Summary

  • AFRINIC records tie the exact name RAXIO DATA CENTRE SMC LIMITED and the Namanve address to active AS328821 and the allocated IPv4 block 102.220.156.0/23. At the July 15, 2026 observation, RIPEstat showed the aggregate and two more-specific /24 routes visible globally from that autonomous system.
  • UG1 has stronger public facility evidence than a brand-only hosting offer. Raxio describes a purpose-built, 1,000-square-metre colocation site with 1.5 MW of IT power, rack-level 2N power, N+1 cooling, two fibre intake points, two meet-me rooms, on-site fuel and round-the-clock operations. UIXP and PeeringDB independently show interconnection activity at the site.
  • Important claims need narrower interpretation. Raxio's current web page, downloadable technical sheet and PeeringDB entry do not use the same rack, cooling, carrier or uptime figures, while the technical sheet specifically describes a Tier III certified design. A customer should attach the current certified scope, built capacity and service-level schedule to the contract rather than rely on the broadest web wording.
  • The public record shows a 24-hour remote-hands and support proposition, named local management and published technical contact fields, but not achieved response times, restoration performance, staffing depth, spare-parts coverage or customer-specific incident results. Local support becomes assurance only when responsibility, authority and escalation are tested against the purchased service.

The legal name reaches a real facility and network

Data-centre procurement often starts with a brand and ends with several different parties. A group company may market a regional portfolio. A local company may sign the contract. A landlord may control the building, utilities may supply power, carriers may carry traffic, and a separate operator may provide remote hands. The first discipline is therefore attribution: identify which name owns each obligation before asking whether the obligation is strong.

The AFRINIC registration for AS328821 provides the cleanest public join for Raxio Uganda. It names RAXIO DATA CENTRE SMC LIMITED as the registrant organisation, gives Plot 781, Block 113, Namanve Industrial Park as the address and records the autonomous system as active. The same organisation appears in AFRINIC's record for 102.220.156.0/23, a provider-aggregatable block covering 512 IPv4 addresses and registered in March 2021.

That exact-name evidence matters. Raxio Group's website markets a multi-country portfolio, and PeeringDB's organisation entry gives the group an Amsterdam address. Those records are useful for understanding the wider brand, but they are not substitutes for the Ugandan counterparty. AFRINIC instead links the local legal name, local address, network resource and named administrative and technical contacts. It does not establish current ownership, directors, financial strength or signing authority. Those still belong in a current Ugandan company extract and executed agreement.

The location also lines up across independent records. Raxio's Uganda page places UG1 in Namanve Industrial Park, about 15 kilometres from Kampala's central business district. PeeringDB's UG1 facility record gives Plot 781, Block 113, Namanve Industrial Park, the same plot used by AFRINIC. PeeringDB labels the property as owner-operated, lists Ugandan technical and sales contacts and reported that its facility record was updated in April 2026.

The alignment supports a restrained but useful conclusion. RAXIO DATA CENTRE SMC LIMITED is not merely a similar name attached to a directory row. It is the named holder of a Ugandan network identity at the address used for Raxio UG1. That does not mean every service promoted under the Raxio brand is delivered by this company, or that every group promise appears in the Ugandan contract. Procurement should state the contracting entity, facility, service boundary, governing terms, invoice party and escalation contacts explicitly.

UG1 sells controlled space, power and access before it sells cloud

The term cloud can blur the product boundary. Raxio's own material is more specific. The UG1 technical sheet says the group's facilities are dedicated to colocation and lists colocation, cross-connections, cages, remote hands, active and dark fibre, customer spaces and relocation assistance among UG1 services. These are infrastructure inputs. A customer or another service provider normally supplies the servers, operating systems, applications, data architecture and much of the recovery design.

That distinction determines who acts during failure. If a customer-owned server stops responding, Raxio may be responsible for rack power, environmental conditions, physical access and a requested remote-hands task. The customer may remain responsible for the server warranty, spare parts, firmware, hypervisor, application, backup and diagnosis. A carrier may own the circuit. A cloud provider reached through the building may own the virtual service. Calling the whole arrangement cloud infrastructure does not merge those obligations.

UG1's public specification is nevertheless substantial. Raxio describes 1,000 square metres of white space, 1.5 MW of IT power and rack densities from 4 kW to 21 kW. The current web page says power is distributed with 2N redundancy from transformer, generator and uninterruptible power supply through to the rack power-distribution unit. It also describes at least N+1 cooling resilience, energy-efficient indirect adiabatic cooling, seven minutes of UPS autonomy and fuel stored on site for 48 hours at full capacity, paired with continuous fuel-delivery availability.

