The Missing Piece in Africa’s AI Blueprint: The Computing Conundrum

The African Union Development Agency (AUDA–NEPAD) recently unveiled a draft artificial intelligence (AI) white paper and policy roadmap outlining the continent’s AI ambitions and strategy. The roadmap sets out six strategic pillars for AI development on the continent: (i) developing human capital in AI; (ii) strengthening AI infrastructure and data management; (iii) creating an enabling environment for AI deployment and development; (iv) establishing a conducive economic climate for AI; (v) building sustainable partnerships; and (vi) establishing the capacity to monitor and evaluate AI strategies in African countries.

To propel Africa’s AI ambitions, the roadmap further lays out several bold recommendations for action, including allocating US$300 million towards the creation of two major initiatives: a US$100 million African Union AI grant and a US$200 million African Union AI investment fund. The funding would be mainly earmarked for supporting early-stage AI start-ups on the continent and AI research.

While the funding allocation and purposes are indeed worthwhile, the roadmap is missing a key component that is needed to unlock the AI revolution: support for investment in critical AI infrastructure, particularly investments in computing power, the engine that powers AI. Africa’s AI infrastructure investment drive needs to be inclusive, so that last-mile communities and underserved populations benefit equally.

The AI Infrastructure encompasses servers, data centers, cloud services, reliable internet access, and consistent electricity supply, all of which are crucial for powering, managing, and scaling AI systems. The continent has a dearth of powerful computers needed to train AI models. Even with access to funding for AI start-ups on the continent, if they don’t have access to computing power, their capacity to innovate AI solutions becomes highly limited. Access to these high-powered computers allows innovators and researchers to study and model complex development and economic problems that would otherwise be challenging to solve.

The value of investing in the AI revolution

Building the critical AI infrastructure is key to unlocking start-ups’ innovation potential. During a recent AI event in Ghana convened to discuss AI policy developments in Africa, two professors highlighted the challenges aspiring AI professionals face at their universities. One professor narrated how the entire department, with more than 100 students and lecturers, shares just 15 AI cloud computing accounts generously provided by Google to the university. The other professor pointed out the necessity for students to run AI models on their computers overnight before the models can execute the actions, with any interruptions in electricity or connectivity jeopardizing the whole process. These challenges highlight examples from relatively well-funded universities in Ghana. In less-developed ecosystems, universities face even greater constraints due to poor funding and infrastructure.

Currently, Africa boasts only a handful of supercomputers. Toubkal, located in Morocco, is the most powerful supercomputer on the continent, having secured a place in the Top 500 list of supercomputers in the world. However, this is not enough computing power to serve the needs of 55 countries on the continent. A recent report published by the Tony Blair Institute for Global Change highlights that even though investment in data centers in Africa has grown by 13 percent more than the global average, cloud computing adoption in Africa remains at 15 percent, significantly lagging behind Europe’s 71 percent penetration rate.

Adding to this challenge is the lack of clear and harmonized data governance strategies and policies in different countries across the continent, despite an Africa-wide data policy framework. This regulatory uncertainty has resulted in data localization policies in some countries that restrict the use of cloud computing beyond national borders or promote on-premises data hosting.

Estimates show the AI revolution could increase Africa’s economy by US$1.5 trillion—about 50 percent of the continent’s current GDP—by 2030. To realize this potential, more investment is needed. Yet, the infrastructure needed to support AI is expensive. Nvidia recently announced plans to build a US$200 million artificial intelligence center in Indonesia in collaboration with a local telecommunications company Indosat Ooredoo Hutchison.

The Indonesian investment highlights the substantial costs associated with constructing state-of-the-art infrastructure tailored for AI development and research—and the related risk of such high costs excluding local communities from participating in and benefiting from this innovation. Interestingly, this figure nearly matches the entire amount earmarked for AI investment in the draft AU-NEPAD roadmap.

Building out Africa’s computational infrastructure—fairly and equitably

Democratizing access to AI infrastructure for computational purposes is needed to accelerate AI research, development, and adoption. This means making AI infrastructure affordable and available to innovators and researchers irrespective of their location. It also means creating tailored solutions for vulnerable and excluded groups to have equitable access to AI infrastructure to solve their unique social and economic challenges.

