UbuntuNet’s 10 Gbit/s Core Still Depends on Campus Networks

Summary

• UbuntuNet is a not‑for‑profit organisation registered in Malawi and governed by its members, serving the national research and education networks of Eastern and Southern Africa; its documents refer to both an “association” and a “Trust”.
• It has assembled regional capacity through operator contracts, points of presence, member routes, and the AfricaConnect programmes, notably including core links at 10 Gbit/s and larger national connections.
• Its portfolio now extends beyond transport: identity, cloud and storage, open science, cybersecurity, and NREN development, with services at varying maturity levels.
• Pooling demand and expertise creates value, but weak campus links, unpaid bills, supplier power, conflicts, grant cycles, and incomplete reporting can prevent capacity from reaching researchers.

Network 3 brought the core into the 10 Gbit/s class

The 2023 report describes a migration from links of about 2.5 Gbit/s to primary links of 10 Gbit/s. It also presents the network as offering at least 20 Gbit/s of total capacity thanks to a redundant architecture. These statements relate to different layers and must not be reduced to a “single throughput”.

The most consistent interpretation is of redundant paths or links at 10 Gbit/s within the core in question, not a universal 20 Gbit/s link between every PoP or to every member. Actual capacity depends on the segment, the provider, and the national access point.

The change remains major. It gives NRENs more headroom, permits better science flows, and reduces the cost per unit. Yet it remains far from the multi‑hundred‑gigabit or terabit backbones of wealthier regions. The comparison must explain the regional economics, not set up a simplistic league table.

The unit‑price drop is an infrastructure outcome

The 2023 report indicates unit‑price reductions of between 50% and 71.5% compared with 2022. NRENs that previously had roughly 130 Mbit/s to 4 Gbit/s were able to access services ranging from 500 Mbit/s to 8 Gbit/s depending on the case.

This reduction comes from pooling demand, long‑term contracts, and better infrastructure utilisation. It shows that the mission is not merely to increase a capacity figure: it is to buy more efficiently so that national networks can pass the benefit on to their institutions.

The trade‑off is a tighter financial margin for the Alliance. When purchasing savings are passed on to members, less cushion remains to absorb an incident, an exchange‑rate shock, or a late contribution. A lower price is only sustainable if recurrent costs are understood and paid.

BotsREN and ZIMREN show how regional upgrades reach national networks

BotsREN was first connected at 10 Gbit/s in 2023, later upgraded to 20 Gbit/s in January 2026. ZIMREN joined the backbone at 10 Gbit/s in June 2025. These figures are national access speeds, not the speed of the entire network.

Nevertheless, they demonstrate that core investment can translate into identifiable capacity at the member level. The final value then depends on how many institutions are connected behind the NREN, their domestic circuits, their equipment, and their ability to use the link.

An NREN with a 20 Gbit/s handoff can still serve campuses constrained by local fibre, a firewall, or a budget. National announcements must therefore be accompanied by measurements of utilisation, loss, availability, and institutional coverage.

The regional layer begins where national networks reach their borders

A student or a researcher rarely sees the name UbuntuNet Alliance when accessing a service. The first visible network is the campus or research institute network. Traffic then passes through a national research and education network such as KENET, RENU, TENET, ZAMREN, TERNET, or MoRENet. UbuntuNet becomes decisive at the next stage: when a national network needs to exchange with another country, join a large scientific facility, reach GÉANT, or benefit from a shared service that is too expensive to replicate locally.

This position explains why the Alliance can be essential without resembling a consumer internet provider. It does not generally sell residential broadband, does not directly connect every university, and does not control local Wi‑Fi. It provides the regional layer of routing and services on top of autonomous national systems. Other regional networks, operators, clouds, exchange points, and the destination institution then complete the path.

The user perceives a single session; operational authority is divided. UbuntuNet can change its routes, upgrade a core link, monitor a regional service, and coordinate an incident. It cannot restore campus power, fix an outdated institutional directory, force a national operator to repair fibre, or guarantee the capacity of a remote application. The point, then, is not to pretend a single owner, but to arrange reliable interfaces between distinct responsible parties.

