Who Owns Africa’s Compute? The Value Chain Behind the Data-Centre Gold Rush

By /

Africa has less than 2% of global data-centre capacity, and MTN’s own leadership puts the continent’s digital-infrastructure funding gap at roughly US$96 billion to 2030 (MTN Group). That gap is now being filled by a scramble of very different players with very different economics. Understanding how they fit together — and who holds the advantage — matters more than any single headline megawatt number. This piece maps the value chain, is candid about where the advantage sits, and looks hard at the role telcos will actually play.

1. The value chain is unbundling — six roles, not one business

The defining feature of this market is disaggregation. What used to be one vertically integrated function is now six distinct roles, each with its own capital profile and return expectations:

Article content
Key value chain map

The strategic point: value no longer accrues to whoever “owns the building.” It accrues to whoever controls the scarcest link — and in Africa that is rarely the concrete.

Article content

Africa’s data-centre value chain: hyperscalers, global developer-operators, local operators, the fibre-connectivity backbone, financial investors and telcos, plus the multi-operator capital-recycling playbook

2. The “buildco” model: how capital recycles through the chain

The engine connecting these roles is the buildco model. A developer with a standardised campus design and deep capital-markets relationships builds fast, signs a hyperscaler on a long lease, then recycles equity by bringing in — or selling down to — financial investors, while retaining operations. The asset behaves like real estate: shell + power + a long lease = a predictable, financeable yield.

This is now an established global playbook, not a single-operator trick. The same recycling logic recurs across every major platform, in several structural forms:

  • Stabilised-asset JVs and sell-downs. Vantage entered Africa with a US$1 billion, 80 MW campus in Johannesburg’s Waterfall City (Vantage) and globally raised a US$9.2 billion equity round (DigitalBridge, Silver Lake), part of a record US$13 billion in 2024 funding more than US$30 billion / 3 GW of development (Vantage). Crucially for Africa, GIC and MEAG committed €1.4 billion (~US$1.5 billion) into Vantage’s EMEA platform — the vehicle that houses its African campus (DCD), followed by a US$1.6 billion GIC- and ADIA-led injection into its APAC platform (Vantage). Digital Realty runs the identical play: a joint venture on stabilised hyperscale data centres in Chicago (Digital Realty) and a US$3.25 billion US Hyperscale Fund (supporting up to US$10 billion of investment) in which it keeps a minority stake plus management fees (Digital Realty). Equinix built its xScale program the same way, jointly investing more than US$7.5 billion with GIC since 2019 and forming a greater-than-US$15 billion US JV with GIC and Canada’s CPPIB in 2024 — institutional partners typically own ~80% while Equinix operates (DCD).
  • Debt and securitisation against leased assets. CloudHQ is raising US$1.4 billion in asset-backed securities secured on two Virginia campuses leased to hyperscalers (DCD); Iron Mountain executed a US$178 million sale-leaseback of UK sites explicitly badged as “capital recycling” (DCD).
  • Whole-platform exits and permanent capital. Where single assets aren’t sold, whole platforms rotate to long-term owners: Aligned Data Centers sold to an AIP / MGX / BlackRock GIP consortium at a ~US$40 billion enterprise value (Macquarie); QTS went to Blackstone (~US$10 billion) and CyrusOne to KKR and GIP (~US$15 billion) (sector M&A record); and Compass Datacenters signed an evergreen permanent-capital partnership with KKR (Compass).

Two structural features decide whether the recycling actually works:

  • Shell vs core. Financing usually splits the “shell” (land, building, power, cooling) from the “core” (servers, chips, racks). In AI-focused facilities the core can be roughly four times the value of the shell, sharply raising capital intensity versus classic colocation (Norton Rose Fulbright).
  • Exit liquidity. Recycling only works if someone will buy the stabilised asset. Single-asset transaction volumes hit back-to-back records in 2024 and 2025, and hyperscale “exits” are spreading into newer markets — South Korea in 2024, Malaysia in 2025 (MSCI). The caveat for Africa: this exit market barely exists on the continent yet — which is precisely why early movers who can manufacture stabilised, financeable assets stand to capture a scarcity premium.

