Nokia turns to AI as Africa’s telecom market enters its next phase

Africa’s telecom story has long been about expanding connectivity—from towers and fibre to 4G and 5G. AI is now changing what those networks are expected to do, pushing them from moving data towards processing it and acting on it in real time.

For Nokia, the Finnish technology company that once dominated Nigeria’s mobile phone market and has spent decades building Africa’s telecom infrastructure, that shift is both an opportunity and a new competitive battleground.

Africa’s telecom market, estimated at $66 billion and projected to reach $90.3 billion by 2030, is attracting fresh investment from global technology companies such as Meta, Google and Microsoft. Their investments in subsea cables, cloud infrastructure and AI are giving them greater control over the infrastructure underpinning Africa’s next phase of digital growth. For Nokia, the stakes go beyond selling network equipment: it must secure a role in that infrastructure or risk becoming a lower-value connectivity provider in the emerging AI economy.

That is the market Nokia is positioning for. The company expects AI-native networks, edge computing, automation and software to become increasingly important, even as much of Africa still struggles with reliable connectivity.

“Africa remains highly unconnected right now,” Danial Mausoof, Nokia’s vice president of Mobile Infrastructure for the Middle East and Africa, said in an interview with TechCabal on August 25. “We’re invested. We’re fully committed to Africa.”

Nokia still sees 4G and 5G as the foundation of its African strategy. Mausoof said it works with major operator groups, including Airtel, Orange, Vodacom, Safaricom, and Maroc Telecom, and has 5G deployments in markets such as Angola, South Africa, and Ethiopia.

But Nokia wants to extract more value from those networks than connectivity alone can provide. Its Artificial Intelligence Radio Access Network (AI-RAN) platform combines existing telecom infrastructure with GPU-based computing to bring AI workloads into the network. Nokia launched its commercial AI-RAN platform on July 15, 2026.

The shift is central to Nokia’s future in Africa. Having largely exited the consumer mobile phone market, it needs to move beyond selling network hardware and competing on price. If operators continue to treat networks mainly as connectivity infrastructure, Nokia risks growing price pressure from lower-cost rivals such as Huawei, which has built a strong presence across the continent.

Nokia’s bigger bet is that software and computing can make the network a more valuable asset. Through AI-native networks and platforms such as AVA Autonomous Operations and Network as Code, it aims to build recurring software revenue, expand into edge computing, and capture more value from 5G and, eventually, 6G. Its relationships with African operators give it market access, but not necessarily a competitive advantage.

Nokia’s bet vs the market bet

That vulnerability becomes clearer as AI infrastructure increasingly consolidates around companies that control both computing hardware and software. Adedeji Olowe, CEO of Lendsqr and a telecom industry expert, is sceptical that Nokia can establish a strong position in the market.

“I don’t think they have a chance, but then I could be wrong,” Olowe said. “Nokia has been in the enterprise market for a while, but the market is shifting significantly, and all the other players are likely ahead of them.”

Olowe points to companies such as Amazon Web Services, Google and Huawei, which have developed their own AI chips and broader technology stacks. That vertical integration gives them greater control over the computing layer on which AI services depend.

Emmanuel Ezenwere, CEO of Arone Technologies, a Nigerian hardware, robotics, and clean energy company, sees the same weakness. “Nokia doesn’t have its own chip,” he said, arguing that this leaves the company dependent on partners for a critical part of the AI stack.

That matters because AI-RAN is more than a telecom software upgrade. Much of the value is shifting to the underlying compute, interconnects, software ecosystems and intellectual property. Nokia controls the network layer, but its AI-RAN strategy relies on NVIDIA for the computing hardware and software that power the AI workloads.

Nokia’s approach is to combine its anyRAN software with NVIDIA’s Aerial software and GPU hardware. The companies are developing an architecture in which GPU-based computing sits alongside Nokia’s existing baseband systems, allowing conventional 4G and 5G traffic to continue running while AI workloads such as inference and training are handled by the GPU layer.

The arrangement gives Nokia access to advanced AI computing without having to develop its own accelerator. But it also raises a question about how much of the resulting advantage Nokia can actually defend. NVIDIA can supply similar computing infrastructure to other network vendors, potentially making the hardware advantage available across the industry.

