“WHO WILL PAY FOR THE ELECTRICITY?” BRIGHT SIMONS CHALLENGES GOVERNMENT OVER GHANA’S $250 MILLION AI COMPUTE CENTRE

IMANI Africa Vice President Bright Simons has challenged the economic and technical assumptions behind Ghana’s proposed national artificial intelligence compute centre, warning that expensive electricity, questions over reliable power supply and the enormous cooling requirements of large data facilities could make the project uncompetitive unless the government produces a convincing operating model before committing substantial public resources. Speaking during the Digital Economy Forum on AI and the Ghana Economy on Wednesday night, Simons directly questioned Communications, Digital Technology and Innovations Minister Samuel Nartey George over whether Ghana can realistically operate a major AI computing facility at electricity prices considerably higher than those available to competing centres elsewhere in the world.

The debate comes as the government moves ahead with an ambitious plan to spend approximately US$250 million on the first phase of a national AI compute centre, alongside another US$20 million intended to seed an artificial intelligence investment fund. The government sees the proposed facility as a critical piece of digital infrastructure that could give Ghanaian universities, researchers, technology companies and public institutions access to computing power that is currently purchased from foreign providers. George has said the first phase could be completed within 24 months and that procurement processes involving the Public Procurement Authority are already underway.

Simons did not argue that Ghana should abandon artificial intelligence or refuse to invest in computing infrastructure. His central argument was that enthusiasm for AI cannot replace a rigorous assessment of whether the economics of operating such a facility will actually work. He warned that compute centres are unusually sensitive to electricity costs because high-performance processors must operate continuously while additional power is required for cooling, networking, storage and backup systems. If Ghana starts with a substantially more expensive energy base than competing destinations, he argued, the country could build an impressive facility only to discover that customers can purchase the same computing capacity more cheaply elsewhere.

Simons told the forum that competitive AI facilities often depend on electricity in the region of three to five US cents per kilowatt-hour, while Ghana could be attempting to operate at rates closer to 10 or 12 cents. In his assessment, that gap would immediately affect the centre’s ability to compete. If the facility passed the higher operating cost to customers, Ghanaian researchers and technology businesses might continue buying computing capacity from international platforms instead. If government deliberately kept prices low to attract users, taxpayers could eventually be required to subsidise the difference between what the centre costs to operate and what customers are able or willing to pay.

That raises one of the most important unanswered questions surrounding the planned investment: whether the US$250 million commitment refers mainly to the capital cost of building and equipping the facility or whether government has developed a long-term funding strategy for electricity, cooling, maintenance, equipment replacement and specialist staff after the centre becomes operational. High-performance AI infrastructure is not a project that ends when the building is completed. Graphics processing units and other specialised processors can become outdated quickly, and a competitive facility may require repeated investment to remain useful as global computing technology advances.

Electricity reliability was another major concern raised by Simons. Even if Ghana succeeded in negotiating a lower tariff for the centre, he argued that price would solve only part of the problem because an advanced computing facility requires extremely dependable power. Sudden interruptions can affect computing workloads and force operators to invest heavily in batteries, generators or alternative energy sources. Ghana has made progress in expanding generation and stabilising parts of its electricity system, but periodic supply constraints and the financial difficulties affecting the power sector remain important considerations when planning infrastructure expected to operate around the clock.

The issue is particularly sensitive because the proposed centre would add a new category of large industrial electricity consumer at a time when households and businesses already complain about power costs. A national AI facility could require dedicated electricity arrangements, its own generation capacity or long-term power purchase agreements to prevent its operation from placing additional strain on existing supply. The eventual business model will therefore matter as much as the technological ambition because government will need to explain whether the centre will pay ordinary commercial tariffs, receive a special rate or generate part of its own electricity.

Large technology companies around the world are confronting the same problem. The rapid growth of generative AI has increased electricity demand from data centres because training and operating advanced models requires thousands of powerful chips working simultaneously. Countries seeking AI investment are therefore competing not only through tax incentives and fibre connectivity but through the availability of large quantities of affordable, reliable electricity. Ghana’s ambition to become an African computing hub consequently places it in competition with countries that may have cheaper hydroelectricity, natural gas, renewable power or larger electricity grids.