3. Do global operators have a structural edge over local players?

Mostly yes — but not everywhere, and the exceptions are the whole investment story.

Where globals win (durable advantages)

  • HQ-level, globally-aligned hyperscaler relationships. A master agreement negotiated in Seattle or Dublin cascades into African allocations. A local player wins the same tenant deal-by-deal, market-by-market, with none of the portfolio leverage. This is the single biggest edge.
  • Cost and structure of capital. Globals tap green loans, securitisation and multi-billion-dollar equity platforms at a blended cost locals cannot match (Vantage’s US$13 billion in a single year is the proof point).
  • Standardisation and speed-to-market, and the balance-sheet credibility to underwrite decade-long power commitments.

Where locals win (why they aren’t simply steamrolled)

  • Power, land, permits and grid navigation — the true African bottleneck. South Africa’s data-centre “gold rush” is already bumping against power and water limits, and regulators are scrutinising inflated “bragawatt” claims (CNBC Africa / Reuters, 2026). Local incumbency here is a genuine moat.
  • Carrier-neutral ecosystem and interconnection density — Teraco’s value was precisely its ecosystem, which is why Digital Realty bought it rather than building from scratch (TelecomTV).
  • Edge and secondary markets — the regional colo demand in Kampala, Kinshasa or Abidjan that is too small for a hyperscale campus but is Raxio’s and OADC’s core. Raxio has now committed over US$380 million, backed by Roha and Meridiam, after a sixfold surge in contracted capacity in H1 2026 (Data Centre Magazine, 2026), with DFI debt from the Emerging Africa & Asia Infrastructure Fund.
  • Sovereignty and regulatory alignment — data-residency rules favour a credibly local story.

The tell: the globals’ preferred route into Africa has been acquiring locals, not greenfield — Digital Realty → Teraco (majority stake at a US$3.5 billion valuation, since raised toward 77%) and Equinix → MainOne (US$320 million) (CRN). The market is pricing exactly those local advantages that money alone can’t quickly replicate. So the honest framing isn’t “global vs local” — it’s global capital and hyperscaler relationships marrying local land, power and ecosystem. The winning local position is being the platform worth acquiring or partnering with.

4. Where telcos fit — and the two things they’re doing at once

Telcos are the wildcard, and their strategy is visibly bifurcating:

  • Exiting the legacy, non-core estate. MTN is exiting two legacy South African data centres in Johannesburg and Cape Town — enterprise-era facilities, not AI-ready (Data Center Dynamics, 2025).
  • Re-entering at the AI frontier — with partners. Simultaneously, MTN is courting US and European partners and has broken ground on a ~US$240 million AI data centre in Nigeria (Mobile World Live). Airtel already runs this as a separate platform (Nxtra).

So a telco’s legacy-DC options reduce to four: carve out and sell to a specialist; sale-and-leaseback to free capital while keeping the space it uses; contribute assets into a JV with a global or regional operator; or retain only edge nodes for its own network and outsource the rest.

5. The tower parallel — and why it only half-holds

The instinct is to map this onto the African tower story (IHS, Helios, ATC): telcos carved out passive infrastructure, sold it to independent operators, freed capital, and leased it back. It’s a useful analogy — and a misleading one.

Where it holds

The carve-out → independent operator → sale-leaseback → shared tenancy playbook is financially identical: turn a cost centre into a shared, yield-bearing asset the infrastructure funds want to own (Data Center Dynamics).