Ezenwere argues that this is where Nokia’s position becomes more vulnerable. Huawei, by contrast, has greater control over its own technology stack, including its Ascend AI processors, networking hardware and software ecosystem. Huawei has not, however, publicly demonstrated Ascend silicon embedded directly inside its base stations, so the two approaches are not technically equivalent.

The distinction is important. Nokia is betting on an open, merchant-compute model that combines telecom infrastructure with NVIDIA’s computing platform, while Huawei is pursuing deeper vertical integration across its hardware and software stack. The question is whether Nokia’s model can deliver better performance, power efficiency and cost at scale.

Nokia is also entering AI-RAN from a weaker position in the traditional RAN market. It ranks behind Huawei and Ericsson in RAN revenue, while the largest vendors control most of the market. Ezenwere argues that this makes Nokia’s AI-RAN proposition dependent on proving that it can deliver meaningful gains over technologies already being developed by its larger rivals.

“On technology and market position, Nokia is third by RAN revenue behind Huawei and Ericsson, with the top five holding 96% and concentration at a ten-year high,” Ezenwere said. “Ericsson has been shipping commercial AI in RAN since June 2026 at roughly 10% spectral efficiency and 20% downlink throughput gains without GPUs across fifteen-plus operators. Nokia’s figures are a roadmap projection rather than delivered results: it claims over 20% today and forecasts 50% by 2027 and above 100% by 2028.”

With Huawei, ZTE, Samsung and Ericsson exploring AI-RAN without putting GPUs directly into the baseband, Nokia will need to show that its more expensive, compute-heavy approach delivers significantly better results.

The economics may be harder than the technology

The economics could be an even bigger constraint in Africa.

Nokia has not published a fixed price for its AI-RAN platform. Telecom infrastructure is typically priced according to capacity, deployment size, number of sites, and software licensing rather than a single retail price. But AI-RAN is more expensive than conventional infrastructure because it adds high-performance computing to the network.

Industry estimates suggest GPU-accelerated processing at centralised C-RAN hubs can cost tens of thousands of dollars per node and, depending on configuration, more than $100,000. Power adds another cost: high-density AI processing can consume several kilowatts, which is significant in markets where operators already spend heavily on electricity, batteries, generators, and diesel.

That makes a continent-wide rollout difficult to justify. African operators are still spending heavily on fibre backhaul, spectrum, 4G and 5G expansion and rural coverage, while generating far less revenue per user than operators in developed markets. Much of their equipment is also priced in dollars, exposing them to currency risk.

AI-RAN is therefore more likely to start where the economics are easiest to defend: high-traffic urban sites, centralised network hubs and enterprise environments where additional capacity or low-latency computing can generate measurable returns.

Mausoof’s description of Nokia’s architecture reinforces that approach. Operators can keep their existing 4G and 5G systems while adding GPU-based computing alongside their baseband infrastructure. Conventional mobile traffic continues through the existing system, while the GPU layer handles AI workloads. A single software layer manages both.

The proposition is less about rebuilding the network than adding an intelligence and computing layer to infrastructure operators already own. Mausoof said Nokia is targeting up to a 2.5-fold improvement in spectral efficiency, potentially allowing operators to extract more capacity from existing spectrum.

That claim, however, remains largely a future-facing proposition. Ezenwere points to a wider industry debate over the actual gains AI can deliver in the RAN. Ericsson has reported commercial AI-RAN deployments producing more modest efficiency improvements without GPUs, while Nokia’s higher figures are based on its own testing and projections. The gap between laboratory performance and measurable operator returns could ultimately determine whether AI-RAN becomes a meaningful business or an expensive upgrade.

AI could nevertheless address some immediate operator problems. Networks can automatically adjust resources to traffic patterns, reduce capacity during low-demand periods and automate maintenance. Mausoof pointed to a Nokia deployment during the Hajj in Saudi Arabia, where the company says its AI-powered automation handled between 10,000 and 15,000 network interactions and changes during a period of intense demand.

That may be a more compelling African use case than putting GPUs everywhere. Where power costs are high and field engineers are difficult to deploy, automation that reduces energy consumption and network maintenance could deliver returns without requiring a complete network transformation.

Africa could test whether AI-RAN can actually pay

The larger prize is enterprise revenue. Mining companies could use dedicated networks for autonomous machinery; ports and logistics operators could combine connectivity, sensors and computer vision; manufacturers could coordinate machines over low-latency networks; and utilities could use connected infrastructure to monitor assets and predict failures.

