AI infrastructure in Nigeria: Closing the GPU gap

The big question for Nigerian AI founders in 2026 is no longer whether local data centers are being built (because they are), but whether the infrastructure coming online is actually designed for the high-density GPU workloads they need.

Key takeaways

Nigeria’s operational data center capacity stands at approximately 86 MW as of 2025, with more than 320 MW of new capacity under construction or planned, representing a nearly fourfold expansion that positions Lagos as West Africa’s emerging interconnection hub.
This pipeline includes the 38 MW Nxtra hyperscale facility at Eko Atlantic and the 100 MW Kasi Cloud campus in Lekki, representing over $2 billion in committed investment through 2027.
Nigeria’s 25 data centers place it second in Africa behind South Africa’s 61, but the country’s 30+ MW operational IT load represents just 0.6% of global installed capacity. This concentration reveals both the gap and the opportunity.
Energy reliability remains the binding constraint, as grid availability fluctuates between 5,000 MW and 6,000 MW against 13,000 MW installed capacity, forcing operators to self-generate at costs between $0.28-$0.33 per kWh.

For years in Nigeria, every AI training run and data request meant routing compute through distant AWS, Microsoft Azure, or Google Cloud Platform (GCP), incurring latency penalties, FX costs, and data sovereignty risks. That constraint is now being dismantled.

As of 2026, Nigeria has 25 operational data centers. Global players like Equinix are now competing with regional operators like Kasi Cloud and Nxtra to capture demand from fintechs, enterprises, and government agencies that are required to keep citizen data onshore.

This article documents exactly who is building Nigeria’s AI infrastructure and what founders should expect over the next 2 years.

A summary of Nigeria’s AI infrastructure

Total data center capacity (operational)	~86 MW
Capacity under construction/planned	320+ MW
Major operators	MainOne (Equinix), Rack Centre, 21st Century Technologies, Kasi Cloud, Nxtra by Airtel, ChamsCorp
Data sovereignty compliance status	Nigeria’s Data Protection Act supports local hosting
Estimated investment committed through 2027	$2 billion+ (includes private hyperscaler and institutional capital)
Fibre backbone	35,000 km currently; expanding to 125,000 km under the BRIDGE project

The compute gap that has been holding Nigerian AI back

When you understand the infrastructure deficit that Nigerian AI builders have been fighting, you’ll grasp why this moment matters.

Premium cost and latency limitations

Every time a Nigerian startup routed AI compute through foreign data centres, they paid egress fees and absorbed FX spreads on dollar-denominated cloud bills.

They also accepted latency penalties of roughly 70-90 milliseconds round-trip compared to the sub-5ms latency a local facility could deliver. That lag can be a deal-breaker for real-time applications, such as fraud detection, voice assistants, and autonomous systems.

Data sovereignty complications

Under the Nigeria Data Protection Act, fintechs, healthtech companies, and government-adjacent AI products must keep citizen data onshore. But when your compute runs offshore, compliance becomes impossible.

That tension has forced many founders to limit their ambitions to lightweight applications or inference-only products.

Poor regional performance

Most of these facilities in Nigeria were never engineered for AI workloads. Until Kasi Cloud’s 2026 launch, Nigeria had only one AI-capable data centre against South Africa’s five.

Workload constraints

Large model training, real-time inference at scale, and multimodal applications were economically unviable locally. So, Nigerian AI talent focused on what worked: lightweight APIs, fine-tuning offshore models, inference-only deployments. The comput-intensive work stayed abroad.

Demand pressure had since triggered investment. Although right now, Nigeria’s 25 data centres represent just 0.6% of global installed capacity, a single number that captures both the gap and the opportunity.

How local data centers are closing the GPU gap

As Nigerian enterprises scaled and fintechs like Moniepoint, M-PESA, and TymeBank reached millions of users, hyperscalers woke up to Africa’s strategic importance, and regulators kept pushing data localisation.

By 2025, the math flipped, and building local infrastructure became cheaper than the accumulated friction of doing everything offshore. Now, it seems the infrastructure pipeline is finally matching the rhetoric.

They are building AI-able facilities

New AI-focused facilities are coming online across multiple hubs.

Airtel’s Nxtra is building a 38 MW hyperscale facility at Eko Atlantic, purpose-built for AI workloads, with its first GPU shipment already received.
MTN’s Genova unit delivered the first phase of its Ikeja data centre in July 2025, with a second phase adding AI-optimised GPU infrastructure by late 2026 at a cost of $240–$250 million.
Kasi Cloud’s Lekki campus, backed by the Nigeria Sovereign Investment Authority, launches its first 5.5 MW AI-optimised phase in April 2026, supporting rack densities from 10 kW to 100 kW with liquid cooling.

