Summary

  • SG1 CyberWorld Data Center Co.,Ltd. is associated with AS136569, an APNIC-registered autonomous system named CyberWorld-SG1. As of 12 July 2026, RIPE routing observations showed no globally visible IPv4 or IPv6 prefixes and no observed neighbours for that ASN.
  • The wider picture is more substantial but must not be collapsed into the profile. A sibling registration, AS136565, is active in Singapore, originates two IPv4 and three IPv6 prefixes, and is listed by PeeringDB at Equinix SG1 and SG3. That supports a CyberWorld network presence in Singapore; it does not prove that AS136569 is live or that CyberWorld owns the Equinix facilities.
  • Thai corporate records describe CyberWorld Data Center Co.,Ltd. as an active Bangkok-registered provider of wired and wireless internet communications. Another CyberWorld ASN, AS136566, is actively routed from a Thai registration. These are useful identity and operating signals, but they do not disclose rack inventory, customer load, utility topology, generator endurance or recovery performance.
  • The operating boundary is the central issue. Equinix publishes the SG1 building's power, cooling and fire specifications; those are facility-operator claims, not automatically CyberWorld service commitments. A tenant can still create a single point of failure through one cabinet, one router, one cross-connect, one power cord or one undersized transit path.
  • The evidence grade is Weak. There is enough public evidence to justify technical diligence, but not enough to treat the named SG1 capacity as demonstrated, currently routed or independently recoverable.

The important fact is the mismatch

Infrastructure stories often begin with a large number: megawatts, racks, square metres or route capacity. This one begins with a mismatch. The APNIC registration for AS136569 names CyberWorld-SG1, assigns the country code TH, and links the number to CyberWorld Data Center Co.,Ltd. at a Bangkok address. The RIPEstat routing-status view, however, showed no visible IPv4 or IPv6 announcements on 12 July 2026. The announced-prefix view returned an empty prefix list, and the neighbour view returned no observed adjacent networks.

Those facts do not show that CyberWorld has no Singapore business. They show something narrower: the autonomous system that gives this profile its SG1 identity was not functioning as a visible origin network in the public routing table at the time of observation. It might be reserved for a future deployment, retained after a migration, used only in a private context, or simply dormant. None of those possibilities can be selected from the registry record alone.

That distinction matters because an ASN is an administrative identifier, not a data hall. A company can hold an ASN before buying transit. It can keep an ASN after moving traffic elsewhere. It can operate equipment inside another company's building without owning the building. It can also deliver a hosted service using provider-assigned addresses that never appear under its own ASN. The visible absence of routes is therefore not a verdict on every service the company may sell. It is a direct limit on what can be claimed about AS136569.

The mismatch is commercially useful. It directs attention away from the reassuring shape of the name and towards the chain that customers actually depend on: legal entity, contracted facility, powered cabinet, edge router, cross-connect, carrier, address origin, support desk and recovery plan. Every link must be current. A name can persist while any one of those links changes.

A real company record is only the first layer

The corporate identity has more substance than a stray network label. Data for Thai, drawing on Thai government open data, lists CyberWorld Data Center Co.,Ltd. under registration number 0105552000659, records incorporation on 6 January 2009, and describes the business as providing internet communications in wired and wireless forms. It lists the company as active and gives a registered address at 12/32 Phaholyothin 48 in Bangkok. A second Thai company-data view reports financial filings through 2024 and categorises the company as a small information-and-communications service business.

That is meaningful evidence of legal continuity. It establishes that the name corresponds to a registered Thai company rather than an untraceable marketing label. It does not establish the scale or location of its technical estate. Registered capital is not data-centre power capacity. A filed business purpose is not an inventory of operating racks. An active corporate status is not proof that a particular network, product or overseas deployment is currently serving customers.

The distinction is especially important here because the legal and network records span Thailand and Singapore. The APNIC entity record links the organisation handle used by several CyberWorld ASNs, but individual network records have different country codes, contact sets and operating labels. The evidence supports common registration context. It does not justify treating every numbered resource under that handle as the same site, service or failure domain.

