Aggregating hyperscaler, operator and private-capital pipelines announced 2023–H1 2026.
Top
AWS / Azure / Google cloud regions · 3.5
33.9% of total
Bottom
Equinix Africa builds · 0.6
5.8% of total
Average
1.7
6 categories
Total
10.32
Sum of series
| Series | Microsoft–G42 Kenya | Cassava AI factories (5 mkts) | Africa Data Centres expansion | Equinix Africa builds | AWS / Azure / Google cloud regions | Telco & PE pipelines (Liquid, MTN, others) |
|---|---|---|---|---|---|---|
| Value | 1 | 0.72 | 1.5 | 0.6 | 3.5 | 3 |
Source: Company announcements and press coverage 2023–H1 2026; IdeaToola triangulation of disclosed figures. Pipelines, not deployed capital.
VerifiedKey Findings
- 1H1 2026 marked the inflection: Cassava Technologies' NVIDIA-powered 'AI factory' came online in South Africa with expansion underway into Nigeria, Kenya, Egypt and Morocco — the continent's first multi-market GPU-as-a-service footprint (Cassava/NVIDIA announcements, 2025–2026).
- 2Microsoft and G42's $1B Kenya digital ecosystem investment — anchored on a geothermal-powered data-centre campus in Olkaria — moved from memorandum to construction through H1 2026, with initial cloud regions targeted for 2027.
- 3Cumulative committed digital-infrastructure investment in Africa passed $10B by mid-2026 when hyperscaler, telco and private-equity data-centre pipelines are aggregated (IdeaToola triangulation of disclosed projects).
- 4Africa Data Centres, Equinix, Digital Realty and Open Access Data Centres all expanded South African and Nigerian capacity in H1 2026; announced continental capacity now targets a doubling of installed MW by 2028.
- 5The binding constraint is power, not capital: data-centre developers are increasingly pairing campuses with dedicated renewables and gas, echoing the Olkaria geothermal model (Kenya) and wheeled-solar structures (South Africa).
- 6The strategic stakes: compute location determines where African AI models can be trained, where data sovereignty can be enforced, and which economies capture the AI services value chain rather than importing it.
The Sprint: From Announcements to Steel in the Ground
For three years, African AI infrastructure was a story of intent. H1 2026 is when intent became construction. The signal event was — Strive Masiyiwa's pan-African technology group — switching on its first NVIDIA GPU cluster in South Africa, the opening move in a five-market 'AI factory' rollout announced with NVIDIA in 2025 and targeting Kenya, Nigeria, Egypt and Morocco through 2027. For the first time, African researchers, banks and governments can rent frontier-class compute *on the continent*, billed in local currency, without exporting data.
In parallel, the Microsoft–G42 Kenya programme — a $1B digital ecosystem commitment anchored on a geothermal-powered campus at Olkaria, Naivasha — advanced from MOU to site works through the first half of 2026. The pairing is deliberate: Kenya's grid is ~90% renewable, and geothermal baseload solves the 24/7 power profile that solar alone cannot. It is the template every serious African data-centre pitch now copies.
Aggregate the disclosed pipelines — hyperscalers, ' expansion programme, 's builds in Johannesburg and Lagos, Digital Realty, OADC, and telco-led facilities from MTN and Liquid — and cumulative committed investment passed $10B by mid-2026, with announced capacity on track to roughly double installed megawatts by 2028.
The Binding Constraint: Electrons, Not Capital
Every developer we track now leads with power strategy before rack strategy. A modern AI data hall draws 50–100MW continuously; South Africa's grid, though stabilised post-2024, cannot host that load at scale without dedicated generation, and Nigeria's grid supplies a fraction of commercial demand. The industry's answer, visible across H1 2026 announcements, is *behind-the-meter* power: Olkaria's geothermal in Kenya, wheeled solar-and-wind in South Africa, and gas-to-power hybrids in Nigeria and Egypt.
This is quietly restructuring the economics. Data centres that arrive with their own power effectively build private utilities — with spillover benefits (grid stabilisation, industrial parks, green-hydrogen feedstock) that host governments are learning to price into approvals. Kenya and South Africa have moved fastest to streamline energy-wheeling and land approvals for exactly this reason.
Africa's Emerging Compute Hubs — H1 2026 Status
| Market | Anchor development | Power strategy | H1 2026 status |
|---|---|---|---|
| South Africa | Cassava AI factory; Teraco/ADC/Equinix campuses | Wheeled renewables + grid | Live & expanding |
| Kenya | Microsoft–G42 Olkaria campus; iXAfrica | Geothermal baseload | Construction |
| Nigeria | Equinix Lagos; OADC; Rack Centre | Gas hybrid + solar | Construction |
| Egypt | Cassava target; government DC programme | Grid + Benban solar | Planned |
| Morocco | Cassava target; sovereign cloud zone | Wind (Atlantic coast) | Planned |
The five markets where AI-grade capacity is being built first.