These details turn a generic resilience claim into components that can be tested. A buyer can ask which two power paths feed its rack, whether both were commissioned under load, how maintenance is isolated, how generator starting is tested, how fuel quality and resupply are governed, and what happens after 48 hours during a regional disruption. Seven minutes of UPS autonomy is not a weakness or strength in isolation; it is the designed bridge to generator power. The evidence that matters is whether the transition, batteries, generators, controls and replenishment process work together under the agreed load.

The same applies to cooling. N+1 describes spare component capacity relative to a design requirement. It does not show current utilisation, heat distribution at a particular rack, maintenance condition or recovery after a controls failure. A 21 kW rack needs a placement and cooling plan, not simply the site's maximum-density label. Acceptance should include environmental readings at the chosen position, alarm thresholds, response ownership and the capacity reserved through the contract term.

Automation sits behind much of this operating model. Building-management systems, energy metering, environmental monitoring, access control and remote-hands tickets turn physical events into alerts and authorised actions. Raxio's sheet says thermal conditions and building-management systems are monitored continuously and describes proximity-card and biometric access. Those systems reduce routine labour only when sensor coverage, alert routing, access review, clock accuracy and fallback procedures are maintained.

A customer should know which records it can receive, how long they are retained and who can override a control during an emergency.

Capacity figures show why the dated service schedule matters

Raxio's public material does not present one immutable capacity snapshot. The downloadable technical sheet says on its opening page that the 1,000-square-metre facility can house up to 250 racks and deliver 1.5 MW of IT power. Elsewhere in the same sheet, a summary says up to 400 racks. The current Uganda page also says the site is designed for up to 400 racks at full scale. That may reflect a distinction between fitted capacity, layout options and eventual expansion, but the public documents do not reconcile it.

Other figures move as well. The PDF says 1.5 MW of installed cooling provision initially and 11 connectivity providers. The current web page says 2 MW of cooling and 15 providers. A changing number can be a positive sign of expansion or a routine publishing lag. It can also refer to different definitions: contracted providers, physically terminated fibre, active services or companies shown in a partner carousel. None of those possibilities should be guessed into fact.

For a buyer, the operational number is the one attached to the purchased hall and date. Total planned racks do not show available rack positions. Installed IT power does not show sellable headroom. A carrier logo does not show a lit path to the customer's cage. The useful capacity schedule states what is built, commissioned, occupied, reserved and available, with power and cooling headroom under the relevant failure state.

The variations also affect migration economics. Colocation replaces capital spending on a building but does not eliminate capacity planning. Customers still need to forecast rack count, power draw, cross-connects, ports, hands-on visits, spares and exit work. If a workload grows from half a rack to a private cage, the provider should show whether adjacent space, power and cable routes can be held. If capacity cannot expand in place, a second internal move can recreate the disruption the customer hoped to avoid.

Published specifications are therefore a good start, not a final bill of materials. The quotation should carry a revision date and identify the actual rack, feeds, metering, power cap, cooling class, fibre path, cross-connect count, access entitlement and remote-hands allowance. It should also state which expansion claims are commitments and which are merely possible future configurations.

Tier language must be read at certificate scope

Raxio's web page calls UG1 Tier III certified and associates that wording with 99.98 per cent uptime. The technical sheet is narrower: it says the site has a Uptime Institute Tier III certified design that is concurrently maintainable with no single points of failure, and elsewhere promotes a 99.982 per cent uptime commitment. PeeringDB's facility note uses 99.9 per cent. These public descriptions are close enough to sound consistent in casual reading, but they are not identical service terms.

Tier certification has a defined scope, and design review is not the same evidence as a constructed-facility assessment or sustained operating performance. The frozen public material did not include the underlying UG1 certificate, its award type, issue date, expiry treatment or scope. A buyer should request that certificate directly and verify it against the exact building and current configuration. The service agreement should then define availability separately: measurement point, calculation period, maintenance exclusions, upstream and customer exclusions, claim process and service credit.

Concurrent maintainability also needs translation into the customer's path. Two power chains can exist while a customer plugs equipment into one feed. Diverse cable entrances can converge in a single customer router. A redundant cooling topology can still leave a local hot spot. A facility design can be strong while a cage or cross-connect is installed in a way that defeats it. Acceptance testing should trace both rack feeds, both carrier routes where purchased and the control points used during maintenance.