Federation is the model, not a provisional substitute for centralisation

UbuntuNet Alliance is an alliance of NRENs, not a holding company. Each member retains its legal personality, its teams, its budget, its domestic infrastructure, its university customers, its public relations, and its pricing. The regional level exists because these organisations choose to purchase certain capacities together, to exchange traffic at common locations, to run shared services, and to advocate for convergent interests.

This model suits a region where telecom markets, currencies, university regimes, procurement rules, languages, and technical maturity vary widely. A centrally owned academic network across all countries would assume a political authority that UbuntuNet does not possess and does not need. Federation lets national autonomy and regional scale be combined, while allowing experience to flow from stronger NRENs to those still building.

Complexity does not disappear; it moves to the interfaces. A slow transfer can be affected by a university firewall, a national circuit, a regional PoP, an international operator, and the remote system. Each party sees only part of the evidence. Hence the need for service definitions, shared telemetry, escalation contacts, change procedures, contribution rules, and workable exit paths.

The legal status is workable, but public terminology is not uniform

Current documents present UbuntuNet as a not‑for‑profit association registered in Malawi. The 2023 annual report and transitional texts also speak of a non‑profit Trust governed by members and Trustees. The file did not retrieve the most recent statutory deed; it would therefore be rash to resolve the discrepancy by assertion.

Practical governance is clearer. UbuntuNet has members, not ordinary shareholders. The documented structure brings together a Members’ Assembly, a Board of Trustees, and a Secretariat led by the CEO. The organisation can employ staff, receive grants, sign contracts, lease or own equipment, and operate services. Nothing points to distributable ownership or conventional commercial valuation.

The nuance matters. Calling it a private telecom company would suggest a shareholder‑return logic. Reducing it to a voluntary community would underestimate its contractual and fiduciary obligations. The safest formulation remains: not‑for‑profit membership body registered in Malawi, while noting the parallel use of “association” and “Trust” in its own records.

The founding problem was about economics and route direction

Universities in the region already had internet access before UbuntuNet was created. What was missing was a sustainable regional mechanism for aggregating academic demand, carrying scientific traffic, and joining global R&E networks on terms designed for education and research. Institutions were buying small international capacities individually, often at high unit cost, while communications between African countries sometimes detoured via Europe.

A commercial connection was useful, but it did not automatically provide direct routes between neighbouring NRENs, predictable paths for data‑intensive science, federated identity, common security, or collective purchasing power. Small communities negotiated weakly, and emerging NRENs struggled to demonstrate their usefulness before they had enough members and capacity.

The Alliance changed the economic unit. Campuses aggregate demand at national level; national networks pool it at regional level. This larger community can plan PoPs, sign long‑term contracts, organise operations teams, and negotiate with global partners. UbuntuNet does not eliminate the commercial market: it enters it as a coordinated buyer and a specialist mission‑driven operator.

Five NREN initiatives formed the first coalition

UbuntuNet Alliance was conceived in the second half of 2005 by initiatives in Kenya, Malawi, Mozambique, Rwanda, and South Africa. The founding base is associated with KENET, MAREN, MoRENet, RwEdNet, and TENET, under the institutional forms of the time. They did not have the same resources or maturity.

TENET and KENET brought operating experience and an established university base. MAREN would later anchor the Secretariat in the host country. MoRENet represented a state‑supported national model. RwEdNet would go on to illustrate both the promise and the fragility of a building NREN: its collapse and later reconstruction would show that the regional layer cannot sustainably substitute for a viable national institution.

This diversity was a strength. A coalition consisting only of mature networks might have remained an interconnection club. A coalition of overly fragile initiatives would have lacked traffic, engineering, and credibility. The mixture allowed real production to be built while developing members’ institutional capacity.

Francis Tusubira helped turn a technical proposal into an institution

Francis “Tusu” Tusubira is closely associated with the founding period and the movement to create African NRENs. The work was not simply about choosing routes or equipment. It required convincing universities, ministries, regulators, operators, funders, and engineers that a regional research network had to be a durable institution, not a temporary project.

It required establishing a legal personality, membership rules, accounts, contractual capacity, a recurring conference, and a shared explanation of how an NREN differs from mere discounted internet access. Tusubira’s contribution belongs to this collective construction; it must not erase the role of the founding networks, the Boards, the technical communities, and the international partners.