Source: Company disclosures; IdeaToola compilation, H1 2026.
Verified"Compute is the new port. The economies that land GPU capacity in 2026–2028 will host the AI services layer for the decade after; everyone else rents it back."
— IdeaToola Intelligence, H1 2026 AI infrastructure review
So What: Who Captures the Value
The infrastructure sprint answers the sovereignty question but opens a commercial one: who fills the racks? H1 2026 showed three demand engines forming. First, banks and telcos — African financial institutions are the earliest large buyers of local GPU capacity, driven by data-residency rules and fraud/credit model training. Second, government AI programmes — Kenya, Egypt, Rwanda and South Africa all advanced national AI strategies referencing local compute. Third, global model labs seeking inference footprint closer to 400m+ smartphone users.
Watch the interconnection layer in H2: submarine cable landings (2Africa, Equiano) plus metro fibre determine whether Lagos, Nairobi and Johannesburg can serve the region, or only themselves. The sprint is real; the network effect that makes it compounding is still being built.
So What? — Strategic Implications
What decision-makers should do about it
Enterprise buyers should negotiate multi-year SaaS contracts now — AI-driven pricing will inflate renewal costs 20–30%.
Cloud migration should prioritise data residency compliance; 14 African markets now have localisation requirements.
Build internal AI/ML capability rather than outsourcing — competitive advantage accrues to firms that own their models.
Strategic recommendations based on IdeaToola Research analysis. Not financial advice.
Predictive Outlook — What Happens Next
Forward-looking analysis · 2026–2031 trajectory
What Happens Next
By end-2026, ~40% of enterprise applications will integrate task-specific AI agents — up from <5% in 2025 (Gartner, 2025).
By end-2027, Gartner expects more than 40% of agentic AI projects to be cancelled on cost, value and governance grounds — winners will be the minority that scaled past pilot.
By 2028, 33% of enterprise software will ship with embedded agentic AI; orchestration and vertical-agent layers capture the durable margin while foundation-model pricing keeps commoditising.
Scenario Modeling
If governance and identity standards (NIST, ISO) mature for autonomous agents
Cancellation rate falls below 25% and enterprise-scale deployments double in regulated sectors (financial services, healthcare).
If foundation-model pricing keeps falling 60–80% per year while capability holds
Per-task agent unit economics flip positive at lower scale; vertical agents in revenue ops and service become the default buy.
If a high-profile autonomous-agent failure triggers prescriptive regulation in the EU or US
Mandatory human-in-the-loop checkpoints for high-stakes actions; enterprise rollouts slow by 12–18 months but trust improves.
Trend Trajectories · 2026–2031
Apps integrating task-specific AI agents (Gartner)
33%+ of enterprise software (2028 anchor)
Agentic AI projects cancelled by 2027 (Gartner)
40%+ of in-flight projects
Organisations scaling a GenAI use case enterprise-wide (McKinsey)
From ~23% in early 2025 to majority by 2028
Share of agentic spend in orchestration + vertical layers (IdeaToola estimate)
~65% of stack spend
Build the Strategy
Turn these predictions into action. Our execution playbooks provide step-by-step frameworks with timelines, owners, and KPIs.
How to Scale SME Acquisition Digitally (Africa Edition)
Banking · Advanced · 12-week sprint
Building STP Onboarding in Emerging Markets
Fintech · Expert · 16-week build
Forward-looking projections based on current market trajectories, institutional research, and IdeaStack analysis. Scenarios represent possible futures, not predictions. Actual outcomes may vary based on regulatory, economic, and technological factors.
Untapped Market Opportunities
Commercial Rooftop Solar
South Africa has 420M m² of underutilised commercial rooftop space. Current 1.2GW installed could grow 6× with wheeling framework maturity.
< 5% of commercial rooftops utilised
R28B
Source: DMRE & GreenCape Market Intelligence Report, 2025
SME Embedded Lending
Embedded lending APIs integrated into accounting platforms could unlock a massive underserved segment with 94% of SMEs relying on informal financing.
Only 6% of SA SMEs have formal credit access
R42B
Source: SARB & FinMark Trust FinScope SME Survey, 2024
Digital Freight Matching
AI-powered load matching across SA's 280,000 trucks could eliminate R14B in wasted capacity annually.
38% of trucks return empty
R14B
Source: Transnet & Road Freight Association, 2024
Ratings and debt metrics reflect latest publicly available data (2025–2026), with some countries undergoing active restructuring. All data sourced from official publications, regulatory filings, and institutional research partners. Figures are indicative and may be subject to revision. Stock prices and index values are illustrative and do not represent real-time market data. IdeaToola does not provide financial advice. Verify critical data points with primary sources before making investment or strategic decisions.