Uganda is preparing to host what organizers are calling Africa’s first artificial intelligence “factory,” a hyperscale, renewable-powered computing hub sited inside the 600MW Karuma Hydropower Plant on the River Nile.

The Aeonian Project will draw on the plant’s power to run sovereign AI infrastructure designed to keep African data and compute on the continent and to accelerate locally-relevant AI development.

The initiative is structured as a 100MW hyperscale facility built in phases. Its initial phase will bring a 15MW AI module and a 10MW sovereign supercomputer — named USIO — online in the second half of 2026. The plan calls for five additional modules to follow, completing the 100MW build-out by 2028. Project partners named at a Nairobi press briefing include Germany’s GIZ, Finland’s HAUS, the European Union Development Fund, and a group of other European investors.

USIO will be built in partnership with NVIDIA, AI infrastructure firm MDCS.AI, and Belgium’s Automation NV, and will rely on NVIDIA’s Blackwell GPU platform. The supercomputer will draw on the Karuma facility’s excess pre-transmission electricity — organizers say about 100MW will be made available to the AI factory — making energy supply a defining feature of the design.

“This is about empowering Africa to control its data backbone responsibly, sustainably, and sovereignly,” Oladele Oyekunle, CEO of Synectics Technologies, said at the briefing.

Energy and cooling are core to the project’s pitch. Aeonian is being designed as one of the world’s greenest AI facilities: it will run entirely on renewable hydropower, use natural river-water cooling, and deploy modular heat-reuse systems and smart infrastructure to limit energy consumption. Schneider Electric East Africa is supplying smart energy and cooling systems.

“By combining new energy with intelligent cooling and modular data center technologies, we are helping build a future where Africa’s data is processed sustainably, securely, and locally,” Ifeanyi Odoh, Schneider Electric’s Country President for East Africa, said.

The project also includes a major connectivity component: a new 2,500km fiber-optic backbone will link Uganda to international subsea cables through Kenya and Tanzania, intended to remove latency barriers and open the facility to regional and global traffic.

Advocates frame Aeonian as a direct response to a long-standing structural weakness: nearly all of Africa’s data organizers cite a figure of roughly 98% that is processed outside the continent. That dependence raises recurring concerns about sovereignty, cost, and the ability of African researchers to train models on local languages and locally relevant datasets.

Niels Van Rees, co-founder of MDCS.AI, emphasized the strategic aim: “In the same way gold and oil once shaped economies, digital tokens will shape the next era of innovation. Africa must not just mine data, but also mint intelligence,” he said, pointing to planned use cases in healthcare, life sciences, higher education, and research.

How Aeonian positions Africa versus emerging AI hubs in Asia and Latin America

Some believe that Aeonian’s ambitions should be read against a broader global pattern in which countries outside North America and Western Europe are creating their own AI infrastructure to secure economic advantage, protect data sovereignty, and capture downstream value from AI.

In Asia, several countries have already built deep AI ecosystems anchored by powerful domestic cloud providers, sovereign investment programs, and dense electronics and semiconductor supply chains. Industrialized Asian economies combine abundant data, large-scale data-center capacity, and proximate hardware supply — an advantage when building and iterating on high-performance AI systems.

National and regional initiatives in parts of Asia have emphasized vertical integration (from chips to cloud to applications), large state or corporate investment in research and training, and close ties between universities and industry. Those ecosystems give Asian hubs rapid scale, local talent pipelines, and low latency to regional markets.

Latin America, by contrast, shows a different trajectory: hubs such as São Paulo and Mexico City have grown as software-and-services centers that focus on local-language NLP, fintech, and agriculture technology. Investments have tended to be smaller in scale than the Asian hyperscale builds but highly targeted toward local commercial needs, with cloud partnerships often formed with global providers while local startups push for solutions tailored to Spanish- and Portuguese-speaking markets.

Aeonian charts a hybrid course. It borrows the hyperscale, sovereign-capacity idea — akin to what some Asian governments and major cloud providers have pursued — but pairs it with an explicitly African policy objective: keep compute and data on the continent to enable models trained on African languages, health, and agricultural data.

Where Asia’s strength is scale, and Latin America’s is localized product focus, Aeonian bets on combining green base-load power with strategic partner networks (chipmaker NVIDIA, infrastructure vendors, and European development finance) to deliver both scale and local relevance.

That positioning gives Kampala a few distinct advantages. First, tying the AI factory to hydropower lowers operating carbon intensity and energy price volatility relative to fossil-fuel-based sites — an attractive pitch for climate-conscious researchers and international partners. Second, on-continent computing reduces friction for African universities and firms that now must run experiments overseas. Third, the sovereign framing and stated interoperability with regional fiber backbones aim to make the factory a continental resource, not a single-country facility.

