Africa’s largest telecom operator, MTN Group, is in advanced negotiations with US and European partners to build a network of data centers across the continent, an ambitious plan aimed at powering artificial intelligence (AI) services and addressing Africa’s severe AI infrastructure gap.

The initiative is part of MTN’s broader effort to diversify revenue beyond traditional telecoms, monetize infrastructure, and position itself as the digital backbone of Africa’s fast-growing economies.

CEO Ralph Mupita confirmed in an interview with Bloomberg that MTN will fund part of the data center build-out directly while partnering with global co-investors, AI infrastructure specialists, and hyperscalers like Microsoft.

“We are now in the commercial negotiation phase and shortlisting partners who can help us scale. Our goal is to conclude these partnerships within the year,” Mupita said.

MTN’s AI data unit, Genova, is at the core of this plan. Genova is already operational, deploying AI across MTN’s 16 markets to optimize network traffic in Nigeria, manage fuel consumption in South Africa, balance energy systems in Benin, and detect fibre cuts in Côte d’Ivoire. Beyond internal use, the new data centers will lease computing power to businesses and governments hungry for AI-driven services.

MTN recently launched West Africa’s largest Tier III data center in Lagos, named the Sifiso Dabengwa Data Centre. With nine megawatts of capacity and cloud infrastructure that MTN claims rivals Amazon, Microsoft, and Google, the facility strengthens the continent’s ability to host advanced digital services.

Africa’s AI Gap

Despite its demographic edge—Africa is home to the world’s youngest population—the continent has less than 1% of global AI data center capacity. South Africa currently dominates what little exists, hosting facilities from Microsoft, Amazon, and Alibaba.

Competitors are also circling. Microsoft and Abu Dhabi’s G42 recently announced a geothermal-powered data center in Kenya, while Airtel Africa, led by billionaire Sunil Mittal, is building AI infrastructure in Nigeria under its Nxtra subsidiary. Airtel has also struck a deal with Xtelify to roll out AI-powered platforms across 14 markets, designed to give its 150,000 field agents real-time insights into customer behavior.

But Mupita acknowledged a core obstacle: reliable electricity. With power infrastructure fragile in many African countries, running large-scale data centers is expensive and complex. He said MTN is exploring renewable and alternative energy sources to ensure viability.

MTN’s Legacy: Leading Every Tech Leap

MTN’s AI ambitions are consistent with its history of taking the lead on transformative technologies in Africa. The company was among the earliest to deploy 4G networks at scale and again led the rollout of 5G services on the continent, launching commercial 5G in South Africa as early as 2020.

In Nigeria, MTN was the first operator to win a 5G spectrum license and turned on its network in 2022, well ahead of most competitors. The move cemented its reputation as the pacesetter for mobile broadband technology in Africa. Telecom industry analysts often note that MTN’s aggressive 5G deployment gave it an edge in offering high-speed services to businesses, governments, and consumers.

MTN is now attempting to repeat this playbook—being first to market with the infrastructure needed to power the next wave of digital innovation by pushing into AI data centers.

Unlike foreign hyperscalers who tend to concentrate capacity in single stable hubs, MTN is spreading its AI infrastructure across 16 diverse markets. This broader footprint, while riskier, could allow MTN to dominate markets overlooked by global players.

Airtel Africa, meanwhile, has opted for partnerships that focus more on software-driven AI solutions rather than heavy infrastructure, signaling a less capital-intensive approach. By contrast, MTN is doubling down on physical infrastructure and betting on long-term control of Africa’s digital backbone.

If MTN’s bet succeeds, it could trigger a domino effect similar to what happened with mobile connectivity. Its early leadership in mobile and 5G adoption forced rivals to catch up and expanded digital access across the continent. Now, with AI data centers, MTN is seeking to position itself as the gateway to Africa’s AI-powered future, one where local businesses and governments no longer rely solely on offshore servers.

Goldman Sachs and Morgan Stanley scored a major legal victory Tuesday after a U.S. appeals court rejected investor lawsuits seeking to hold them liable for losses tied to the March 2021 collapse of Archegos Capital Management, the $36 billion family office run by Bill Hwang.

In a unanimous 3–0 decision, the 2nd U.S. Circuit Court of Appeals in Manhattan found that Archegos was not an insider that owed fiduciary duties to the companies whose stock positions it amassed. As a result, the court concluded the banks could not be held responsible for alleged market-timing or tipping before Archegos’ meltdown.