The minimum responsible description is therefore modest: CyberWorld Data Center Co.,Ltd. is an active Thai company with a long-lived internet-services registration and several APNIC number-resource records. Whether the subject called SG1 represents live customer capacity requires facility and routing evidence beyond the corporate filing.

The active Singapore network sits next door in the records

The strongest positive signal is not AS136569 but AS136565. Its APNIC record names CyberWorld, uses country code SG, gives 20 Ayer Rajah Crescent in Singapore, and links to the same organisation handle. RIPE's routing-status record for AS136565 showed two IPv4 prefixes, representing 512 addresses, and three IPv6 /48 prefixes visible on 12 July 2026. The announced-prefix list included 103.20.122.0/24, 103.20.123.0/24, 2405:2500:25::/48, 2405:2500:26::/48 and 2405:2500:65::/48.

Visibility was broad in the RIPE collector view, and the last-seen observation was current on the publication date. IPinfo's AS136565 page independently describes a small hosting network, lists the same two IPv4 blocks, and reports responsive addresses measured from Singapore. That is much stronger evidence of an operating Singapore network surface than the SG1 label on a dormant ASN.

The adjacency picture is small. RIPE observations showed AS6939 and AS33229 on the left side of paths to AS136565. IPinfo classifies Hurricane Electric's AS6939 as an upstream and also records AS33229. These observations support internet reachability through at least two visible adjacent networks, but public path inference is not a substitute for contracts or fibre maps. It cannot prove whether both paths carry full routes, whether both have adequate commit during a failure, whether they terminate on separate routers, or whether their cables enter the building through diverse ducts.

This is the article's most important boundary. The active AS136565 evidence strengthens the case that CyberWorld has operated a network in Singapore. It does not make AS136569 active by association. It also does not reveal which customers, if any, are placed behind either number. A buyer should insist that a proposal name the production ASN and prefixes, rather than accepting a family of related registrations as interchangeable proof.

SG1 appears to be a location claim, not an ownership claim

The Singapore address in the active ASN record is revealing. Twenty Ayer Rajah Crescent is Equinix SG1, a nine-storey colocation facility operated by Equinix. PeeringDB's facility record also identifies the address and operator. CyberWorld's PeeringDB network profile lists AS136565 as a network service provider with two IPv4 and three IPv6 prefixes, and records facility presence at Equinix SG1 and Equinix SG3.

That combination supports a specific, useful statement: a CyberWorld network has declared presence in two Equinix Singapore facilities, and its live routes are consistent with a Singapore operating surface. It does not support the stronger statement that CyberWorld owns, builds or controls those data centres. Equinix is the facility operator. CyberWorld may be a tenant, customer, reseller, network entity or service provider within the buildings. The exact commercial role is not publicly disclosed.

This ownership line changes how resilience should be assessed. Equinix can be responsible for the building shell, utility intake, generators, uninterruptible power, cooling plant, fire systems, security and cross-connect fulfilment. CyberWorld can be responsible for its cabinets, rack power distribution, routers, servers, optics, configurations, transit orders and customer support. A customer service survives only when both layers work and when the contract correctly joins them.

The PeeringDB profile is self-maintained and was last updated in October 2022. It discloses no traffic volume, no looking glass and no public exchange attachment for the network. Facility presence is still valuable evidence, but it is not a current cabinet schedule. The customer should ask whether both listed facilities remain in use, which one carries production, whether workloads are duplicated, and whether the two sites can operate independently.

The label SG1 can therefore be read as a clue to place. It should not be read as a deed to the building or a certification of the tenant's service.

Facility specifications do not automatically flow through to the tenant

Equinix publishes substantial specifications for SG1: 10,146 square metres of colocation space, 2N UPS and generator redundancy, N+1 cooling, pre-action fire suppression, aspirating smoke detection and elevation above the stated 100-year base flood level. Those details make the facility legible. They still do not tell a CyberWorld customer how its own service is wired.