Edge computing is central to this model because it allows data to be processed closer to where it is generated. That could allow operators to sell network capabilities—not just bandwidth—based on latency, reliability, security and throughput.

But the market for these services is still narrow. Large industrial companies may have clear reasons to pay for real-time AI processing, while many African businesses are still moving basic workloads from legacy systems to cloud platforms. The demand for sophisticated AI inference at the network edge is therefore likely to emerge first from a small group of technologically advanced enterprises.

For operators, this means changing how they monetise their networks. Enterprise customers would need to pay for specific performance or business outcomes, while operators would need to develop use cases with businesses rather than simply sell conventional data packages.

The technology is still some way from widespread commercial deployment in Africa. Nokia is discussing AI-RAN use cases with operators in the region, although Mausoof said they cannot yet be named because the projects remain in trial stages. Trials are expected to begin as early as Q1 2027, followed by further announcements at Mobile World Congress 2027 and commercialisation in 2028.

Mausoof expects broader adoption in Africa to extend beyond 2030 because markets differ significantly in technology maturity, spectrum availability, infrastructure and operator capabilities. More advanced 5G markets are likely to move first, while others remain focused on expanding basic connectivity.

That creates a difficult commercial calculation for Nokia. It is trying to sell a technology that promises greater efficiency and new revenue for operators who are still under pressure to fund basic network expansion.

The competitive landscape makes the challenge harder. Microsoft, Google, Amazon and NVIDIA are building positions across cloud and AI, while Ericsson and Huawei remain entrenched in telecom infrastructure. The hyperscalers have advantages in cloud infrastructure, AI models, developer ecosystems and capital, while NVIDIA controls a critical part of the computing stack.

Nokia’s potential advantage is therefore narrower: its RAN expertise, existing operator relationships and installed base. But even those advantages are under pressure. If NVIDIA can supply similar accelerated computing platforms to Ericsson, Samsung and other vendors, Nokia’s head start may not become a durable moat.

Huawei presents a different challenge. Its advantage in Africa is not simply access to cheaper financing or Chinese state support. It has an installed base, competitive pricing and the ability to sell a broader portfolio spanning telecom networks, fibre, cloud and government technology. That gives it more opportunities to capture spending as African governments and enterprises build digital infrastructure.

For Nokia, the question is whether its AI-RAN strategy can turn its existing network footprint into a stronger position in that ecosystem.

Africa could be a crucial test. GSMA data shows that 3G and 4G networks cover about 85% of the population, yet more than 60% remain offline. High smartphone and data costs continue to limit usage, even where coverage exists.

That makes Africa a difficult market for expensive AI infrastructure—but potentially a useful one for technologies that lower network operating costs. The strongest near-term case for AI may therefore not be autonomous networks or GPU-powered cell sites, but using AI to reduce energy consumption, automate operations and squeeze more capacity from existing infrastructure.

Nokia’s long-term bet is that those efficiencies will eventually lead to something bigger: networks that become programmable platforms for AI and enterprise applications.

“Network infrastructure in Africa has to be scaled up, and that is our priority—to continue to drive connectivity,” Mausoof said. “Infrastructure will evolve into AI-native platforms that will be programmable, software-led and edge-enabled.”

The question is whether Nokia can turn that vision into a defensible business. Its partnership with NVIDIA gives it access to the computing power it needs, but not necessarily control of the technology stack. Its operator relationships give it access to customers, but not guaranteed demand. And its AI-RAN technology may improve network economics, but operators still need a compelling return on investment.

To win in Africa, Nokia will likely need more than a technically capable AI-RAN platform. It will need deep partnerships with operators, enterprise use cases that generate measurable revenue or savings, and large-scale deployments that turn its existing network footprint into an advantage.

That is ultimately what is at stake: whether Nokia can use AI to make its telecom infrastructure more valuable—or whether the next layer of Africa’s digital infrastructure is captured by companies that control the compute, software and capital behind it.

True scale demands moving beyond surface-level integrations to robust execution. We’ve filtered the noise out of Moonshot 2026, optimising the conference strictly for high-calibre connections between startup founders, global financial operators, enterprise leaders, and individuals rewiring Africa’s technical frameworks. Get 20% off Early Bird tickets for a limited time.