Circuit Judge Maria Araujo Kahn, writing for the panel, said there was no proof that Goldman and Morgan Stanley “agreed to act in Archegos’ best interest” or that they tipped preferred clients about Archegos’ travails — facts central to the investors’ claims. The ruling upholds a March 2024 dismissal by U.S. District Judge Jed Rakoff in Manhattan and resolves seven related appeals (In re Archegos 20A Litigation, Nos. 24-1159, 24-1161, 24-1162, 24-1166, 24-1173, 24-1177, and 24-1178).

Plaintiffs had accused Goldman, Morgan Stanley, and other prime brokers of using advance knowledge of Archegos’ inability to meet margin calls to dump billions of dollars of Hwang’s favored stocks — including ViacomCBS, Discovery, and five Chinese companies such as Baidu — thereby accelerating price declines and inflicting losses on other shareholders.

How Archegos imploded

Archegos built massive, concentrated positions using total return swaps and similar contracts, creating what regulators and market participants called highly leveraged, opaque exposure (estimates of related exposure at the time approached $160 billion). When prices of core holdings fell, Archegos could not meet margin calls, triggering a rapid unwind that produced billions in losses for counterparties.

Credit Suisse suffered some of the largest write-downs and was later acquired by UBS; Nomura and several other banks also reported heavy losses.

Bill Hwang and Archegos’ former CFO Patrick Halligan were convicted of fraud in July 2024. Hwang received an 18-year sentence; Halligan received eight years; both are appealing and remain free on bail.

While the appeals court shielding Goldman and Morgan Stanley from these investor suits narrows civil exposure, legal and regulatory consequences continue to ripple through the banking sector. In July, Goldman, Morgan Stanley and Wells Fargo agreed to pay a combined $120 million to settle a suit by former ViacomCBS shareholders who alleged conflicts of interest tied to prime broker relationships with Archegos.

Why Archegos changed the conversation about derivatives and hedge fund transparency

Beyond courtroom rounds and shareholder lawsuits, Archegos left an outsized imprint on policy debates in Washington and at U.S. financial regulators. The episode exposed how large exposures can accumulate off-balance-sheet via swap contracts and prime broker arrangements, and it triggered renewed scrutiny of disclosure, market structure, and risk controls.

Archegos was not a registered hedge fund; it operated as a family office and used derivatives to achieve large economic exposure without holding the equivalent equity positions on public records. That opacity — and the speed with which the unwind produced extreme losses for some global banks — prompted lawmakers, regulators, and market participants to press for reforms aimed at reducing systemic blind spots:

• Policymakers and market observers focused on the role of total return swaps and other non-cleared derivatives in allowing private vehicles to magnify risk without public disclosure. Proposals discussed in the aftermath included expanded reporting requirements to regulators for large swap positions and enhanced transparency around prime broker flows.
• Regulators and industry groups revisited how prime brokers monitor. counterparties’ aggregate exposures, the adequacy of intraday margin calls, risk-based concentration limits, and how quickly firms can and should act when clients approach distress. The speed of the Archegos unwind prompted calls for more conservative margining and earlier risk mitigation.
• The crisis fed a debate over whether more types of economically significant swap exposure should be centrally cleared or otherwise standardized to reduce bilateral counterparty contagion.
• Lawmakers questioned the regulatory blind spot that family offices occupied compared with registered investment advisors and funds. Some policymakers argued for tailored disclosure rules for very large private investment vehicles that wield systemic economic influence.

Regulatory activity and political attention

In the months after the collapse, congressional committees held hearings, asking bank CEOs and regulators to explain what happened and what steps had been taken to prevent a repeat. Regulators—including staff at the Securities and Exchange Commission and the Commodity Futures Trading Commission, as well as bank supervisors—said they were reviewing the episode to identify gaps in supervision and market safeguards.

Crucially, the Archegos fallout did not produce a single, sweeping legislative overhaul. Instead, it sharpened the agenda across multiple forums.

Thus, Tuesday’s 2nd Circuit ruling removes a major avenue of investor claims against Goldman Sachs and Morgan Stanley tied to the Archegos collapse and cements the legal finding that Archegos was not a corporate “insider” in the way the plaintiffs alleged.

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.