Where it breaks down — three ways

  1. A tower is passive; a data centre is an operating business. Power procurement, cooling, uptime SLAs, interconnection and AI fit-out make it far less “pass-through” — and the shell-vs-core split (core ≈ 4× shell) is a complexity towers never had.
  2. The reversal risk is real. MTN’s ~US$2.2 billion move to take control of IHS Towers (Total Telecom, 2026) has prompted the pointed question of whether these sale-and-buyback cycles are strategy or financial engineering. Any telco carving out DCs today should ask whether it is structurally exiting or merely renting back an asset it will want to own again.
  3. The anchor-tenant asymmetry — the crucial difference. In the tower model, the telco remains the anchor tenant with predictable, contracted, decade-long demand: that guaranteed cash flow is exactly what makes a tower portfolio bankable. Data centres are the opposite. Telco in-house demand is shrinking, not committed. Operators have spent a decade moving IT and business-support systems into the public cloud — for some, hyperscale cloud is now cheaper than running their own private data centres (Amdocs) — and network workloads are following: AT&T runs its 5G core in Microsoft Azure and DISH built its entire 5G network on AWS (Urgent Communications; AWS). Migration of network functions is still early — STL Partners counts only ~12 production 5G-core deployments on public cloud, and some telcos resist outright (“I will never be putting our core on a hyperscaler,” per Verizon Business) (STL Partners; Light Reading) — but the direction of travel is unambiguous: virtualisation and AI are reducing telcos’ own physical footprint, not anchoring it.

That asymmetry is the insight investors should internalise: a telco data-centre carve-out does not come with the built-in anchor-tenant demand that made the tower model financeable. Its value must be underwritten on third-party demand — hyperscaler, enterprise, or sovereign — from day one.

6. The connectivity layer: continental fibre as enabler — and target

There is a layer beneath the data centre that decides whether any of this works: the terrestrial fibre backbone carrying traffic between subsea landing stations, cloud on-ramps and inland compute. Africa’s terrestrial fibre reached roughly 1.3 million km by 2025, but it remains fragmented and unevenly distributed (Africa Data Centre Association). Three continental-scale players dominate the independent long-haul layer:

  • Liquid Intelligent Technologies (Cassava) — Africa’s largest independent network at ~110,000 km across 25+ countries, including the first terrestrial Cape Town–Cairo route (Liquid).
  • Bayobab (wholly owned by MTN, formerly GlobalConnect) — ~141,000 km of open-access fibre reaching 54 countries, expanding via the US$320 million Project East2West with Africa50 (MTN). MTN has been consolidating its fibre into Bayobab and positioning it as an open-access platform able to take on external partners (Light Reading).
  • Telesonic (Airtel Africa) — a fibre-wholesale carve-out managing Airtel’s ~75,000 km terrestrial network across 14 markets, plus subsea capacity (SubTel Forum).

Note the pattern: Bayobab and Telesonic are themselves telco carve-outs — the same “unbundle the passive infrastructure into an open-access platform” logic we saw in towers and data centres, now playing out in fibre. That makes them natural partnership or investment targets, and natural interconnect partners for data-centre operators who need backhaul to landing stations and cloud regions.

Does financial stress make a player like Liquid a target? Liquid spent a fraught 18 months staring down a US$620 million bond due September 2026 and a rand term loan due February 2026 — described locally as a “fight to avoid financial ruin” (TechCentral). It has since executed a substantial deleveraging: a US$195 million equity injection from Cassava, new credit facilities, an NVIDIA equity investment, and a STANLIB minority stake in sister company Africa Data Centres whose proceeds cut Liquid’s debt (Cassava), capped by a US$300 million Eurobond oversubscribed 2.5x within an ~US$855 million recapitalisation (Africa Global Funds) and a Fitch upgrade to ‘B-‘ (Fitch). The lesson cuts both ways: leverage on a capital-intensive, pan-African fibre asset created real refinancing risk and, at the trough, take-out optionality — yet the heavily oversubscribed bond shows the asset is prized even when the balance sheet is stretched. The target thesis is real, but the window narrows as these players deleverage and pivot toward AI-adjacent revenue.

Will hyperscalers keep investing in African fibre? They already are — and the trajectory mirrors their global playbook of owning strategic subsea and backhaul rather than leaving it to carriers:

  • Google built Equiano (Portugal–South Africa, with branches to Nigeria, Togo and Namibia), its 14th subsea investment (Submarine Networks), as part of a US$1 billion, five-year Africa commitment (Quartz). It followed with Umoja, the first direct Africa–Australia route — and, tellingly, built Umoja’s terrestrial leg through Kenya, Uganda, Rwanda, the DRC, Zambia, Zimbabwe and South Africa in partnership with Liquid (Google Cloud). Google is also adding four subsea-cable landing/interconnect hubs on the continent (Techpoint).
  • Meta completed the core of 2Africa — at 45,000 km and 180 Tbps, the world’s largest subsea cable system — in November 2025 (DCD), with Bayobab among its landing partners.