The lasting result was a body capable of receiving programme funds, negotiating long‑term capacity, and representing a regional community. That institutional capacity is what made the AfricaConnect phases possible.

Amsterdam provided a first legal vehicle; Malawi became the regional home

Registration documents were filed in Amsterdam in 2006. For an organisation still forming, a Dutch structure offered a familiar framework for European partners and eased contracts, projects, and international relations before a durable base could be established within the region.

As the Alliance matured, a European domicile became less coherent with its mission. A new constitution was adopted in Malawi in 2012, registration followed in 2013, and the archives describe a transfer of assets and commitments to the Malawian structure. The Secretariat is now in Lilongwe.

This shift brought governance closer to members, while leaving a terminological ambiguity between association and Trust. The history must therefore be told as a process: conception in 2005, initial formalisation in the Netherlands, then legal and administrative re‑planting in Malawi.

The first transit to GÉANT proved usefulness before a full backbone existed

Records indicate that an R&E transit to GÉANT was implemented in 2008. Even limited, this service gave members a path to European and global scientific networks. It showed that a regional organisation could deliver operational value before assembling a mature continental backbone.

The GÉANT relationship went beyond connectivity. European partners brought expertise in procurement, project management, identity, security, and operations, together with access to EU funding mechanisms. This cooperation accelerated regional building, while creating a lasting dependence on the timetables and priorities of external programmes.

The fairest reading is neither wholly imported infrastructure nor complete independence. UbuntuNet used global partners to build regional capacity governed by its members. The strategic question remains what portion of that capacity can become recurrent after projects end.

Long‑term contracts turned pooled demand into infrastructure

Small annual contracts expose a regional network to price increases, low capacity, and negotiating breakdowns. The fifteen‑year agreements signed with WIOCC in 2013 and SEACOM in 2016 gave UbuntuNet a longer horizon for arranging its routes and pricing.

A long‑term commitment can lower unit cost, stabilise planning, secure access to selected routes, and facilitate capacity upgrades. It also creates obligations: supplier quality, technology evolution, maintenance, exchange‑rate risk, and recurrent payments. The Alliance may have to pay an international bill even if a national member pays late.

These contracts show that aggregation is not only a policy idea. It becomes infrastructure when an organisation can bundle needs, sign on behalf of several members, install equipment, and manage restoration. But control remains shared: fibres, landing stations, and transport networks are largely owned by third‑party operators.

The 2014 go‑live moved the Alliance from promise to production

In July 2014, UbuntuNet Alliance and DANTE announced the commissioning of regional links at 622 Mbit/s. The figure looks modest by today’s standards, but it represented a real change for networks that had previously been buying hundreds or tens of megabits at high cost.

This milestone created a production domain: real traffic, incidents, on‑call rotas, configuration, contracts, and member expectations. From that point, an outage or a policy error was no longer just a project concern; it affected institutions and researchers.

The go‑live also provided a base for later extensions. PoPs, operator relationships, NOC procedures, and billing mechanisms could be strengthened rather than invented afresh in each new phase.

AfricaConnect2 widened the programme and strengthened eastern resilience

AfricaConnect2, announced in 2015 as a €26.6 million programme covering several regions, included UbuntuNet, WACREN, ASREN, and GÉANT. This sum does not constitute the Alliance’s revenue; it funds a collection of networks, partners, capacities, and activities.

For UbuntuNet, the phase enabled wider links, added members, and finalised the SEACOM contract connecting Kampala, Dar es Salaam, and Amsterdam. This path improved diversity in the eastern part and reduced dependence on a single route or supplier.

AfricaConnect2 also reinforced the pan‑African logic: multiple regional networks could develop their own infrastructure within a common framework. Progress remained asymmetric, however, because national maturity, markets, and geographies varied widely.

AfricaConnect3 made the services portfolio as important as the backbone

AfricaConnect3, with a total programme value of €37.5 million, supported the network, pricing, training, identity, cloud, and open science until April 2025. The Alliance was no longer judged only on the number of circuits, but on its members’ ability to use common infrastructure.

The phase produced Network 3, new PoPs, faster national access, and price reductions. It also consolidated federated identity, repositories, Utafiti Africa, AfricArXiv, cybersecurity, and institutional support.