They are leasing GPUs

GPU leasing and on-demand compute are becoming real.

Cassava Technologies, Africa’s first NVIDIA Cloud Partner, is deploying AI Factories across the continent (starting in South Africa, then Nigeria), offering GPU-as-a-Service and AI-as-a-Service to local developers.
In August 2025, Udutech launched the Africa GPU Hub in Lagos to offer GPU rentals, specifically optimized for the African market, with prices starting at less than $1 per hour.

They are partnering with global players

Partnerships with global cloud providers are deepening.

Equinix’s $22 million LG3 facility opens in Q1 2026 and integrates with Equinix Fabric for direct cloud on-ramps. This means Nigerian businesses can securely connect to AWS, Microsoft Azure, and GCP from Lagos with private, low-latency connections rather than routing through Europe.

The facility positions Lagos as a strategic interconnection hub for West Africa, linking subsea cables like Equiano and 2Africa to local enterprises.

They are attracting investment

Private capital is leading, with public backing.

The Nigeria Sovereign Investment Authority’s stake in Kasi Cloud signals a sovereign commitment, but most of the funding is private through 2027.
The AI-powered super app, Vesti, is pivoting into infrastructure with Zyra Data, a new division building a Tier 2 data centre scheduled to go live in late 2026. The $20 million facility combines debt and equity.

How far the gap has closed, and what remains

Significant improvements from 2023 to 2026 include:

Nigeria now has 30+ MW of operational IT capacity from 25 data centres, with more than 320 MW under construction or planned.
The fibre backbone is expanding from 35,000 km to 125,000 km under Project BRIDGE, backed by $500 million from the World Bank and $100 million from the EBRD.
Total investment committed through 2027 exceeds $2 billion.
The Nigerian data center Market is expected to grow from $374.05 million in 2026 to $782.8 million in 2031.

The timeline to regional parity is realistic

South Africa leads with 61 data centres and accounts for the bulk of Africa’s 360 MW of operational capacity.

Nigeria’s pipeline suggests 2-3 years to close the per-capita gap with Kenya (19 centres), and 2028-2030 to approach South Africa’s absolute capacity.

But gaps remain

Large-scale model training remains constrained until 100-MW facilities come fully online.
Real-time inference for latency-sensitive apps is improving, but not yet enterprise-grade across all sectors.
Multimodal workloads requiring high-density GPU clusters are still 12-24 months from widespread availability.

Structural bottlenecks persist

Power reliability. Grid fluctuates between 5,000 and 6,000 MW, against 13,000 MW installed, forcing operators to self-generate at $0.28- $0.33/kWh.
Technical talent for data centre operations. Microsoft’s training push addresses users, not infrastructure operators.
Cooling. Cooling for high-density GPU racks remains first-mover territory; Kasi Cloud’s liquid-to-chip technology is not yet the industry standard.

Who is building Nigeria’s AI infrastructure layer?

Here are the operators, investors, hyperscalers, and the connectivity backbone that make it all possible:

The flagship: Kasi Cloud

Kasi Cloud is a purpose-built AI data centre campus. Located on a 42-hectare site in Lekki, Lagos, the facility broke ground in April 2022 and is designed for AI from day one rather than retrofitted onto legacy infrastructure.

Key specifications:

Total investment: $250 million.
Phased capacity: 5.5 MW launching April 2026; scalable to 100 MW fully built out.
Rack density: Engineered for 10 kW to 100 kW per rack.
Cooling: Liquid-to-chip cooling with magnetic-drive technologies and triple HEPA filtration.
Backing: Nigeria Sovereign Investment Authority (NSIA) is a strategic partner.

Major Operators and Hyperscalers

Nxtra by Airtel

Nxtra is building a 38 MW hyperscale facility at Eko Atlantic, Lagos, as part of Airtel’s continent-wide network of nearly 50 edge and core data centres. The facility is designed for hyperscale cloud providers, enterprises, and AI workloads, with a 2028 go-live target.

MainOne (Equinix) and Rack Centre

Both operators are actively expanding capacity for higher-density workloads. MainOne’s Lekki data centre and Rack Centre’s Lagos facility serve as anchor colocation providers for Nigerian enterprises.

21st Century Technologies and ChamsCorp

These remain active players in Nigeria’s colocation and government-facing digital infrastructure space.