Simons also raised the less frequently discussed question of water. Large data centres can consume substantial amounts of water depending on the cooling system used, local climate and design of the facility. Processors generate heat continuously and must be kept within controlled temperatures, meaning operators can use evaporative cooling systems, chilled-water systems, air cooling or newer liquid-cooling technologies. The exact water requirement of Ghana’s proposed centre will depend on its design, but Simons argued that government should establish those numbers before making major commitments rather than discovering the environmental and utility requirements after construction begins.

His concern is particularly relevant in communities where water systems already face pressure from population growth and infrastructure limitations. A large facility should not be designed in a way that forces residents and commercial users to compete with computing infrastructure for treated water. Ghana could reduce that risk by selecting water-efficient cooling technology, using recycled water, locating the centre near suitable non-potable water sources or incorporating advanced closed-loop cooling systems, but each option affects capital and operating costs.

Simons invoked Ghana’s previous experience with major infrastructure projects, including challenges surrounding the desalination plant developed to increase water supply in parts of Accra, as a reason for detailed planning before government commits itself. His point was broader than the specific history of that plant. He argued that technology projects can look attractive at the announcement stage while becoming financially difficult when governments underestimate operating costs, long-term contractual obligations or the difference between technical capability and commercial sustainability.

The IMANI executive also questioned what has fundamentally changed in Ghana’s technology ecosystem to make the current AI project more likely to succeed than previous efforts to develop locally controlled digital platforms. He pointed to the experience of the late technology entrepreneur Herman Chinery-Hesse, who spent years advocating Ghanaian-built software and digital services but often struggled to secure large-scale government adoption. Simons asked why a new generation of state-backed technology projects should succeed if institutional problems that undermined previous local initiatives have not been clearly identified and corrected.

That question goes directly to the relationship between government procurement and Ghana’s technology industry. Ghana has produced software engineers, financial technology companies and digital platforms capable of competing internationally, yet government institutions have also relied heavily on foreign vendors for some major technology projects. If the national AI compute centre is intended partly to strengthen local capacity, its procurement structure will determine whether Ghanaian engineers and companies develop lasting expertise or whether most of the technical value is captured by foreign equipment and service providers.

George has presented the project as a way of reducing precisely that dependence. During the same forum, he argued that Ghanaian researchers, universities and companies currently purchase computing capacity from providers outside the country and that a domestic centre would allow some of that spending to remain inside Ghana. Government also believes locally controlled infrastructure could improve data sovereignty by reducing the need for sensitive public-sector information to be processed entirely on foreign AI platforms.

The minister disclosed that analysts working around government officials are already using international AI systems to process documents, including material he considers sensitive. His concern is that Ghana can pass data-residency and cybersecurity rules but still lose practical control over information if public servants routinely upload government documents into external systems. The planned infrastructure is therefore being positioned not merely as a commercial data centre but as part of a national strategy to give public institutions access to AI tools operating within a more controlled environment.

One proposal is the development of a Ghanaian language and government-focused AI system sometimes described by the minister as “Ghana Chat.” The idea is to build a local model capable of handling government and Ghana-specific information without requiring every task to be sent to established foreign platforms. George has acknowledged that such a system would not initially perform at the same level as the most advanced international models because those systems have been trained on vastly larger quantities of data and developed through enormous investments.

The government’s argument is that Ghana has to begin somewhere if it wants local capability instead of remaining permanently dependent on external providers. George has said the country cannot expect to solve every AI infrastructure problem within a single administration and has described the compute centre as an initial foundation rather than the final stage of Ghana’s artificial intelligence ambitions. The first phase is targeted for completion within 24 months, after which further expansion could follow depending on demand and available investment.

Simons’s challenge is that beginning somewhere does not remove the need to establish whether that starting point is economically sound. A US$250 million public commitment is substantial in a country managing debt, infrastructure needs and competing demands in health, education, roads and energy. If the centre is expected to make money or at least recover a meaningful portion of its operating costs, the government needs to demonstrate who its customers will be, what they currently pay international providers and why they would move their workloads to the Ghanaian facility.

