Summary
- The exact label "IX TELECOM S.A." is incomplete. Spanish commercial-register notices show that the company now called AFR-IX Telecom S.A. began as WAPOP Africa Interconnect Exchange S.L., became AFR-IX Telecom S.L. in 2015 and converted to a public limited company in 2022. Its current legal notice links that same Barcelona registration to the AFR-IX business.
- South African operating authority sits with Computer and Satellite Electronics (Pty) Ltd, or CSE. ISPA identifies CSE as the locally registered, licensed company, says it is wholly owned by AFR-IX, and lists both network-service and communications-service licences. That distinction matters for contracts, access plant, repair obligations and regulatory accountability.
- AS60171 is visibly operating, not merely reserved. RIPE routing observations show current IPv4 and IPv6 announcements, while NAPAfrica and INX-ZA list AFR-IX in Johannesburg, Cape Town and Durban. These records support real peering and transit operations, but port listings do not disclose utilisation, leased route ownership, spare capacity or shared failure domains.
- AFR-IX says it supplies international capacity to fibre and fixed-wireless partners in South Africa. Those announcements fit an aggregation business in which CSE/AFR-IX controls routing and service management while local partners control towers, fibre drops and customer installations. Neither announcement provides circuit identifiers, delivery tests, route diagrams or restoration results.
- The appropriate evidence grade is Medium. The legal chain, local licence holder, live routing, multi-city exchange presence and local staff are observable. Physical diversity between cities, independent subsea exits, backup-power duration, contracted headroom, access ownership and field-repair performance remain undisclosed.
The missing letters decide which company is being tested
An infrastructure assessment can go wrong before it reaches the first router. "IX Telecom" is not a unique name, and the suffix "S.A." does not identify a South African company. A separate Malaysia-based business trades as IX Telecom Sdn. Bhd. and operates AS56286 in Asia. Its routes, facilities and customers do not support any claim about the entity examined here. The South African clues point elsewhere.
The decisive link is the Spanish commercial record. A 2013 notice in the Boletin Oficial del Registro Mercantil records the incorporation of WAPOP Africa Interconnect Exchange S.L. in Barcelona under registry sheet B 440369. A 2015 notice records the change of that same company name to AFR-IX Telecom S.L. A 2022 notice records its conversion into AFR-IX Telecom S.A., preserves sheet B 440369 and expands the stated corporate purpose to construction, operation, installation, management and maintenance of telecommunications infrastructure.
AFR-IX's current legal notice completes the chain. It names AFR-IX Telecom S.A. as the website owner, gives Spanish tax number A66087420 and cites the same Barcelona register book and page. The exact public evidence therefore supports AFR-IX Telecom S.A., not an independently verified company whose full legal name is simply IX TELECOM S.A.
This is more than a typographical correction. If an assessor accidentally attaches the Malaysian AS56286 to the Spanish parent, the apparent network jumps from Africa to Southeast Asia. If the assessor treats "S.A." as "South Africa", a Spanish legal form becomes a geographic claim. If the assessor assumes the initials mean an independent internet exchange, a commercial transit network begins to sound neutral before its role has been established.
The safest reading is narrow. The entity label in the Overview identifies an existing company entry. The operating evidence analysed here belongs to the AFR-IX group and its South African business because the underlying description, geography and legal history converge there. Where a record names AFR-IX or CSE rather than the shortened label, this article says so. It does not transfer evidence from the unrelated Malaysian company, and it does not assume that every group claim is performed by the Spanish parent itself.
That identity discipline is the first infrastructure control. A customer buying a circuit needs to know which company signs the order, which company holds the licence, which autonomous system carries the route, which operator owns the access tail and which organisation answers when the line fails. The missing "AFR-" is small on a page and large in an outage.
IX is a history, not a certificate of neutrality
The letters IX normally carry a precise association in network engineering. An internet exchange point supplies a shared switching environment in which autonomous networks exchange traffic. Neutrality concerns the rules of access, governance and commercial treatment of participants; it is not conferred by two letters in a company name.