Investment and Talent Development

Microsoft

Training

In February 2025, Microsoft announced a $1 million investment to train Nigerians in AI and digital skills over two years, partnering with Tech4Dev and Data Science Nigeria. This complements a broader pledge to train 3 million Africans across South Africa, Kenya, Nigeria, and Morocco by 2026.

Infrastructure

Microsoft’s $1 billion Africa-wide AI and cloud investment (announced May 2024 in partnership with G42) includes a geothermal-powered data centre in Kenya. Nigeria-specific data centre details are still developing, but Microsoft’s fibre and cloud backbone increasingly supports Nigerian AI workloads.

Google

Google announced a $1.7 million grant in October 2024 to support AI talent development in Nigeria, part of a broader $5.8 million commitment to digital skills across Sub-Saharan Africa.

The foundation: Submarine cables and fibre networks

Hardware is only as valuable as the connectivity that powers it.

Submarine cable landings

The 2Africa and Equiano cables (both landing along the Lagos coast) dramatically strengthen the economics of connectivity. Equiano, Google’s private subsea cable, landed in Nigeria in 2022 and provides three times more network capacity than all previous cables combined. This enables local compute viability by reducing international bandwidth costs and latency.

Project BRIDGE

Nigeria’s flagship fibre backbone expansion, backed by $500 million from the World Bank and $100 million from the European Bank for Reconstruction and Development (EBRD), is expanding the national fibre network from 35,000 km to 125,000 km. Total investment committed through 2027 exceeds $2 billion.

National Broadband Alliance for Nigeria (NBAN)

Launched in February 2025, NBAN brings together government agencies, private sector leaders, and non-profits to drive broadband penetration toward the government’s targets: 70 per cent penetration by 2025, 25 Mbps minimum in urban areas, and 80 per cent population coverage by 2027

What GPU access looks like for Nigerian startups today

Most startups are still routing workloads through AWS, GCP, or Azure, paying egress fees, absorbing FX spreads on dollar-denominated cloud bills, and accepting 70-90ms latency penalties.

Kasi Cloud’s AI-optimised racks come online in April 2026 and will be a purpose-built AI infrastructure. Pricing hasn’t been published yet, but local colocation eliminates egress fees and FX spreads, which should reduce the total cost of ownership.

However, startups must also budget for IP transit and cross-connect fees to global cloud on-ramps, such as AWS Direct Connect or Azure ExpressRoute via Equinix Fabric, which remain significant line items despite egress savings. The wild card is power, as operators self-generate a structural cost that shapes every pricing decision.

In practice, Nigerian startups run hybrid models. Lightweight inference happens locally where possible: fintechs with strict data residency may colocate at Rack Centre or MainOne for storage. But training runs still go offshore. The developer experience gap is real; even where GPU infrastructure exists locally, tooling, MLOps support, and ecosystem maturity lag the US and EU regions by years.

FAQs

Can Nigerian startups access GPUs locally in 2026?

Yes, driven by a surge in AI-ready data center construction and the deployment of local GPU-as-a-Service (GPUaaS) platforms.

How do local data center costs compare to AWS or Azure?

Local colocation eliminates egress fees and FX spreads, but energy costs (self-generation) mean the total cost of ownership depends on workload type.

Is Nigeria’s power grid reliable enough for data centers?

Unfortunately, No. Grid availability fluctuates between 5,000 and 6,000 MW, against an installed capacity of 13,000 MW.

Conclusion

Nigeria’s AI infrastructure gap is closing, measurably, and at a pace that reflects both private capital conviction and strategic national intent. With 86 MW operational, 320+ MW in pipeline, and a purpose-built AI-optimised facility launching in April 2026, the foundation for local compute is being laid.

But the gap between investment announcements and developer-accessible infrastructure remains wide. Energy reliability still imposes structural costs. Enterprise adoption cycles remain gradual. And the tooling and talent ecosystem (e.g., data centre operations engineers, MLOps specialists, cooling experts) lags more mature markets by years.

The current trajectory, if sustained, will produce a genuinely competitive AI development environment in Nigeria within 2-3 years.

Citations

Disclaimer!

This publication, review, or article (“Content”) is based on our independent evaluation and is subjective, reflecting our opinions, which may differ from others’ perspectives or experiences. We do not guarantee the accuracy or completeness of the Content and disclaim responsibility for any errors or omissions it may contain.

Related Story: Responsible gaming: How AI can aid safer play?

The information provided is not investment advice and should not be treated as such, as products or services may change after publication. By engaging with our Content, you acknowledge its subjective nature and agree not to hold us liable for any losses or damages arising from your reliance on the information provided.

Always conduct your research and consult professionals where necessary.