If government intends the centre primarily as strategic national infrastructure rather than a profit-making business, that should also be made explicit because the evaluation criteria would be different. A facility built for data sovereignty, public research, universities and national security could justify some public subsidy in the same way governments fund laboratories or communications infrastructure that do not produce immediate commercial returns. The debate therefore depends partly on defining what the centre is actually supposed to achieve.

George has gone beyond domestic use and suggested that Ghana could eventually sell computing services to customers elsewhere in Africa. Government believes building capacity larger than Ghana’s immediate requirements could position the country as a regional AI hub and allow the facility to earn revenue from universities, businesses and governments across the continent. That strategy could increase the economic case for the centre if Ghana secures enough customers, but it also intensifies the importance of the electricity-cost question because regional clients will compare Ghana’s prices with alternative facilities in Africa, Europe, the Middle East and global cloud platforms.

The government has projected that artificial intelligence could eventually generate as much as US$5 billion in annual economic benefits for Ghana. That projection covers the wider impact of AI across agriculture, finance, government, healthcare, education and other sectors rather than revenue from the compute centre alone. The figure nevertheless forms part of the political case for making large early investments before other African countries secure an advantage in the emerging market.

There has not yet been a detailed public methodology demonstrating exactly how the US$5 billion projection would be distributed between sectors or when those gains are expected to materialise. That absence strengthens Simons’s argument for independent technical scrutiny because major public investments are easier to assess when assumptions about demand, pricing, employment, energy and economic benefits can be examined openly.

Simons urged the government to invite think tanks and independent experts to submit technical memoranda and challenge the project before key decisions are finalised. His argument is that scrutiny should not be interpreted as opposition to innovation. Instead, government should expose its assumptions to engineers, economists, energy specialists, data-centre operators and financial analysts who can identify weaknesses before contracts are signed and equipment is purchased.

Such an assessment could examine the planned computing capacity, expected utilisation, electricity tariff, backup-power design, cooling system, water demand, staffing needs, cybersecurity architecture and replacement cycle for expensive processors. It could also compare the cost of building a state-backed facility with alternatives such as negotiating sovereign cloud arrangements, renting dedicated international capacity or establishing public-private partnerships in which private operators absorb a larger share of commercial risk.

The question of skills has already emerged separately from the infrastructure debate. Ghanaian technology companies have said the country does not currently possess enough specialised AI engineers, data scientists and data-centre professionals to support the rapid expansion envisioned under the national strategy. Hubtel has disclosed that it previously struggled to recruit the technical specialists it wanted for its own AI work, illustrating that buying processors alone will not automatically create an artificial intelligence ecosystem.

Running the proposed centre will require professionals capable of operating high-performance computing clusters, maintaining networking equipment, managing cooling and power systems, securing the infrastructure against cyberattacks and helping researchers use the resources efficiently. If many of those positions initially have to be filled by international contractors, Ghana will need strong skills-transfer provisions to ensure local expertise grows over time.

The government is attempting to address the human-capacity challenge through the One Million Coders programme and other training initiatives. AI focal persons have also been appointed across government ministries, while Cabinet ministers have received introductory AI training and programmes are being prepared for senior civil servants. The strategy is designed to ensure that Ghana develops both infrastructure and people capable of using it rather than treating the compute centre as a standalone construction project.

A separate national data-exchange hub is also being developed to connect approximately 48 government registries. The idea is that artificial intelligence becomes substantially more useful when government agencies can analyse reliable information from healthcare, education, population, agriculture and other systems rather than operating isolated databases. The compute centre would provide the processing power, while the data-exchange infrastructure would make information accessible under controlled rules.

That creates another area requiring careful oversight because greater data integration can improve public services while simultaneously creating serious privacy and cybersecurity risks. A system capable of connecting dozens of government databases becomes highly valuable to the state but also highly attractive to criminals and hostile actors. Government has therefore announced plans for central technical review of public-sector AI acquisitions and says cybersecurity, data protection, interoperability and sustainability will be considered before systems are approved.