AFR-IX's original corporate name explains why the association arises. WAPOP Africa Interconnect Exchange was incorporated in 2013. Packet Clearing House still carries a historical directory row for a Barcelona exchange of that name with zero listed participants. That row is useful as history, not as proof that a current exchange exists. The 2015 legal renaming and the present service catalogue point to a different identity: AFR-IX sells IP transit, dedicated internet access, private circuits, managed networks and colocation. Its PeeringDB entry classifies AS60171 as a network service provider, not an exchange.
The distinction can be seen in current interconnection records. PeeringDB describes AFR-IX's peering policy as selective and says a contract is required. A selective commercial network can be a valuable participant at neutral exchanges. It does not become neutral merely by connecting to them. The neutral venue and the carrier using that venue are separate operating layers.
In South Africa, those venues include NAPAfrica and the community-run exchanges operated by INX-ZA. NAPAfrica's participant list shows AS60171 at its Johannesburg, Cape Town and Durban fabrics. INX-ZA's member record identifies AFR-IX as a full member and gives ports at JINX, CINX and DINX. These are strong signals because the exchange operators, not only the carrier, publish the participation.
They also show what neutrality actually buys. At a shared exchange, AFR-IX can reach local access providers, content networks and other carriers without sending every packet through a paid international transit path. The result can be lower latency, lower unit cost and less dependence on one upstream. Yet a port at an exchange does not oblige every other participant to peer, does not reveal private interconnection terms and does not guarantee that the fibre from AFR-IX's router to the exchange survives a building, duct or metro outage.
The neutral claim should therefore be divided into three questions. Is the data-centre or exchange venue open to competing networks? The venue operators publish evidence that it is. Does AS60171 participate there? Multiple independent lists say yes. Is AFR-IX itself a neutral exchange, or are its products delivered without preference, lock-in or dependence on specific suppliers? The reviewed evidence does not establish that broader proposition.
This distinction matters commercially. A regional ISP buying transit from AFR-IX needs the carrier to make routing choices, prioritise incidents and enforce contracts. Those are the actions of an accountable operator, not a neutral switch. The useful promise is not neutrality. It is controlled connectivity with enough alternative paths that one failure does not erase the service.
South African authority sits with CSE
The Spanish parent is only one part of the operating structure. The clearest South African record is maintained by the Internet Service Providers' Association. Its membership page for Computer and Satellite Electronics (Pty) Ltd says CSE provides internet and data services under the AFR-IX Telecom South Africa name, is registered locally and is wholly owned by AFR-IX. It lists individual electronic communications network service licence 0117/IECNS/JAN/09 and individual electronic communications service licence 0117/IECS/JAN/09.
The licence types separate two powers that marketing language often blends. South Africa's Electronic Communications Act distinguishes the provision of communications networks from the provision of communications services. An ECNS holder can establish and operate network facilities, while an ECS holder can provide communications services to customers. The licence does not prove that CSE owns every fibre strand, tower, router or international wavelength it uses. It establishes a local legal basis for network and service activity.
AFR-IX's own 2021 ISPA announcement says it acquired CSE in 2020. ISPA's independent member entry is more useful than the announcement because it names the registered company, licence numbers, address, contact channels and offered services. The Wireless Access Providers Association also lists AFR-IX as a Gauteng vendor member, another sign that the business sits in the wholesale and access-provider ecosystem rather than presenting itself only from Barcelona.
The ownership boundary should be drawn explicitly. AFR-IX Telecom S.A. is the Spanish group parent and owner of the brand and website. CSE is the named South African member and licence holder. AS60171 is registered through the RIPE region to AFR-IX Telecom S.A. Local fibre network operators, tower owners, data-centre companies, exchange operators and cable-capacity suppliers control other assets needed to deliver a South African circuit. A single customer service can cross all of those boundaries.
That division affects recovery. A BGP routing error may be handled by AFR-IX's network engineers. A failed router at a colocated site may require local remote hands or a CSE engineer. A cut metropolitan tail may depend on the contracted fibre owner. A failed fixed-wireless sector belongs to the wireless partner. A submarine fault is repaired by a cable consortium and specialist vessel. The customer sees one outage, but the restoration clocks belong to several organisations.