This mission extension is logical: an under‑used fibre because of missing identity, storage, or skills does not yield its full scientific value. It increases, however, the number of services that must be maintained. Each platform adds software, data, security responsibilities, and a recurrent cost.

AfricaConnect4 opens a new investment window and sets a sustainability deadline

AfricaConnect4 began in 2026 with a further €40 million of European support over four years across sub‑Saharan African regional networks. The amount is programme‑wide and cannot be presented as a grant exclusive to UbuntuNet.

The new phase envisages national and campus components, cybersecurity, climate sensors and data, EUMETCast, federated computing, GPUs, and open science services. This orientation recognises that a strong regional core is not enough if institutions can neither access it properly nor turn the capacity into research workflows.

The deadline is also financial. Equipment, repositories, identities, stations, and clusters may remain in place after 2030, while the initial funding will have ceased. Every service must therefore identify its permanent operator, its budget, its charging model, its skills, and its conditions for exit or migration.

AS36944 defines UbuntuNet’s public routing boundary

UbuntuNet uses AS36944. Routing registries and PeeringDB confirm its identity, some peers, announced prefixes, and valid RPKI state for observed origins. They give a useful view of the public border, not a complete map of every private circuit and service.

PeeringDB displays a declared traffic category of 5–10 Gbit/s, about 1,200 IPv4 prefixes, and 150 IPv6 prefixes. bgp.tools observes a smaller set of directly originated prefixes. The differences are normal: one service measures announcements and declared relationships, the other sees routes from its collectors.

None of these numbers should replace core capacity or a member’s throughput. The ASN is evidence of operational existence and an interface with the BGP system; it does not measure availability, peak traffic, or physical diversity.

PoPs and peering create options, but a map is not an asset register

The 2023 report identifies PoPs or core sites in Cape Town, Mtunzini, Maputo, Dar es Salaam, Nairobi, Amsterdam, Gaborone, and Johannesburg, with Djibouti and Mombasa described as planned or under development. Older pages mention London and other historical locations.

Discrepancies may reflect different dates, different categories between core PoP, operator handoff, and exchange presence, or the removal of some sites. A public map does not always make clear who owns the router, the fibre, the rack, the port, or the contract.

The new Johannesburg PoP and peering at NAPAfrica have improved options in the south. But resilience depends on genuine diversity of ducts, providers, power, colocation, and control‑plane. Two lines on a map can share the same cable or the same metropolitan risk.

The Lagos–Cape Town route makes the pan‑African federation more concrete

The 2025 and 2026 developments connected WACREN to the South African environment via ZAOXI, SANReN, and TENET, and then to UbuntuNet Alliance. The partners presented this architecture as the first effective direct interconnection between African regional research networks on African soil, bypassing Europe.

The claim must remain attributed and limited to RENs. It does not mean that no commercial or private link previously existed between the north and south of the continent. The innovation lies in the institutional and operational link between regional research organisations.

The mechanism matters: more African scientific traffic can stay on continental paths, reduce some detours, and facilitate inter‑regional projects. The service nevertheless depends on operators, exchange points, BGP policies, and national partners. No single body controls the entire journey.

Membership figures describe different categories and dates

The 2023 report counted 15 active members and more than 3.5 million users across more than 1,000 institutions in 13 member countries. A March 2026 audit call lists NRENs from 18 countries. The public home page shows 16 profiles. These numbers are not necessarily contradictory.

They can distinguish between legal members, active members, connected NRENs, observers, published profiles, or countries in a conflict situation. Lesotho was welcomed in November 2025, and the 2026 elections show a community still evolving.

The editorial rule is simple: date and define each total. Membership does not prove a production connection, and a national connection does not prove that all universities have equivalent access.

Campus infrastructure is the litmus test of regional investment

A fast regional link does not fix inadequate local fibre, outdated Wi‑Fi, unstable power, a saturated firewall, or a team without diagnostic tools. The researcher’s experience is limited by the weakest segment in the chain.

That is why AfricaConnect4 includes national and campus components. The programme recognises that core investment can remain under‑used if institutions cannot absorb it. Data transfers, scientific instruments, and cloud also require servers, storage, and end‑to‑end configurations.