Following are four key recommendations on how the AU, the private sector, and academia can advance investment in computational infrastructure to support the AU artificial intelligence roadmap, and how AU member state governments can develop policies that incentivize that investment.

1. Establish a shared fund for open computing access.

The AU, in collaboration with its member states as well as global development partners, should create a common pool of funding for open and accessible AI computing infrastructure. This could be similar to the Europe Joint HPC initiative, which is a fund that is dedicated to investing in Europe’s Supercomputers.

This fund should focus on making computing power accessible to researchers and innovators in the continent. The access to computing power would enable researchers, innovators, and other ecosystem players at national and regional levels to utilize AI computing power that would have otherwise been out of reach for them. Consequently, that would not only democratize access to computing but also broaden the range of innovation and ideas coming from different countries and regions on the continent.

2. Encourage an AI computing credit program.

To further democratize access to computing capacity, the strategy should encourage countries to develop AI credit schemes that support budding researchers and innovators with subsidized or free computing resources. Universities could serve as an entry point for running these credit schemes. A similar program currently runs at the Stanford Human Centred Artificial Intelligence. With such a system, researchers could local and international computing resources for educational, model testing, and deployment purposes before they are commercialized. A strong partnership with global computing resource partners to run such AI computing credit schemes will help complement local capacity.

3. Establish integrated hubs for AI development.

To overcome some of the hurdles of limited computing capacity, poor connectivity, and unreliable electricity faced by AI aspirants on the continent, the AU strategy should propose governments establish multi-country integrated hubs for AI development. Countries should be encouraged to set up dedicated regional AI hubs and zones with incentives for AI-related infrastructure providers to co-locate.

To create a conducive environment for AI research and development, the hubs should have access to reliable power supply, high-speed internet, and other computing services to support both public and private usage. The United Nations Development Programme together with the G7 are pushing for the democratization of cloud computing across ecosystems for AI hubs.

The AU and African governments could develop incentives to support these initiatives for AI hubs, which in turn will support key provisions of the African Union Digital Transformation Strategy and the African Continental Free Trade Area while reducing set-up costs and ensuring equitable access to data and AI computing resources. Further, innovators and researchers can have access to shared computing resources within the hubs for short-term or long-term AI project development.

4. Upgrade and optimize existing cloud data center resources.

With limited immediate funding for cloud data infrastructure, Africa needs to make the most of its existing resources to lay the groundwork for AI advancement. As of 2023, the continent has more than 100 data centers, with the majority located in South Africa, Nigeria, Kenya, Mauritius, and Angola. Combined, these countries host about 76 of these facilities.

The AU strategy should call for strategic collaboration and partnership between public sector entities and private sector actors such as cloud and data center operators, to focus on upgrading these existing resources to offer various types of cloud infrastructure for AI computational services. This approach of optimizing current resources would significantly lower the costs of computing power, making it accessible and affordable to AI innovators including students, start-ups, and researchers.

Implementing this targeted investment approach would transform existing cloud infrastructure to meet basic AI computing demands and help pave the way for the creation of faster and more powerful supercomputers for future AI initiatives.

Going beyond the AU roadmap

The AU’s AI roadmap is a pivotal opportunity for Africa to harness the transformative potential of artificial intelligence for development. However, it is imperative to prioritize the development of robust AI infrastructure as a foundational step towards realizing this vision. The race to establish advanced AI cloud infrastructure is well underway in other regions, and Africa must not be left behind.

Despite challenges, stakeholders need to forge strategic partnerships with tech companies aimed at making AI infrastructure accessible and affordable across the continent. Moreover, investments in digital infrastructure should focus on enhancing computing capacity and optimizing existing resources to meet the growing demands of AI applications. Policymakers must reconsider data localization policies to facilitate the establishment of shared cloud infrastructure and enable seamless data mobility across national borders.

Ultimately, meeting the AI infrastructure needs of Africa will require innovative approaches that extend beyond traditional grants and investment funds outlined in the AUDA-NEPAD roadmap. By prioritizing AI infrastructure development, Africa can lay the groundwork for unlocking the full potential of artificial intelligence to drive inclusive growth and sustainable development across the continent.