It also affects what should appear in a service order. The order should name the contracting company, service demarcation point, access supplier, autonomous system, protected or unprotected status, committed information rate, handoff interface, maintenance window, mean time to repair and service-credit regime. If the service is sold as diverse, the order should state what is diverse: carrier, duct, building entry, metro route, long-haul route, cable system, landing station, router, power feed or all of them.
The evidence supports a real South African operating company inside the AFR-IX group. It does not support treating the Spanish parent, CSE and every contracted access network as one physically unified asset base.
AS60171 is visible in the places that matter
The routing evidence is substantially stronger than the initial name evidence. AFR-IX operates AS60171, and that autonomous system is observable from public route collectors. A RIPEstat routing-status query on 10 July 2026 reported 31 IPv4 prefixes covering 20,224 addresses and ten IPv6 prefixes, with its announcements visible to all reporting RIPE RIS peers in both protocol families at the query time. Counts change as customer routes move, so they should be read as a dated observation rather than permanent capacity.
The routing record proves several limited but important things. AS60171 originates or carries publicly visible address space. Its routes propagate internationally. IPv6 is not merely listed as a future feature. The network is far beyond the stage of holding an unused ASN or presenting a map without a route.
Cloudflare's routing view for AS60171 independently identifies AFRIX-AS and displays announced space, connectivity and BGP activity. CAIDA's AS Rank entry places AFR-IX in a large observed customer and neighbour graph. Neither service can determine where every router sits or which physical cable carries a given path. Together with RIPE observations, they support the conclusion that AFR-IX is acting as a transit network for more than its own small pool of addresses.
The South African exchange evidence adds physical anchors. PeeringDB lists AS60171 at Teraco facilities in Johannesburg, Cape Town and Durban and at Africa Data Centres JHB1 in Midrand. It lists NAPAfrica ports in all three cities and a JINX port in Johannesburg. NAPAfrica corroborates the three-city presence. INX-ZA corroborates ports at its Johannesburg, Cape Town and Durban exchanges. Its JINX looking glass has shown the AFR-IX IPv4 session up and receiving more than a thousand routes, while its DINX looking glass has shown the Durban session up with a similar order of routes.
This is credible operating evidence because it joins control-plane visibility to named interconnection sites. A route collector alone could see an announcement delivered from Europe. A facility list alone could be stale. Current exchange sessions in South Africa make it much more likely that AS60171 has active routers and cross-connects in the country.
There is still room for caution. PeeringDB is maintained by participants and facility operators; it is not an audited asset register. A session can be delivered over a remote peering service rather than a dedicated router in every listed building. Exchange route-server status confirms a session to that fabric, not the ownership or physical path of the circuit feeding it. Public BGP also exposes logical adjacency, not spare optics, line-card redundancy or restoration inventory.
The appropriate conclusion is stronger than "thin footprint" but narrower than "fully mapped network". AFR-IX has a visible, multi-city South African routing presence. The public cannot reconstruct the physical network underneath it.
A port speed is not usable capacity
Capacity claims become slippery when three different quantities are presented as one. Design capacity is what an interface, optical system or cable could carry under defined conditions. Installed capacity is what has actually been equipped and commissioned. Usable capacity is what remains available after current traffic, protection reservations, oversubscription, failures and commercial limits are accounted for.
PeeringDB lists 100Gbps NAPAfrica ports for AFR-IX in Johannesburg and Cape Town and 10Gbps in Durban. It also lists a 10Gbps JINX connection. INX-ZA's own member page, however, lists 1Gbps at JINX, 1Gbps at CINX and 10Gbps at DINX. The difference may reflect a later upgrade, a lag in one directory, separate ports or a simple entry error. It is exactly why a public port field should not be turned into a claim about delivered customer capacity.
Even a confirmed 100Gbps physical port would not say how much traffic passes through it. It could be lightly used and provide generous headroom. It could carry many downstream customers and become busy at peak time. It could be one member of a protected pair, or the only path into a city. It could reach two exchange fabrics through one metro duct. A line rate measures an interface, not a service's failure tolerance.