The public security description is similarly concrete but incomplete. The site lists a perimeter wall, road blocker, guards, closed-circuit television, intruder alarms, card and biometric readers, a secure delivery area, early smoke detection and gas suppression for technical areas. Those are meaningful layers. They do not disclose access-review frequency, visitor sponsorship, camera retention, incident history, lost-card handling, biometric fallback, contractor control or whether a customer can retrieve its own access evidence.

This is not an argument for publishing sensitive design details. It is an argument for controlled customer assurance. A regulated buyer can review certificates, test reports, maintenance records and access samples under confidentiality. A smaller buyer can still walk the route from gate to rack, inspect its feeds, observe an access request and review a recent maintenance or generator-test record. The public claims identify what should be demonstrable.

AS328821 is operating evidence, not a facility warranty

The network record is one of the strongest parts of Raxio Uganda's public identity. AFRINIC assigned AS328821 and 102.220.156.0/23 to the exact local company in March 2021. RIPEstat's autonomous-system overview marked AS328821 announced on July 15, 2026. Its announced-prefix record showed the /23 and both component /24s visible throughout the returned July 1 to July 15 window.

The routing-status observation for the aggregate identified AS328821 as origin, dated the first observation to May 2021 and showed the route visible to all 326 reporting IPv4 peers in that dataset. This is direct evidence that the company-linked network identity was being used, rather than sitting as an unannounced allocation. The two more-specific routes were also originated by AS328821.

The observed external path was narrower than the facility's carrier menu. RIPEstat's routing-consistency response showed AS37100 in BGP as the immediate peer for AS328821. AS37100 belongs to SEACOM, which Raxio lists among providers at UG1. This says the public prefixes reached global collectors through SEACOM at observation time. It does not show the paths purchased by a colocation customer, and it does not prove that Raxio lacks dormant, private or physically diverse connectivity.

Carrier neutrality is a choice surface, not a claim that every route uses every carrier. The site says 15 connectivity providers terminate fibre, and it names two meet-me rooms, two fibre intake points and diverse cable pathways. A customer can use that environment to buy separate services. The customer still has to select carriers, order cross-connects, configure routing and test failure. If both circuits depend on one duct, one metro fibre operator or one remote route, two invoices may not yield useful diversity.

The route-origin security result also needs precision. RIPEstat's RPKI validation endpoint returned unknown for the /23, with no validating route-origin authorisation in its view. Unknown is not invalid: it means the observation did not find a cryptographic authorisation against which to validate this origin. The route remained globally visible. A buyer concerned with routing hygiene should ask whether Raxio plans to publish authorisations for the aggregate and more-specifics, how route changes are approved and how filters are maintained with its upstreams.

None of this route evidence measures rack uptime. A prefix can remain visible while an application, server, cross-connect or power feed fails. A facility can remain healthy while an upstream route is withdrawn. Network records identify an operating control surface and one dependency. Service assurance requires mapping the customer's addresses and circuits onto that surface and deciding who restores each layer.

UIXP makes locality useful, but not automatic

The local interconnection evidence extends beyond Raxio's own marketing. In February 2022, UIXP reported that it had expanded into Raxio, with Google as the first peer at the new location and an inter-site link intended to let networks exchange traffic from either UIXP point of presence. UIXP also said it deployed servers at Raxio for operational systems and local services. That is direct evidence of exchange infrastructure and equipment placed at the facility.

The current UIXP connected-networks page lists Raxio Data Center, AS328821, as an open-policy entity since 2022. PeeringDB's facility page lists UIXP and eight networks at UG1, including RENU, Roke Telkom and several regional providers. The exact facility population in public directories can lag or use voluntary reporting, but the independent records agree on the central fact: UG1 is part of Uganda's interconnection environment, not merely a building that advertises carrier access.

That matters because local traffic can avoid unnecessary international transit when content and users are connected to the same exchange ecosystem. It can reduce latency, international capacity use and exposure to distant route failures. It can also make local copies, caches and public services more practical. Yet a rack at UG1 does not make an application locally routed by itself. The customer's network or provider must connect, advertise the right prefixes and exchange traffic under suitable policy.

Data locality has an even wider boundary. Hosting a server in Namanve establishes a physical primary location if the contract and inventory identify that server. It says nothing by itself about backup copies, cloud control planes, monitoring data, support tickets, administrator access, security logs, payment records or disaster-recovery systems. Raxio's page advertises access to public, private and hybrid cloud providers, which can be valuable precisely because data and control may cross supplier boundaries.

A sovereignty schedule should therefore name the local legal processor, primary facility, backup sites, support-access countries, subprocessors, retention periods and deletion evidence. It should distinguish customer content from telemetry and account data. If a customer requires all copies to remain in Uganda, that condition must cover remote support and backups, not only the rack address. If regional recovery is allowed, the countries and transfer controls should be explicit.

