By: William SongAabo

Ghana has bought herself a fast car. On 24 April 2026, President John Mahama launched the National Artificial Intelligence Strategy 2025 to 2035, together with a 250 million-dollar compute centre intended to make Ghana West Africa’s hub for responsible artificial intelligence [1], [2]. The strategy projects that AI will contribute 500 billion cedis to gross domestic product by 2035 [2]. The difficulty lies not with the car but with the road. In 2024, Ghana’s network operators recorded approximately 5,600 breaks in the fibre carrying the nation’s internet traffic, an average of roughly fifteen a day [3]; in a single incident off Côte d’Ivoire, four of her five submarine cables were severed at once and left unrepaired for weeks. A strategy of this ambition cannot operate on infrastructure of this fragility. The foundations must be established first, and in a deliberate sequence.

This is not an objection to the strategy, which is a serious document: it identifies data and infrastructure as pillars and proposes a Responsible AI Authority [4]. The objection concerns sequence. Three foundations bear the load: a fibre backbone that remains in service, a firm supply of power, and public data that government can use. Each must be substantially built before AI is scaled across sectors, and before the compute centre absorbs the budget and the attention. Where the sequence is inverted, a fourth cost arises, one the strategy scarcely acknowledges: Ghana retains the data and the risk, while the value and the control reside abroad. In short, the financing and the messaging have outpaced the plan.

Consider the fibre first. Ghana’s backbone is concentrated within the Accra, Kumasi and Takoradi triangle and thins beyond it. National internet penetration stood just below 70 per cent in early 2024 [5], an aggregate that conceals a wide disparity between Greater Accra and the rural north. Where the cable is present, it remains exposed: road construction is the single largest cause of the 5,600 breaks, interrupting service to households, clinics and businesses for hours [3]. The international connection is more precarious still: on 14 March 2024, a single fault off Abidjan severed four of the five cables serving Ghana at once; most of the country lost access, and repairs continued into May [6]. The applications the strategy envisages presuppose a connection that is continuous, not merely fast. Faster internet has raised employment elsewhere in Africa [7], but only where it is sustained. A diagnostic instrument in a district hospital that cannot reach its server is not a slow instrument; it is an inoperable one.

Power is the second foundation, and its constraint is financial before it is physical. Load-shedding, or dumsor, persisted through 2024 and 2025, although dependable generating capacity already exceeds peak demand of roughly 4,000 megawatts [8]. The binding constraints are fuel supply and debt settlement, not a shortage of installed plant: energy-sector debt exceeded 3 billion dollars in early 2025 [9], and the government spent approximately 1.5 billion servicing it and restoring lapsed international guarantees [10], six times the price of the compute centre and directed toward the system on which the centre would depend. A national AI facility is a continuous load that must operate without interruption; the strategy does not specify what sustains it, or the clinics and agricultural offices dependent on its output, through a difficult week on the grid.

Data is the third foundation, the one about which the strategy is most optimistic and the one Ghana is furthest from establishing. The routine records of the state, for health, farmers, businesses and taxation, reside in disconnected systems, and the centralised electronic health record has already been discontinued once [11]. The spatial record is in poorer condition: there are no authoritative base maps, no complete digital cadastre or land register, no addressing system in consistent use, and no open geoportal [12]. Yet lands and natural resources is a designated priority sector [4], and AI cannot be developed for land administration while much of the country’s territory is unmapped and unregistered. The Lands Commission has recognised this: in 2023 it entered an 85 million dollar, five-year partnership to digitise records and construct the missing infrastructure [13]. The Data Protection Commission, charged with regulating personal data, remains underfunded, understaffed and dependent on the executive for its appointments [14]. Establishing these foundations is years of unglamorous institutional work, not an administrative formality to be completed alongside the sector deployments that depend on it.