AFR-IX's IP transit page markets scalable bandwidth from 100Mbps to multiple 100Gbps, direct peering, redundant paths and service-level guarantees. Its dedicated internet page offers interfaces up to 10GbE, non-oversubscribed bandwidth, managed customer equipment and monitoring. These are clear product specifications. They do not disclose total installed line capacity, the ratio of sold commitments to protected capacity, average utilisation by city or the conditions under which backup paths become congested.
This gap matters most during a failure. A network may have enough capacity in normal operation because traffic is divided across west-coast subsea cables, east-coast cables, local peers and content caches. When a cable group fails, the remaining international paths inherit the load. When a metro ring is cut, traffic may converge on one intercity link. Capacity that looked spare can disappear before routing has settled.
Buyers should ask for evidence matched to their service rather than a global headline. Useful evidence includes 95th-percentile utilisation on the customer-facing port, peak utilisation on the normal and protected backbone paths, the capacity reserved for failover, the number of services sharing the access tail, and a test showing that the backup route carries the committed rate. A simple before-and-after traceroute is not enough because traffic can move logically while remaining inside the same cable corridor.
The same caution applies to point-of-presence counts. AFR-IX's South Africa page says it has more than 20 PoPs in major SADC countries; its broader corporate pages use larger worldwide totals. No reviewed public page supplies a dated PoP inventory with site, operator, active equipment, capacity and readiness. A sales node, remote port, router, rack and staffed facility can all be called a PoP. The count becomes operationally useful only when the definition is consistent.
AS60171's public routes establish usable connectivity at the moment of observation. They do not reveal how much usable capacity would remain after the failure that matters to a specific customer.
International means surviving failures on both coasts
South Africa has a comparatively rich set of international cable landings. That national diversity is valuable, but an individual carrier benefits only from the systems on which it has contracted capacity and the terrestrial routes that can reach them. A map of cables near the country is not a map of AFR-IX capacity.
AFR-IX said in 2021 that the arrival of the Africa Coast to Europe cable in South Africa strengthened its IP transit and Ethernet services by connecting Cape Town and Johannesburg nodes. Its current South African page refers more generally to multiple submarine cable systems. Those statements support access to international capacity, but they do not name current protected quantities, landing-station contracts or restoration priorities.
The Internet Society's report on the March 2024 West Africa outage shows why names and routes matter. ACE, SAT-3, WACS and MainOne were simultaneously offline after damage off Cote d'Ivoire, affecting connectivity in 13 countries including South Africa. Networks with east-coast alternatives could reroute, but replacement paths absorbed additional load. In May 2024, a separate east-coast outage followed damage to SEACOM and EASSy off KwaZulu-Natal.
Those events define a credible stress test for AFR-IX. A resilient South African international service should not merely buy two logical circuits that enter the same Cape Town landing system or share the same west-coast fault domain. It should combine independently routed terrestrial backhaul with capacity on cable systems whose marine paths, landing stations and onward European routes do not fail together. For some customers, an east-coast route to Durban or Mtunzini and a west-coast route to the Cape would be more meaningful than two suppliers on one coast.
Teraco's infrastructure demonstrates how the choice can be made. Its Durban DB1 description gives access to SEACOM, EASSy, METISS and 2Africa and says DB1 connects to the Johannesburg campus through a choice of carriers over diverse regional fibre routes. Its Cape Town diversity note says CT1 and CT2 are more than 30 kilometres apart and on different power grids. Those are facility capabilities, not evidence that AFR-IX has bought and tested every available option.
The South African AFR-IX footprint at Johannesburg, Cape Town and Durban is a good starting geometry. Three cities can support local peering, coastal diversity and an inland aggregation point. The missing layer is the set of actual paths between them. Are Johannesburg-Cape Town and Johannesburg-Durban delivered by different long-haul operators? Do they share a duct on the first kilometres out of the data centre? Does traffic automatically prefer a local South African exit, or does policy send it through Europe? How much capacity remains if a coastal route is lost?