George’s strongest defence of domestic computing infrastructure is ultimately about national control. He argues that Ghana should not enter an AI-driven future in which the country generates valuable data while foreign companies own the models, processors and cloud infrastructure used to analyse it. From that perspective, the US$250 million facility represents an attempt to build a strategic asset whose importance may become clearer over decades rather than through short-term financial returns alone.

Simons’s counterargument is not that foreign dependence is desirable but that national ownership must still operate within economic reality. A Ghanaian facility that is too expensive to use could fail to attract researchers and companies, while a heavily subsidised centre could become another recurring burden on the national budget. A technically sophisticated building is not automatically a successful digital economy project if utilisation remains low.

The public debate therefore goes beyond whether one supports or opposes artificial intelligence. Both sides of the discussion broadly accept that AI will influence Ghana’s economy and that computing capacity will become increasingly important. The disagreement concerns how quickly Ghana should invest, what form that investment should take and how much evidence government should produce before spending hundreds of millions of dollars.

For universities and researchers, domestic computing capacity could be transformative if it is affordable. Training large models and conducting advanced simulations can be prohibitively expensive when institutions must buy access from foreign cloud platforms using scarce foreign currency. A subsidised national research allocation could allow students and academics to work on problems specific to Ghana that may never attract investment from global technology companies.

Agriculture is one area government believes could benefit. Satellite data, soil information, weather predictions and crop records could be analysed to help farmers plan planting, predict disease and target extension services. Healthcare systems could potentially use AI for diagnostics and resource allocation, while government could analyse population data to determine where schools, hospitals and infrastructure are needed most.

Those potential applications are real, but their value depends on high-quality data and functioning institutions. Artificial intelligence cannot compensate for incomplete records, poorly maintained databases or government agencies unwilling to share information responsibly. Investments in data quality may therefore be just as important as investments in processors.

The centre’s eventual location will also matter enormously. A suitable site needs strong fibre connectivity, reliable electricity infrastructure, physical security, access to cooling resources and protection from flooding or other environmental risks. Its proximity to major power infrastructure could reduce transmission challenges, while its cooling design will determine whether water becomes a significant concern.

Renewable energy could form part of the answer to Simons’s electricity challenge. Ghana has substantial solar potential, and a combination of grid power, dedicated solar generation, battery storage and other sources could reduce long-term energy costs. However, high-performance compute centres require continuous power, meaning intermittent renewable generation normally needs substantial storage or backup capacity to support round-the-clock operation.

Natural gas could provide another source, particularly if the centre secures a dedicated generation arrangement, but fuel pricing and Ghana’s existing energy-sector debts would need consideration. Hydroelectricity offers relatively low operating costs but is constrained by national demand and changing water conditions. There is therefore no single obvious energy solution, reinforcing calls for the government to publish a detailed technical plan.

The water question similarly depends on design rather than simply the existence of a large facility. Some modern data centres use cooling systems that consume far less water than traditional evaporative systems, although those alternatives can require more electricity or higher capital expenditure. Choosing the correct technology will therefore involve trade-offs between energy use, water use and construction costs.

International technology companies are themselves wrestling with those trade-offs as AI demand expands. Data-centre projects in the United States and Europe increasingly face questions over electricity generation, grid connections and water consumption, while some communities have resisted new facilities because of concerns about utilities and environmental impact. Ghana is therefore entering an industry where the infrastructure challenges Simons highlighted are being debated globally rather than being unique to Ghana.

What is unique is Ghana’s fiscal position and the opportunity cost of committing public money. Wealthier countries can absorb unsuccessful technology projects more easily than developing economies with limited budgets. Ghana consequently has stronger reasons to ensure that the design is financially robust before procurement reaches an irreversible stage.

That is why the emerging argument over the AI compute centre could become one of Ghana’s most important technology-policy debates. Government sees a narrowing window in which countries that invest early can capture artificial intelligence opportunities, while critics want evidence that urgency will not become an excuse for weak scrutiny. Both perspectives reflect legitimate concerns: moving too slowly could leave Ghana dependent on infrastructure built elsewhere, while moving too quickly could lock the country into an expensive project that struggles to compete.