The fourth cost is ownership. A country that adopts AI more rapidly than she develops her own data, infrastructure and expertise becomes a supplier of raw data and a purchaser of finished systems. The servers and the trained models are located abroad, and so, in effect, is the judgement those systems produce: a land claim flagged, a diagnosis proposed. Two influential analyses of the political economy of AI in the global South, often grouped as digital colonialism, set out this risk explicitly [15], [16]. It does not await the centre’s opening; it accumulates each year Ghana procures AI faster than she builds what operates it. Professor Isaac Wiafe of the University of Ghana has stated the matter plainly: AI must serve Ghana’s development rather than deepen her digital dependence [17]. Observing the sequence is how Ghana remains a producer rather than a supplier.

The strongest objections warrant considered replies. The first is that Ghana has achieved technological leaps before, from fixed lines to mobile telephony and bank branches to mobile money, and that an aspirational strategy attracts investment, so that moderating its ambition would weaken the signal to investors. Both points have merit, and neither withstands the absence of power and data. The history is also misremembered: mobile money succeeded because a network of cellular towers and more than 100,000 agents had been established beforehand [18], and a genuine leap rests on infrastructure that already exists, not on infrastructure merely promised [19]. The second and more serious objection is that the injunction to address the fundamentals first can harden into a permanent pretext for inaction. The response is to define the sequence, not abandon it: a threshold applied sector by sector, so AI is scaled in health once clinical data and connectivity can support it, in land administration once the cadastre exists, not before. The Responsible AI Authority can administer that test, and it prevents “not yet” from becoming “not ever”.

A further objection is more difficult: because a data centre takes years to construct, Ghana should begin now, while political commitment and financing remain available. Sovereign compute capacity is worth possessing. But a centre that cannot be reliably powered or reached is a stranded asset, and compute capacity without domestic data is only partial sovereignty, since the models still learn from data held elsewhere. The centre should therefore be built in parallel with the power and data commitments that determine whether it justifies its cost, not ahead of them.

The fiscal position removes the element of choice. Ghana completed her International Monetary Fund Extended Credit Facility programme in mid-2026 and moved to a non-financing arrangement, subject to a primary-surplus target and a debt anchor of 45 per cent of GDP by 2034 [20]. Little room remains, beyond debt service, for clinics, schools or the repair of fibre. Within that constraint, every cedi committed to the compute centre is a cedi withheld from fibre protection, firm power or land records. The sequence will be determined deliberately, or imposed by whichever obligation falls due first.

In practice, the sequence is a programme of work with designated owners. The Ministry of Communications hardens the backbone and secures a second, independent international link. The Ministry of Energy, with GRIDCo and the Electricity Company of Ghana, guarantees firm supply to the centre and to priority facilities. A national data programme constructs the land and administrative records and resources the Data Protection Commission. The Responsible AI Authority conducts two or three funded pilots, each with an owner and a deadline, only in sectors that satisfy the threshold. The compute centre is still built, from the resources the other three leave; it simply ceases to be the headline.

Ghana’s ambition reflects the correct instinct, and this caution is not an argument for inaction. It is an argument about sequence: fibre, power and data first, then AI scaled into each sector as it becomes ready, with the compute centre following rather than leading. A fast car on a broken road does not reach its destination sooner; it strands the traveller between towns. Once the road is built, the same vehicle that appears reckless today becomes the one that brings a farmer a planting forecast she can rely on, and a nurse in a rural clinic a diagnosis she can trust, both built and governed in Ghana rather than rented from abroad. The outcome depends on the road, not the car.

References

[1] Ghana News Agency, “Ghana to invest $250m in AI Centre, launch national strategy,” Accra, Apr. 2026. [Online]. Available: https://gna.org.gh/2026/04/ghana-to-invest-250m-in-ai-centre-launch-national-strategy/

[2] Ghana Fact, “Ghana’s National AI Strategy launched: key points you need to know,” Apr. 2026. [Online]. Available: https://ghanafact.com/ghanas-national-ai-strategy-launched-key-points-you-need-to-know/

[3] Ghana News Agency, “Ghana’s telecom sector suffers over 5,600 fibre cuts in 2024,” Accra, May 2025. [Online]. Available: https://gna.org.gh/2025/05/ghanas-telecom-sector-suffers-over-5600-fibre-cuts-in-2024/

[4] Ministry of Communication, Digital Technology and Innovations, Republic of Ghana, “Ghana National Artificial Intelligence Strategy 2025-2035,” Accra, 2025.