Public BGP cannot answer those questions. Different AS paths may still ride the same fibre, and one AS path can move across multiple physical systems. A meaningful proof would combine contract records, route drawings, carrier letters of authority, optical demarcation details and a controlled failover test. Customers do not need every commercially sensitive coordinate. They do need enough evidence to know that "multiple" describes independent failure domains rather than multiple product names.
The local network is partly somebody else's network
AFR-IX's South African strategy becomes clearer in two recent announcements. In March 2026, the company said it would provide international connectivity to Konekt SP, a fixed-wireless wholesaler. In June 2026, it said it would provide high-capacity IP transit to fibertime, a pay-as-you-go fibre provider serving township communities.
The announcements are credible descriptions of commercial intent, and the counterparties have independent public footprints. Konekt describes itself as a South African connectivity wholesaler. Fibertime publicly offers uncapped service at up to 100Mbps with daily payment options. Neither counterparty page reviewed for this article independently sets out the AFR-IX contract, circuit size, start date or service results. The relationships should therefore be treated as first-party announcements until operating measurements or counterpart confirmation add detail.
The architecture they imply is still important. AFR-IX is not claiming to dig every fibertime trench or maintain every Konekt radio. It is positioning itself behind the access networks, supplying international reach, routing, interconnection and managed service. That is a normal wholesale structure. It lets a local builder concentrate capital and labour on homes, towers and customer equipment while a carrier aggregates traffic and buys international scale.
It also creates a chain of dependencies. For a fibertime household, the service may pass through an optical terminal, neighbourhood fibre, aggregation equipment, a fibertime handoff, an AFR-IX edge, a South African exchange or upstream, a long-haul route and an international cable. For a Konekt-served location, the first segment may include customer radio, tower sector, tower power, microwave or fibre backhaul and a Konekt core before AFR-IX sees the traffic. A fault at any segment can look like "the internet is down".
Responsibility needs a demarcation. AFR-IX's dedicated internet description says it coordinates local operators and supplies managed customer-premises equipment. That can simplify procurement, but it makes incident ownership more important. If AFR-IX is the single commercial contact, its operations team must be able to identify whether a fault belongs to the access partner, local power, its own router, an exchange, a transit provider or a cable system, then keep the customer informed while another company repairs it.
The partner structure also changes the meaning of coverage. Konekt's claimed high-site footprint is not an AFR-IX-owned tower estate. Fibertime's drops are not AFR-IX-owned access plant merely because AFR-IX supplies transit. The group may be able to quote service across a wide area by combining licensed authority, partner networks and managed equipment. That is commercial reach, not a single owned physical network.
The right test is not whether AFR-IX owns everything. Asset-light aggregation can be efficient and resilient when suppliers are diverse and contracts are enforceable. The test is whether the operator knows each dependency, has alternate suppliers where failure would be severe, and can restore service across organisational boundaries.
The regional economics reward aggregation and punish idle redundancy
The wholesale carrier earns revenue by combining demand that would be uneconomic to serve one circuit at a time. A local ISP may need a few gigabits today and more next year. AFR-IX can aggregate that demand with other customers, maintain larger interconnection ports, buy transit and cable capacity in bulk, and operate routing expertise that each downstream provider would struggle to staff alone.
Peering improves the arithmetic. Traffic exchanged locally at JINX, CINX, DINX or NAPAfrica can avoid a paid international path and reduce latency. Popular content may remain inside South Africa. International transit remains necessary for destinations that are not reached through local peering. The carrier's margin depends on the blended cost of local cross-connects, exchange ports, metro and intercity transport, upstream commitments, support labour, equipment, facilities and cable capacity.
This creates a tension around resilience. An unused backup circuit looks expensive in a monthly account, especially when customers compare prices. Yet a backup that becomes saturated during the first major cut is not a backup for committed service. The operator must buy enough idle or lightly used headroom to carry important traffic during failure, then recover that cost through pricing, shared protection or differentiated service levels.