George responded to the wider concerns raised during the forum by acknowledging that many of the questions were legitimate while maintaining that government believes it is developing solutions. He remains confident that Ghana should invest early rather than regret missing the AI transition later. The minister has also stressed that the first phase is only one part of a much longer national strategy and that policymakers should not promise that every technology gap can be closed immediately.

The next phase will therefore be crucial. Procurement is underway, the government has publicly identified a 24-month target and a substantial financial commitment has been announced. Before construction and equipment purchases accelerate, attention is likely to shift toward the details Simons wants made public, including the centre’s energy model, expected capacity, water requirements, ownership structure, commercial pricing and long-term operating budget.

Ghanaians will also want to know precisely where the US$250 million will come from and whether the entire amount represents direct government expenditure, external financing, private investment or some combination of sources. The government has been clearer about the broad allocation than about the final financing structure, making that another area where greater disclosure would help the public assess the project.

The proposed US$20 million AI fund introduces another dimension because government expects private financial institutions to contribute alongside the state. Banks, insurers, pension funds, telecommunications companies and technology businesses could eventually participate if incentives and governance structures are attractive. A similar partnership approach may ultimately prove relevant for the compute centre itself if private investors can reduce the financial risk carried by taxpayers.

The debate also arrives at a moment when Ghana is trying to position Accra as a regional technology hub. International data-centre companies are expanding across Africa, while countries including South Africa, Kenya, Nigeria, Morocco and Egypt are competing for cloud and AI investment. Ghana’s political stability, undersea cable connections and growing digital-services sector provide advantages, but electricity pricing and specialised skills could determine whether those strengths translate into major compute investment.

Simons’s challenge therefore goes to the heart of the government’s regional ambition. If Ghana wants to sell computing services beyond its borders, it must compete on more than patriotism or government policy. Customers running expensive AI workloads will compare performance, reliability and price. If a provider in another country can deliver the same computing power significantly more cheaply, many commercial users will choose the cheaper option regardless of Ghana’s national strategy.

The government could still justify a higher-cost centre if it provides services that cannot easily be outsourced, including sensitive public-sector computing, sovereign datasets or research considered strategically important. That again demonstrates why the centre’s mission must be clearly defined before its commercial viability can be judged.

For now, the most significant outcome of the Digital Economy Forum may be that Ghana’s AI ambitions are entering a more rigorous phase of public scrutiny. The country has moved beyond general statements about embracing artificial intelligence and is now confronting the practical questions that accompany expensive infrastructure. Electricity, water, data, engineers, financing and customers are no longer secondary details because they will determine whether the project succeeds.

Simons wants those questions answered before Ghana locks itself into a large expenditure. George believes the risks can be managed and argues that Ghana cannot afford to remain permanently dependent on foreign computing power. Between those two positions lies the central policy decision facing the country: how to build enough national capacity to participate meaningfully in the AI economy without creating an expensive facility that taxpayers must continually support.

That is why the question of who ultimately pays for the electricity is more than a line from a television debate. It captures the central economic challenge behind the entire project. Ghana can purchase advanced processors and build a modern data centre, but every hour those machines operate will require power, cooling, maintenance and highly skilled personnel. Unless those recurring costs are competitive and supported by consistent demand, the true price of the centre could extend far beyond the US$250 million announced for its first phase.

Government now has an opportunity to answer those concerns through transparency rather than rhetoric. Publishing a detailed feasibility assessment, allowing independent technical review and explaining the energy, water and commercial models would help separate legitimate national ambition from questions that remain unresolved. If the numbers support the project, such scrutiny could strengthen public confidence. If they reveal weaknesses, correcting them before construction would be far cheaper than trying to redesign an operating facility later.

Ghana’s ambition to become a serious African player in artificial intelligence is not in dispute. The debate now concerns whether the infrastructure supporting that ambition is being designed with the same sophistication as the technology it is intended to host. Bright Simons has placed that question squarely before the government, and with procurement moving ahead and the first phase targeted within two years, the answers are becoming increasingly urgent.