The outcome should not be charged to a single organisation. UbuntuNet provides and coordinates the regional layer. The NREN must build the national network. The university must invest in its systems. Government, operators, and funders shape the conditions. A good measure tracks the service all the way to the researcher instead of stopping at the flag on a map.

Lilongwe governs while Kampala concentrates network operations

The Secretariat is in Lilongwe, Malawi. Organisation and recruitment documents place the NOC in Kampala, Uganda. This separation is consistent with a regional body: administration, governance, and technical operations do not need to share the same building.

The Secretariat manages projects, contracts, members, finances, and partnerships. The NOC monitors links and equipment, opens tickets, coordinates with operators, and supports NRENs. Separation can increase resilience and bring operations closer to a technical ecosystem, but it demands clear responsibilities, secure access, and continuity plans.

Such a distributed organisation must avoid concentrating knowledge in a few people. Documentation, on‑call rotas, access backups, succession, and skill transfer matter as much as monitoring tools.

Shared operations must strengthen national capacity, not replace it

Not every NREN can afford a 24/7 team. UbuntuNet can offer monitoring, diagnostics, preventive maintenance, and first‑ or second‑level support. The NREN keeps its customers, its policies, its suppliers, and responsibility for its domestic network.

Sharing is effective when it fills a gap while building local autonomy. It becomes problematic if a member ceases to understand its own configurations or to be able to leave the service. Action rights, change windows, log retention, contacts, costs, and a transition plan must therefore be defined.

The regional NOC cannot repair a national fibre or change a university firewall without authority. Its value comes from correlating evidence, escalation, and collective memory. The right goal is an NREN that is stronger because of the service, not a member dependent on an overstretched Secretariat.

Federated identity forms a second network based on institutional trust

Connectivity carries packets; identity federation allows a service to trust a person authenticated elsewhere. The home institution maintains the account and asserts attributes. The national federation coordinates metadata and rules. The regional and global layers allow a service provider to recognise that identity.

UbuntuNet does not hold all the region’s student passwords. It operates or supports proxies, metadata, federations, and the journey to eduGAIN. Security still depends on local directories, the account lifecycle, multi‑factor authentication, certificates, and attribute quality.

The benefit is enormous: libraries, clouds, repositories, compute, and research platforms can be accessed with institutional credentials. The risk is also large: a compromised institution can propagate false assertions. Federation extends trust; it must therefore extend controls, audits, and withdrawal procedures.

Eduroam turns campus identity into a mobility service

Eduroam lets a student or researcher use their credentials at a visited institution. The authentication request may traverse local, national, regional, and global proxies before returning to the home institution.

The experience seems simple, but it depends on certificates, realms, RADIUS, secure Wi‑Fi, and maintained accounts. UbuntuNet supports or runs the regional layer for several countries, without being able to correct an expired credential or a misconfigured access point.

Success should not be measured solely by the number of servers installed. One must track active institutions, successful authentications, incidents, resolution time, and configuration security.

EduID.africa and eduGAIN open the way to wider trust systems

EduID.africa is operated with WACREN and ASREN as a hosting federation or template for countries that do not yet have a mature national federation. The aim is to provide an entry path without waiting for each state to reproduce the whole structure immediately.

MAREN and TERNET joined eduGAIN in 2025, broadening the connection of Malawi’s and Tanzania’s federations to the global system. This admission establishes institutional interoperability; it does not guarantee that every university already has adequate directories, attributes, and assurance practices.

The regional layer is useful when it accelerates national building and remains exit‑ready. It would be risky if it became a permanent opaque substitute. The architecture must allow metadata migration, clear governance, and shared responsibilities.

Research4Life turns federation into concrete access to knowledge

The 2025 Research4Life pilot let institutions in Kenya, Malawi, Uganda, and Zambia use their home credentials to access scientific resources remotely. A follow‑up reported 66 institutions integrated across 15 countries.

The value goes beyond convenience. Shared institutional passwords are hard to secure, revoke, and attribute. Federation can offer individual access, linked to affiliation, without creating a separate account for each resource.

The figure remains dated and does not mean universal availability. Institutions still have to manage accounts and attributes, and Research4Life keeps its own eligibility rules. The case nevertheless shows how a regional identity infrastructure yields a benefit directly perceivable by the researcher.