The downstream economics are equally demanding. Fibertime advertises a R5 daily single-device option and other low-cost plans. AFR-IX does not receive that full retail amount. Access construction, home equipment, payment collection, customer support and local operations sit ahead of the transit bill. Low retail revenue makes efficient aggregation essential and leaves little room for waste. Konekt's wireless access can avoid some trench cost, but tower leases, sector capacity, radio replacement, backhaul and backup power replace it.
These are reasons to value AFR-IX's multi-city peering footprint. They are not reasons to assume unlimited capacity. A wholesale operator serving many small networks can experience correlated growth when streaming, software downloads or a major event lifts demand across customers at once. If most downstream traffic converges on the same Johannesburg router or international path, geographic coverage can conceal a central choke point.
The company publishes a selective peering policy and a large set of interconnection locations. It does not publish South African traffic by city, peak-to-average ratios, oversubscription policy, committed downstream load or failover headroom. Such details are commonly confidential, but customers buying business-critical service can request them under nondisclosure. A carrier can protect commercial terms while still proving that its design meets the contracted service level.
There is a second concentration risk in local support. A small South African team can be fast because it knows the sites and suppliers. The same team may be stretched by simultaneous fibre cuts, power failures and customer equipment incidents. The economics favour central monitoring from Barcelona and partner field work in each market. Recovery depends on whether escalation contracts and spare stocks turn that distributed arrangement into real response capacity.
The regional carrier proposition is plausible because AS60171 already aggregates many routes and appears at the right exchange points. Its economics remain resilient only if the cost saved through shared infrastructure is not achieved by sharing every failure domain too.
The power boundary begins inside the carrier hotel
AFR-IX's public facility list includes Teraco campuses and Africa Data Centres JHB1. Those operators supply hardened colocation environments, but the facility and the network equipment inside it have separate power designs.
Africa Data Centres' JHB1 specification describes multiple incoming feeders, 2N uninterruptible power at rack level, battery autonomy and generator fuel for a minimum of 48 hours. Teraco markets resilient power and multiple building entries across its South African sites. These are positive facility attributes. They do not disclose the feed selected for a particular AFR-IX rack, whether dual power supplies connect to separate distribution paths, or how the carrier powers equipment outside those buildings.
The equipment chain extends beyond the router. An exchange port needs a switch fabric and cross-connect. An intercity service needs optical amplifiers and regeneration sites. A metropolitan access tail may traverse street cabinets. Fixed-wireless service needs tower power and customer equipment. Fibre at a home is passive along the route but still depends on powered equipment at each end. A generator at one carrier hotel does not keep the entire chain alive.
Power diversity must therefore be specified at each failure domain. At a core site, useful questions include whether routers have dual power supplies, whether feeds come from separate UPS systems, how generator transfer is tested and how fuel is replenished during a prolonged grid event. On a long-haul route, the issue is which huts or repeater sites need power and whether the supplier's service level covers them. At a tower, battery runtime, generator access, theft protection and refuelling matter. At the customer premise, the network can remain healthy while the local router and optical terminal go dark.
The AFR-IX support page says its Barcelona operations centre monitors terrestrial and submarine infrastructure continuously and provides a South African NOC telephone number. Monitoring reduces detection time. It does not add battery minutes. An alarm can identify the dark site, but restoration still depends on remote control, automatic transfer, local hands, access permission, spares and fuel.
Facility diversity also requires route diversity. Teraco says its buildings have multiple entry points, allowing carriers to deploy separate approaches. A carrier using two cross-connects that leave through one building entry has logical duplication but physical concentration. A carrier present at both CT1 and CT2 can gain greater separation, but only if the equipment, metro routes and power grids are actually used independently. PeeringDB lists AFR-IX at CT1, not CT2, so the broader campus capability should not be credited to AFR-IX without additional evidence.
The public evidence supports well-equipped host facilities. It does not disclose AFR-IX's rack-level power configuration, intercity optical power design, tower dependencies or customer-premises backup policy. Power resilience remains a contractual and engineering question, not a consequence of a data-centre logo.
Local support is a chain of people, permissions and spares
AFR-IX's South African page names a country manager, two field engineers and a service-delivery engineer. Its managed-services page says field engineers work locally while a 24-hour operations centre monitors the network. Those details are more informative than a generic claim of global support because they identify a local technical function.