Locality also depends on customer control. The buyer should know who owns domain registration, DNS, encryption keys, address space, carrier contracts and recovery media. Raxio's /23 belongs to the provider; a customer should not assume addresses are portable on exit. A migration plan may need renumbering, DNS changes and new allow-list entries even if the hardware can be moved. The value of a local facility is strongest when these dependencies are documented before an incident.

Remote hands turns resilience into a labour question

Raxio says trained engineers provide remote hands around the clock and that its operations, security and facilities functions run continuously. The current Uganda page names a local general manager and a vice-president responsible for technical operations. PeeringDB publishes a Ugandan technical email and telephone number as well as separate sales contacts. These are useful signs of accountable local presence.

They are not performance records. The public sources reviewed here did not state severity definitions, acknowledgement targets, restoration targets, remote-hands response times, included monthly time, after-hours charges, spare-parts inventory or the number of staff on a shift. They also did not publish achieved availability, incident frequency, maintenance outcomes or customer restore tests. A continuously open channel can still depend on an engineer, carrier or vendor who is not immediately available.

Colocation makes the division of labour particularly important. Remote hands may be permitted to read an indicator, reseat a cable, replace a customer-supplied disk or escort a vendor. It may not include diagnosis, operating-system access, network reconfiguration or application recovery. A rushed instruction can also cause damage if rack labels, authorisation and change records are weak. The service should define who may request work, how identity is checked, whether a second approval is required and how completion is evidenced.

Local support quality can be tested without manufacturing an emergency. During acceptance, open a ticket through the ordinary and urgent channels. Ask the team to identify a rack, verify both feeds, report an environmental reading and execute a reversible physical task. Review the timestamps, photographs or console evidence and escalation. Then test a scenario that crosses boundaries, such as a failed carrier circuit or customer server, to confirm that Raxio can reach the responsible supplier while keeping the customer informed.

Recovery also requires spares and authority. If customer equipment fails, the site can only replace what is available and authorised. The customer should place labelled spares on site or agree a vendor response, document compatible parts and maintain secure access instructions. If the problem is facility power, cooling or access control, Raxio needs its own runbooks and escalation. If the problem is routing through SEACOM, the network team needs a carrier path. One support number should lead to the correct owner rather than hide the chain.

The most revealing support metric is not how fast a ticket is closed. It is how quickly service is restored safely, with the cause, actions and remaining risk recorded. A response-time target can encourage a quick acknowledgement. A restoration objective, update interval and post-incident review make the labour promise operational. Buyers should ask for all four.

Assurance should follow the service a customer actually buys

RAXIO DATA CENTRE SMC LIMITED has a credible public foundation. The legal name, local address, autonomous system and IPv4 block line up. The facility specification identifies power, cooling, security and connectivity components. UIXP confirms a meaningful interconnection role. PeeringDB shows a populated facility and current contact surface. These facts justify treating UG1 as operating infrastructure rather than as a name awaiting proof of existence.

The next step is not to turn every positive clue into a blanket guarantee. It is to assemble a service-specific evidence chain. For identity, obtain the current company extract, signing authority and exact contractual counterparty. For the facility, obtain the current Tier certificate and scope, commissioned-capacity statement, rack assignment and maintenance evidence. For power and cooling, trace the customer's feeds and review recent failover and generator tests. For connectivity, identify carriers, physical paths, addresses, route policy and acceptance results.

For data control, attach primary and backup locations, subprocessor and support-access schedules, retention and deletion terms, encryption ownership and an exit plan. For automation, obtain access, monitoring and change records relevant to the customer's service. For support, agree severity, response, update and restoration clocks; authorised requesters; remote-hands tasks; spare-parts responsibilities; and escalation beyond the site. Finally, run a restore or failover exercise before the workload becomes difficult to move.

The public inconsistencies should be resolved in that process, not treated as a verdict. A site can grow from 11 to 15 carriers or from an initial fit-out toward 400 racks. A web page can round an uptime figure differently from a technical sheet. A design certificate can coexist with later facility certification. The provider should be able to state which explanation applies and supply the dated record.

The central judgement is therefore balanced. RAXIO DATA CENTRE SMC LIMITED has more public operating evidence than its formal name alone suggests: a real Namanve facility, local exchange participation, live routed resources and a detailed physical-service proposition. The remaining gap is customer-level assurance. UG1 becomes dependable for a workload when the exact legal promise, fitted capacity, route, data location, automated controls and human response can all be followed from a public clue to a tested contractual outcome.