But the project also faces headwinds that have dogged other nascent hubs. Building and retaining specialized AI talent is expensive and time-consuming. Training a local workforce to operate, optimize, and innovate on large models will require intensive education and upskilling programs. Reliable and affordable last-mile connectivity remains an obstacle in parts of Africa despite the planned fiber backbone, and regional policy coordination — on data protection, research ethics, and export controls — will be essential but difficult.

Finally, long-term commercial viability will likely depend on whether continental demand materializes at scale and whether local startups and institutions can pay for premium compute.

Ghana’s banking sector is bracing for another major shake-up as the Bank of Ghana (BoG) moves to tighten prudential rules, compelling financial institutions to reduce their non-performing loan (NPL) ratios by the end of 2026.

The new measures, announced in August 2025, are seen as part of a broader regulatory posture shaped by the lessons of the country’s landmark 2017–2018 banking sector clean-up.

According to Fitch Ratings, the anticipated improvement in asset quality will be driven largely by accelerated loan write-offs and a more favorable operating environment. The new rules mandate that all regulated financial institutions maintain NPL ratios below 10%. Institutions exceeding 15% will face immediate restrictions on dividend and bonus payments, while those between 10% and 15% will be penalized if they fail to comply within two consecutive years.

Fitch Ratings Findings

As of mid-2025, only four of the country’s 23 banks had NPL ratios below the 10% threshold. More than half reported ratios above 15%, underscoring the scale of the challenge. Still, Fitch believes most banks will be able to bring their NPL ratios below 15% by the end of 2026, mainly through strategic write-offs.

The agency, however, warned that six banks may struggle to meet capital adequacy requirements once regulatory forbearance tied to losses on cedi-denominated government bonds expires at the end of 2025. These banks, weighed down by problem loans and thin capital buffers, are expected to face the greatest difficulty in complying with the new rules.

Ghanaian banks have grappled with weak asset quality for more than a decade. The problem deepened during the sovereign debt restructuring launched in December 2022, which sent the sector’s NPL ratio surging to 26.7% by Q1 2024 from 14.8% at the end of 2022. Payment delays to government contractors, sluggish credit growth, and macroeconomic instability compounded the situation. By mid-2025, the ratio had only modestly improved to 23.1%.

“Most Ghanaian banks have not paid dividends in recent years due to the sovereign default and their reliance on related regulatory forbearance,” Fitch said.

The agency added that the expiration of forbearance at the end of 2025, combined with the threat of dividend restrictions, will push banks to accelerate NPL reduction.

Write-offs as the Key Tool

Encouragingly, the sector’s NPL ratio excluding fully provisioned loans stood at just 8.5% at the end of H1 2025. This suggests banks can execute substantial write-offs without triggering additional provisions. At the same time, strong pre-impairment operating profits, supported by high-yielding sovereign securities, provide a cushion to absorb new provisions without eroding capital. With net loans accounting for only 19% of total banking assets as of April 2025, write-offs are expected to be the main compliance tool.

Fitch also pointed to improving macroeconomic conditions. Ghana’s Long-Term Issuer Default Rating was upgraded to ‘B-’ with a Stable outlook in June 2025, following the country’s successful normalization of relations with most external creditors. The cedi has strengthened, inflation is projected to decline sharply, and banks are expected to benefit from a more stable operating environment.

“Improved operating conditions should help attenuate problem loan generation and support stronger loan growth,” Fitch noted, though it cautioned that persistent payment arrears to government contractors remain a significant risk.

Still, the agency observed that foreclosures and restructurings are unlikely to materially reduce NPL ratios before the new prudential limits take effect. Slow legal processes and lengthy cure periods for restructured loans remain barriers.

Lessons from the 2017–2018 Clean-up

The BoG’s tough stance on NPLs reflects a regulatory posture shaped by its 2017–2018 banking sector clean-up, one of the most consequential reforms in Ghana’s financial history. During that period, the central bank revoked the licenses of nine insolvent banks and merged others to restore stability to a sector crippled by poor corporate governance, undercapitalization, and mounting bad loans. The exercise cost the state nearly GHS 21 billion in bailouts and fundamentally reshaped the industry.

That clean-up underscored the systemic risks posed by unchecked bad loans and weak balance sheets. Regulators now appear determined to avoid a repeat of the crisis, using stricter prudential rules and credible enforcement threats—such as dividend bans and bonus restrictions—to compel banks to clean up their books before vulnerabilities metastasize.