They still do not establish field capacity. Four named roles may be part of a larger team, and a web page may not list everyone. Staff count alone would not answer whether engineers are based near Johannesburg, Cape Town and Durban, carry the required site credentials, or can reach a remote Konekt high site after severe weather. A regional service also depends on the access suppliers' technicians.
Repair time is built from several intervals. Monitoring must detect the symptom. An engineer must isolate the likely domain. The correct supplier must accept the ticket. A technician needs permission to enter the site or road reserve. The repair crew needs transport, safety equipment, test gear and the right spare. A fibre cut may require locating the damage, exposing the cable, splicing it and testing every affected strand. A failed router module may be restored quickly if a compatible spare is in the building and slowly if it must clear customs.
This is why a 24-hour phone number and a restoration commitment are different products. The phone proves there is a channel. A meaningful support commitment sets acknowledgement, diagnosis, dispatch, update and restoration targets. It defines severity, excludes or includes access-provider delays, and assigns service credits. For a wholesale customer, it should also specify whether AFR-IX can open tickets directly with every underlying carrier or must wait for an intermediary.
The local CSE licence and ISPA membership strengthen accountability. A South African customer has a named local company, local contact details and an industry code of conduct. That is valuable when a dispute concerns service or communication. It does not make a third-party fibre crew arrive faster. Operations quality has to be demonstrated through incident records, not inferred from membership.
Spares are the quiet part of labour. AFR-IX has not publicly identified its South African stock of routers, line cards, optics, power supplies, customer devices, radios or fibre repair materials. A carrier can reasonably keep that inventory confidential. A customer can still ask whether critical modules are held in each city, whether configurations are backed up, whether replacement equipment is pre-authorised for site entry and whether access partners have reciprocal spares obligations.
The visible team and support channels move the assessment above a purely remote reseller. The unknown distribution of people and spares prevents a stronger judgement about recovery performance.
Six failure drills would reveal the real network
The first drill is loss of a customer access tail. Disconnect a protected test circuit's primary fibre or radio path at the agreed demarcation. Measure detection, failover, packet loss, available throughput and customer notification. Confirm that the alternate route does not share the same pole line, tower backhaul, duct or building entrance. If the service is unprotected, measure diagnosis and dispatch instead of pretending automatic recovery exists.
The second drill is a Johannesburg facility failure. Remove the logical availability of the primary router or site and observe whether traffic enters the network in Cape Town or Durban without manual customer reconfiguration. The test should show route convergence, surviving capacity and session stability. A second router in the same rack is useful for hardware failure but does not answer a building or metro outage.
The third drill is a west-coast international failure similar to March 2024. Withdraw the preferred west-coast path in a controlled window and show where traffic exits South Africa, how latency changes and how much headroom remains. Then repeat for the east coast. The result should identify cable systems or at least independent coast and landing domains without revealing sensitive commercial pricing.
The fourth drill is loss of commercial power. Verify dual-feed operation at the core sites, generator transfer and alarm escalation. For access partners, sample a tower, street node or customer device rather than testing only a data-centre rack. Record the duration for which each part of the service remains available and the conditions for refuelling or battery replacement.
The fifth drill is a spare-and-crew exercise. Declare a failed optical module in Cape Town or Durban, locate the replacement, obtain site access and complete a simulated change. This exposes expired credentials, missing adapters, configuration gaps and unrealistic travel assumptions before a customer outage does.
The sixth drill is congestion during failover. Drive a representative protected load, remove one backbone or upstream path, and confirm that the committed rate survives without excessive loss or latency. Routing diversity without capacity diversity can make a service reachable but unusable. This is the distinction between a route collector seeing a prefix and a customer completing work.
None of these tests requires publication of a full confidential topology. AFR-IX could publish a short assurance statement describing the domains tested, date, broad result and corrective actions. Large customers could receive detailed evidence under contract. Exchange operators and facility providers already publish enough location information to make broad claims testable.