Cloud services combine regional infrastructure and commercial procurement

The cloud strategy has two branches. On the one hand, community‑operated or‑hosted infrastructure: UbuntuNet Open Science Cloud, file transfer, Drive, hosting, collaboration, and experimental environments. On the other, frameworks for access to commercial services, notably the Redington/AWS agreement announced in 2024.

A regional cloud can keep more skills and control with members. A hyperscaler offers a depth of services that a small organisation cannot replicate. UbuntuNet can aggregate demand and simplify support, but it does not own AWS and does not determine its availability or pricing.

The 2023 report documents an installation in Lusaka and equipment for Kigali, without providing a consolidated capacity, utilisation, tenant count, or common SLA. Installed hardware does not by itself create a mature multi‑region cloud: orchestration, identity, backup, security, support, and funding are decisive.

Support services address ordinary frictions as much as big science

Utafiti Africa aggregates funding and training opportunities free of charge. The service exceeded 1,100 active users in 2024. It does not award grants or guarantee funding; it reduces the discovery cost for researchers.

The catalogue also includes file transfer, collaboration, hosting, and tools such as Drive, Meet, or Learn. These services can avoid duplicate purchases and provide a better option than e‑mail or consumer storage. Their maturity levels and operators differ, and some older pages still use development language.

A one‑off terabyte transfer is not the equivalent of a sustained scientific flow. Large instruments demand storage, end‑point tuning, monitoring, and sometimes dedicated circuits. The portfolio’s credibility depends on precisely bounded use cases.

Open science has become a second infrastructure layer

A network can move a paper without making it discoverable, citable, or preserved. Many institutions lack repositories, metadata, persistent identifiers, and staff. UbuntuNet entered open science because these functions also rest on hosting, storage, identity, certificates, backups, and durable organisations.

UbuntuNet and Access 2 Perspectives have been hosting the AfricArXiv infrastructure since late 2023 on DSpace. AfricArXiv remains community‑driven; the Alliance is not its sole owner. UbuntuNet was also selected as the Southern node of the Africa Open Science Platform.

The responsibility is long‑term. A repository must maintain software, content, rights, security, and preservation. Users must be able to export their data if a partnership changes. Connectivity is easier to replace than a neglected scientific heritage.

Repository and identifier management demands permanent institutional work

The Research Repository Stewardship Programme trains librarians and data managers. Its first Portuguese‑language workshop took place in Mozambique in July 2026, proof that localisation is part of operations, not just communication.

Institutions need deposit policies, metadata, rights, backups, updates, preservation, and a budget. Multi‑tenant platforms reduce duplication but create a shared failure domain. The Network Adoption Fund projects seek to turn NRENs into sustainable service providers.

DataCite, DOIs, ORCID, and the 2026 DOCiD partnership with TCC Africa add an identification layer. An identifier does not fix poor metadata. UbuntuNet can coordinate acquisition, integration, and regional practices while keeping global governance and local responsibility visible.

Cybersecurity capacity is distributed across people and institutions

Security work includes regional bootcamps, national workshops, identity and systems training, and participation in R&E communities. There is no evidence of a single SOC monitoring all members. The model instead strengthens national and institutional teams and creates trusted contacts.

This distribution fits the federation. Laws, data, skills, and incident procedures remain national. UbuntuNet can share tools and knowledge, but cannot automatically remediate a compromised device on a campus.

Financial training is part of operational security. In 2026, a programme brought together managers from 15 countries. Without budget, compliance, contribution collection, and currency management, a support or capacity contract can expire. Administration is therefore a component of resilience.

Governance ensures continuity despite incomplete public information

The documented structure places the Assembly above the Board of Trustees and the Secretariat. The Assembly provides legitimacy; the Board supervises; the Secretariat executes, manages contracts, and runs services.

Professor Madara Ogot was appointed CEO for a four‑year term from 1 February 2022. 2026 sources still present him as CEO, but no separate renewal announcement has been found. Professor Hellicy Ng’ambi became the fourth Chairperson on 1 January 2024. Miriam Chahuruva and Sabelo Dlamini were elected Trustees on 29 May 2026.