Regional Parallels

The regulatory tightening also mirrors broader trends across West Africa. Last month, Fitch disclosed that while most Nigerian banks are expected to exit regulatory forbearance by December 2025, some may continue under forbearance, facing strict penalties, including dividend bans. The Central Bank of Nigeria, like the BoG, is pressing banks to enter 2026 with stronger capital buffers and cleaner balance sheets.

In Ghana, the new measures signal that regulators have drawn clear lessons from the past that proactive supervision and early intervention are less costly than emergency rescues.

A tentative framework to keep TikTok operating in the United States is moving closer to reality, with sources telling CNBC that the agreement will bring new investors together with existing stakeholders in ByteDance and should close within the next 30 to 45 days.

Oracle, which has long been floated as the U.S. partner to handle TikTok’s data and technical operations, is expected to retain its cloud role under the arrangement. Treasury Secretary Scott Bessent has described the package as a “framework” that emerged from recent talks, and U.S. officials say commercial terms have been in place for months even as negotiations were paused and restarted amid broader trade tensions.

If finalized, the deal would be less about a blockbuster takeover and more about structuring TikTok’s U.S. operations to meet national-security demands while leaving the platform commercially viable. CNBC’s reporting suggests the capital infusion for the U.S. entity could be relatively modest and that no public-listing plan is expected for the new unit — a practical, compliance-first approach rather than a growth-driven reorganization. Oracle’s continued cloud role and the presence of both new and existing investors appear designed to reassure Washington that U.S. user data and critical infrastructure will be insulated from inappropriate foreign access.

However, the political backdrop remains volatile. Congress passed legislation in 2024 requiring ByteDance to divest TikTok or face a ban in the United States, and the White House has twice extended the compliance deadline. President Donald Trump delayed enforcement multiple times and has signaled willingness to keep granting extensions while a deal is negotiated — a process that sources say has already seen deadlines shifted several times this year. Treasury officials say the framework’s timing may hinge on diplomatic conversations between Trump and Chinese President Xi Jinping.

How this mirrors — and diverges from — earlier forced restructurings

The TikTok arrangement shares features with past U.S. actions aimed at Chinese tech companies but also departs from prior playbooks in important ways.

In the case of Huawei and ZTE, Washington relied largely on export controls, government procurement bans, and sanctions that choked off access to U.S. components and markets. Those measures were designed to hobble the companies’ global competitiveness rather than to create viable U.S.-based substitutes; the outcome was effectively exclusion from large swaths of the U.S. market rather than a negotiated, commercially managed divestment. The U.S. response to Huawei emphasized national-security containment through supply-chain restrictions and exclusion from government contracts.

By contrast, the proposed TikTok framework is a form of negotiated remediation: a tailored restructure that keeps the service running for American users while attempting to seal off perceived security risks through governance, ownership, and operational safeguards. The Oracle cloud arrangement echoes earlier proposals in which U.S. firms assumed technical custody roles to address data-residency and software-update concerns. That approach aims to preserve consumer access and investor value while meeting political requirements — a middle path between outright ban and unfettered foreign ownership.

The Ant Group episode from 2020 illustrates another relevant precedent, but one from Beijing’s side. Ant’s IPO was abruptly halted after Chinese regulators moved to restructure the fintech giant under stricter financial controls — a government-led reconfiguration to serve public-policy objectives rather than a market-driven sale or divestment. Ant’s experience shows how national authorities can force sweeping corporate re-engineering when systemic risks are at stake; in TikTok’s case, the U.S. is seeking a negotiated private-sector fix to an analogous set of national-security concerns.

Key differences are therefore strategic: Huawei and ZTE were largely shut out via export and procurement controls; Ant Group was reshaped by its home regulator; TikTok’s resolution is likely to be an international, commercially oriented compromise that leans on U.S. corporate partners and investor commitments to deliver technical and governance assurances.

Where the risks lie

Even a signed framework will not eliminate political or legal risk. Congress could demand stricter oversight or reject the structure if lawmakers judge the safeguards insufficient. Conversely, Beijing could veto or complicate an arrangement that it perceives as an expropriation of a Chinese-owned asset. And the modest size of the planned investment — if accurate — raises questions about whether the new U.S. entity will have the financial heft to stand alone down the line. Treasury officials and White House negotiators have tried to thread that needle of preserving U.S. data security while avoiding a destabilizing rupture in a platform used by tens of millions of Americans.

The bottom line: The emerging TikTok framework is best read as a hybrid response that borrows elements from prior U.S. tech containment strategies while seeking a more pragmatic, market-friendly outcome. It aims to square three competing imperatives: national security, consumer continuity, and commercial viability. How well it succeeds will hinge on the precise mechanics of ownership, governance, and operational control — and on the political winds in Washington and Beijing over the weeks ahead.