[5] S. Kemp, “Digital 2024: Ghana,” DataReportal, Feb. 2024. [Online]. Available: https://datareportal.com/reports/digital-2024-ghana

[6] Internet Society, “2024 West Africa submarine cable outage report,” Internet Society, Reston, VA, USA, Apr. 2024. [Online]. Available: https://www.internetsociety.org/resources/doc/2024/2024-west-africa-submarine-cable-outage-report/

[7] J. Hjort and J. Poulsen, “The arrival of fast internet and employment in Africa,” American Economic Review, vol. 109, no. 3, pp. 1032-1079, Mar. 2019.

[8] Energy Commission, Republic of Ghana, “2025 Energy Outlook for Ghana: demand and supply outlook,” Accra, Dec. 2024. [Online]. Available: https://www.energycom.gov.gh/

[9] Ghana News Agency, “Energy sector debt hits $3 billion; Energy Minister-designate calls for urgent action,” Accra, Jan. 2025. [Online]. Available: https://gna.org.gh/2025/01/energy-sector-debt-hits-3-billion-energy-minister-designate-calls-for-urgent-action/

[10] Ministry of Finance, Republic of Ghana, “Mahama administration pays US$1.470 billion to clear energy-sector debt and restore World Bank guarantee within first year,” Accra, Jan. 2026. [Online]. Available: https://www.mofep.gov.gh/news-and-events/2026-01-12/mahama-administration-pays-us1.470-billion-to-clear-energy-sector-debt-and-restore-world-bank-guarantee-within-first-year

[11] Y. A. Ofori-Adjei, “Beyond a single system: lessons from Ghana’s discontinued national EHR and the imperative for an interoperable digital health ecosystem,” Ghana Medical Journal, vol. 59, no. 4, pp. 177-180, 2025.

[12] R. Owusu Ansah, Z. Abubakari, B. Quaye, D. Todorovski, M. C. Chipofya, and C. Lemmen, “Navigating the terrain of digital transition: Ghana’s journey of developing a digital land information system,” Land, vol. 13, no. 4, art. no. 528, 2024.

[13] Graphic Online, “Lands Commission to roll out $85m project,” Accra, Sep. 2023. [Online]. Available: https://www.graphic.com.gh/news/general-news/lands-commission-to-roll-out-85m-project.html

[14] Business and Financial Times, “Website privacy standards: is Ghana’s Data Protection Commission falling short?,” Accra, Feb. 2025. [Online]. Available: https://thebftonline.com/2025/02/24/website-privacy-standards-is-ghanas-data-protection-commission-falling-short/

[15] A. Birhane, “Algorithmic colonization of Africa,” SCRIPTed, vol. 17, no. 2, pp. 389-409, 2020.

[16] S. Mohamed, M.-T. Png, and W. Isaac, “Decolonial AI: decolonial theory as sociotechnical foresight in artificial intelligence,” Philosophy & Technology, vol. 33, pp. 659-684, 2020.

[17] University of Ghana, “AI must serve Ghana’s development rather than deepen digital dependence, Prof. Isaac Wiafe argues,” Accra, 2026. [Online]. Available: https://www.ug.edu.gh/news/ai-must-serve-ghanas-development-rather-deepen-digital-dependence-prof-isaac-wiafe-argues

[18] T. Suri and W. Jack, “The long-run poverty and gender impacts of mobile money,” Science, vol. 354, no. 6317, pp. 1288-1292, 2016.

[19] R. M. Mutiso, “Five rules for technology leapfrogging in Africa,” Science, 2025, doi: 10.1126/science.adz9028.

[20] International Monetary Fund, “IMF Executive Board completes the sixth review of Ghana’s arrangement under the Extended Credit Facility and concludes the 2026 Article IV consultation,” Press Release PR/25/260, Washington, DC, USA, Jul. 2026. [Online]. Available: https://www.imf.org/en/news/articles/2026/07/27/pr25260-ghana-imf-exec-board-completes-6th-rev-arr-ecf-concludes-2026-aiv-consult-rev-req-36mo-pci