Sources do not publish a fully reconciled Board after the election, nor all corresponding departures. This gap is not proof of vacancy, but it limits journalistic precision and underscores the value of a maintained public register.

2023 accounts show a large mission and a very thin cushion

The most recent complete financial report found shows $2,404,779 in revenue, $2,393,911 in expenses, and a surplus of $10,868. Connectivity services provided $1,902,318, or 79.1% of revenue, and network operating costs reached $928,032.

The surplus is small relative to the mission and the risk. An operator invoice, a currency swing, or a contribution delay could absorb it. The drop in purchasing costs was passed through to NRENs, which serves the mission but reduces the available margin.

The CEO noted that late payment of operating and maintenance contributions was seriously compromising the Alliance’s ability to meet its commitments. No more recent audited accounts have been found; it would therefore be incorrect to estimate cash or solvency from programme announcements.

Donors, operators, currencies, and conflicts are part of the operating model

UbuntuNet’s autonomy is real but conditional. AfricaConnect funds the major generations. Operators provide a large share of the physical infrastructure. Clouds control their platforms. Contracts are often in euros or dollars, while members receive budgets in local currencies. Conflicts can reduce an NREN’s activity even if the regional core works.

The collapse of RwEdNet in 2018 and its reconstruction in 2024 show that a national institution can fail despite the existence of a regional route. Members in conflict zones can retain their status while having intermittent service.

These dependencies define the work: diversify, document, train, build reserves, and refuse to confuse sovereignty with isolation. The goal is to increase options and skills, not to pretend that global suppliers have disappeared.

Public data are strong on programmes, weaker on current operations

UbuntuNet publishes many milestones, elections, connections, trainings, and launches. These sources allow history and responsibilities to be reconstructed. They are less complete for current performance. The file contains neither a route‑by‑route register, a multi‑year availability history, a full traffic series, per‑link utilisation, a public incident database, current headcount, nor audited accounts beyond 2023.

Public databases also measure different things. The PeeringDB traffic category is neither core capacity nor a member’s throughput. bgp.tools shows observed origins and relationships, but not all private connections. A PoP can mean a router, a handoff, or a wider metropolitan environment.

An absence of published incidents does not prove an absence of outages; a lack of metrics does not prove poor quality. The defensible conclusion is that programmes, contracts, and upgrades are well documented, while an independent SLA and a full dependency map cannot be calculated. Publishing more availability, restoration, utilisation, and adoption data would strengthen member governance.

Regional autonomy is built through options, not the elimination of all dependence

The most useful digital sovereignty is practical. UbuntuNet increases regional control by pooling procurement, keeping more traffic local, developing engineers, and operating identity and repositories. It cannot make the region independent of global standards, commercial cables, equipment supply chains, hyperscalers, or scientific partners.

Links with WACREN, ASREN, SANReN, TENET, and GÉANT illustrate a pan‑African federation with no single owner. Lagos‑Cape Town reduces some European detours, but still depends on operators, exchanges, and routing policies. It is an improvement in choice and locality, not the creation of a closed network.

Every measure strengthens one dimension while retaining a dependency: the AWS agreement eases access but increases exposure to a hyperscaler; a regional repository improves local governance but requires software, identifiers, and labour; a long‑term contract lowers cost but raises supplier and currency risk.

The strategic test is the ability to change: can a member switch providers without losing the federation? export its data? move a workload? keep its identity? Autonomy advances when these answers become positive and recedes when a central dependency becomes opaque and un‑exit‑able.

UbuntuNet matters because it turns separate networks into a usable research environment

The central achievement is as much institutional as technical. Independent national networks can pool demand, share operations, connect to the world, and build services that would otherwise be absent or duplicated. The backbone, identity, repositories, cloud, and training are distinct, but answer the same problem of fragmentation.

The organisation remains unfinished. Capacity is still modest relative to climate, astronomy, AI, and other data‑intensive sciences. Public evidence is stronger on milestones than on utilisation and availability. Member finances vary, suppliers and funders retain an important place, and service maturity is uneven.

These limits do not lessen the subject’s interest. UbuntuNet is neither a separate African internet nor a finished continental platform. It is the common layer that makes a larger system possible. Its success will be measured by its members’ ability to use, fund, monitor, and leave the services without losing their autonomy.