinds of questions researchers are able to pursue.

In this respect, the workshop aligned strongly with UCT's Vision 2030 commitment to research that addresses the complex problems facing Africa and the world. It reinforced the idea that infrastructure is not an end in itself, but an enabler of research that can contribute to health, education, language inclusion, environmental resilience and other urgent societal challenges.

Governance, fairness and the question of access

Alongside technical design, the workshop highlighted governance as one of the most important questions facing the initiative. Shared compute is a scarce and valuable resource, which means that decisions about access, allocation and priority-setting will shape who benefits from the infrastructure and under what conditions.

Participants reflected on how ACI might balance different ideas of fairness, efficiency, sustainability and public value, particularly in a context where institutions and researchers across the continent face unequal levels of infrastructure access and support.

Dr Georgina Rakotonirainy, lecturer in the Department of Statistical Sciences, UCT, observed:

"Fairness is not universally defined. It must be defined. The big question is, what is fairness? Proportional fairness, equal resource share. Temporal fairness, equal waiting time. Institutional fairness, equal access across institutions. Social fairness, where we want to accommodate underrepresented communities. They can conflict with one another."

This question of fairness sits at the heart of the initiative. It speaks not only to how compute resources will be distributed, but also to how ACI can contribute to a more inclusive and equitable AI research ecosystem on the continent. For a project hosted within a university committed to transformation, excellence and sustainability, these governance questions are inseparable from the technical design of the infrastructure itself.

ACI and UCT's vision for African research

As the first workshop of the ACI project, the Cape Town gathering created a shared space for experts across Africa to reflect on what a fair, effective and sustainable compute initiative should look like in practice. It brought together the technical realities of infrastructure design with a wider set of questions about access, research priorities, skills development and collaboration across institutions and borders.

For UCT, this work resonates strongly with Vision 2030's ambition to be a university that produces research for impact, supports innovation in teaching and research, and contributes to solving Africa's problems through collaboration and knowledge production. By convening this workshop, UCT and its partners helped open an important conversation about the kind of infrastructure needed to support African AI research on African terms.

The workshop laid the groundwork for the next phase of the African Compute Initiative: refining the technical architecture, strengthening governance models, understanding the scale and nature of demand, and building the partnerships required to support shared AI infrastructure across the continent.

As ACI moves forward under the leadership of Associate Professor Jonathan Shock, the questions raised in Cape Town will remain central: how to build infrastructure that is not only technically strong, but also usable, fair, collaborative and grounded in African research priorities. In that sense, the workshop was more than a planning exercise. It was an early step in shaping the conditions for a stronger and more inclusive African AI research ecosystem.