The same drills would clarify the boundary with Konekt and fibertime. AFR-IX should be able to show where its responsibility begins, how quickly it identifies a partner-side fault and whether the partner's backup route has sufficient capacity. The access partner should be able to test its own segment without assuming the international carrier will mask every local failure.
Repeated tests are more valuable than a one-time certificate. Networks change as ports are upgraded, suppliers are replaced and customer traffic grows. A path that was diverse last year can become concentrated after a migration. Recovery evidence should follow the network as closely as its sales map does.
What the evidence settles, and what it leaves open
The evidence settles the broad identity. The operating group is AFR-IX, descended from the Barcelona company formerly called WAPOP Africa Interconnect Exchange. AFR-IX Telecom S.A. is the current Spanish parent named in its legal notice. CSE is the licensed South African company identified by ISPA as wholly owned by AFR-IX.
It settles that a network operates. AS60171 has current, widely visible IPv4 and IPv6 routes. Exchange operators list it in Johannesburg, Cape Town and Durban. Current looking-glass observations have shown active route-server sessions. The network is not supported only by announcements or a coverage graphic.
It settles that the South African business has people and channels. The company names local engineering roles, publishes a local office and support number, and appears in local industry associations. It also markets transit, dedicated access and managed equipment in terms that correspond to observable routing activity.
It does not settle the exact asset register. Public material does not say which South African metro or intercity fibres AFR-IX or CSE owns, which it leases, where every PoP is located, or which partner owns each access segment. It does not give route kilometres, tower counts attributable to AFR-IX, customer circuit counts or installed backbone equipment by city.
It does not settle capacity. Exchange-port fields and product ceilings are not utilisation data. There is no public South African traffic series, protected headroom figure, city-by-city commitment total or failover throughput result. Claims of non-oversubscription apply to products as marketed and would need a service-specific measurement to verify.
It does not settle physical redundancy. Multiple cities, exchanges, upstream names and cable references are encouraging. None proves that a particular customer receives separate ducts, carriers, coastlines, landing stations and power systems. Public routing views cannot see shared civil infrastructure.
It does not settle recovery. The NOC and local team can detect and coordinate incidents, but no public mean time to repair, backup-power duration for AFR-IX equipment, spare inventory, field-crew coverage map or incident postmortem was found. The difference between the Barcelona monitoring centre and South African field responsibility remains material.
Those limits do not make the network fictitious. They define the difference between an observable carrier and an independently verified resilient service. AFR-IX has crossed the first threshold. Customers still have to test the second at the circuit level.
The acronym earns trust only after the route does
AFR-IX has more operating substance than the shortened label suggests. The group has a traceable legal history, a local licensed subsidiary, a live autonomous system, multi-city South African exchange participation, named local staff and recent wholesale customer announcements. That combination supports a Medium network evidence grade.
The downgrade from Strong is not about whether packets move. Public observations show that they do. It is about what happens when the preferred path, powered site, access supplier or repair resource disappears. The company publishes the ingredients of resilience but not the physical composition or test results for South African services.
The most useful disclosure would be a dated South African network assurance note. It should list active cities and exchange fabrics, distinguish owned from leased and partner access, describe independent coast and landing domains, state failover headroom in broad bands, identify local support coverage and summarise recent route, power and congestion tests. It should reconcile differing public port speeds and define what counts as a PoP.
For customers, the practical response is equally concrete. Contract with the correct local entity. Record the access owner and demarcation. Require physical rather than merely logical diversity. Ask how many failures the committed rate can survive. Test the backup path under load. Keep customer-premises power in scope. Escalate through one accountable service owner while preserving visibility into the supplier chain.
"IX" can evoke a neutral meeting place where networks interconnect. AFR-IX's present value is different: it is a commercial carrier that can join those meeting places and connect smaller networks to them. That role can be important to South African fibre and fixed-wireless operators, especially when low retail prices make independent international engineering uneconomic.
But neutrality is a property of governance, and resilience is a property of paths. Neither comes from an acronym. AFR-IX has proved that AS60171 is present. It still has to prove, service by service, that international connectivity remains usable when the ordinary route is no longer there.