The strategic read: hyperscalers self-provision the subsea and long-haul routes that are core to their latency and resilience, but they partner with continental fibre players for the terrestrial reach they don’t want to build. That makes Liquid, Bayobab and Telesonic enablers first — the terrestrial complement to hyperscaler subsea — and targets second, where balance-sheet stress or strategic fit makes an equity stake or acquisition cheaper than building. Expect the same “own the strategic core, partner or buy for local reach” pattern that defines the data-centre layer to repeat in fibre.

7. What this means for investors and operators

Four success factors decide outcomes across the chain — each an underwriting question before it is an engineering one:

  1. Contracted demand quality. Whose demand underpins the asset, for how long, and how credible is it — especially where the telco seller is not the anchor?
  2. Power. Firm, affordable, ideally clean power for 10–15 years remains the largest determinant of viability.
  3. Capital structure and partnership design. The right pairing of global capital/relationships with local land, power and ecosystem — via JV, sale-leaseback, or acquisition — is the core value-creation lever.
  4. Ecosystem and interconnection. Carrier-neutrality and cloud on-ramps, not raw megawatts, drive pricing power.

Generic capital will chase the Tier-1 hubs (Johannesburg, Nairobi, Lagos). The outsized returns will accrue to those who can correctly price the partnership structure — and the demand risk — that generic models miss.

8. How Digitalthings assists

This is the terrain Digitalthings has worked for over a decade. We are a boutique, investment-and-strategy-focused advisory firm specialising in digital infrastructure, cloud and broadband — combining deep African local insight with global best practice, and assigning senior practitioners to every mandate. Our founder led the Africa digital-infrastructure practice of a global engineering consultancy before establishing the firm; our track record spans the exact assets and transactions this value chain is built on.

We support hyperscalers, operators, telcos, investors, governments and DFIs across the transaction and asset lifecycle:

  • Transaction & investment advisory — technical and commercial due diligence, red-flag reviews, and investment-grade techno-economic models for data-centre, fibre and tower assets. We have delivered a red-flag review of an FTTH operator for a leading African tower company, technical and commercial due diligence on a regional fibre operator spanning Uganda, Zambia, Kenya, Rwanda and the DRC, and supported a leading South African fibre operator’s acquisition of a metro fibre network.
  • Strategy & business planning — market sizing, open-access and wholesale strategy, partnership/JV structuring, and financial modelling. We have built open-access FibreCo business plans and financial models across six African markets and advised a major Southern African fibre holding group (parent of leading FTTH and dark-fibre operators) on strategic review.
  • Feasibility & market entry — including cable-landing-station feasibility across Liberia, Togo, Guinea Conakry and Côte d’Ivoire, and open-access fibre market entry for a pan-African wholesale fibre operator (DRC).
  • Programme & interim management — we led programme management and transaction advisory for a national open-access broadband rollout in Malawi, and served as interim CEO of a metropolitan open-access broadband SPV (a South African municipality’s vehicle, with mobile-network anchor tenants).
  • DFI & institutional advisory — supporting infrastructure investors and development finance institutions (including engagements linked to leading tower companies, African infrastructure private-equity firms and a US development finance institution) on the digital-infrastructure opportunity.

If you are underwriting a data centre, evaluating a telco carve-out, or structuring a local–global partnership in Sub-Saharan Africa, the difference between a good and a bad outcome is local, senior, investment-grade diligence — and a clear-eyed view of who actually holds the demand. That is what we do. Let’s talk.

Prepared by Digitalthings. Sources are hyperlinked inline. Deal values and market estimates vary by methodology and reporting date; figures cited reflect the most recent publicly available data as of mid-2026 and should be validated against primary sources before transaction use.

Sources

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top