A 2N building can contain a single-corded server. Two independent power paths can meet in one rack transfer switch. A tenant can order one cross-connect, place both edge routers in one cabinet, connect both carriers through one patch panel, or use a management server that depends on the same production path it is meant to repair. The facility's design reduces common risks; it does not eliminate bad or economical choices inside the tenant boundary.

The distinction between redundancy and maintainability is well described by the Uptime Institute's Tier framework. Tier III is about removing any capacity component or distribution path for planned maintenance without affecting IT operations; Tier IV adds tolerance of an unplanned failure. Uptime also warns that a certified facility can still experience IT outages if systems are not operated and managed effectively. No public evidence reviewed here shows an Uptime certification held by CyberWorld for the SG1 service, and no such certification should be inferred from the host facility's marketing.

The customer-level proof should be a one-line diagram and a rack-level allocation. It should show A and B power from the facility handoff to every critical device, including any automatic transfer switch used for single-corded equipment. It should show the failure state when one feed, one PDU, one UPS path or one generator train is unavailable. It should also show that monitoring, console access and out-of-band management stay alive in that state.

Without that evidence, the building specification remains encouraging context rather than a service guarantee.

Installed capacity is not sellable capacity

Data-centre capacity is commonly described at three different levels that should never be confused. Design capacity is what a completed architecture is intended to support. Installed capacity is what has been built and energised. Sellable or usable capacity is what can be allocated to customers while preserving the promised redundancy, cooling margin and maintenance state. Recoverable capacity is what remains after a fault.

Public evidence for the subject does not disclose any of those numbers. There is no verified rack count, contracted kilovolt-ampere allocation, measured IT load, generator allocation, cooling envelope or sold-versus-available figure tied to AS136569. The absence is not proof that capacity does not exist. It means any statement of quantity should come from a current facility schedule, invoice, acceptance record or metered report rather than from the company name.

The difference becomes acute when density rises. A cabinet may have physical room for more servers but no remaining A-feed or B-feed headroom. A hall may have spare floor tiles while cooling distribution prevents another high-density rack. A tenant may have a 10-gigabit port but an upstream commit or firewall that cannot sustain that load. A provider can therefore advertise a nominal unit of capacity that is not safely deployable under the contracted resilience model.

Customers should request normal-state and degraded-state figures. Normal-state utilisation should be shown separately for A power, B power, cooling allocation, edge ports and committed transit. Degraded-state headroom should be calculated after the largest expected component or path is removed. If a service is sold as concurrently maintainable, the provider should demonstrate that maintenance does not consume the headroom reserved for a subsequent failure.

For SG1, current route absence adds another measurement: the provider should show which active ASN and prefixes carry the service. A powered rack without a production route is not usable internet capacity. A route from a sibling ASN may be entirely legitimate, but the contract and network diagram must say so.

Power proof begins below the utility headline

At the building level, Equinix states 2N power redundancy for SG1. That is a strong facility claim. At the tenant level, CyberWorld still needs to show how the service receives and uses it. The relevant path runs from utility and generator systems through switchgear, UPS, distribution, cabinet PDUs, device power supplies and the management plane. A fault at the narrowest part of that path defines the service outcome.

Dual utility feeds are often used as shorthand for resilience, but utility count alone can mislead. Two feeds may share a substation or transmission exposure. They may enter the same switchboard. They may be operationally unavailable during a common grid event. The Uptime Institute's discussion of Tier misconceptions emphasises onsite generation because public utility supply can be interrupted even in reliable markets.

Generator endurance is also more than a tank-size number. Equinix's public SG1 specification gives a diesel-storage figure, but a CyberWorld customer needs the service consequence: which tenant loads are on backed-up paths, how long the site can run at the relevant load, what refuelling contracts exist, whether fuel pumps and controls share failure points, and how runtime changes when cooling is included. The Uptime Institute's fuel-system guidance treats tanks, pumps, valves, controls and distribution as one topology and uses 12 hours at design load as a baseline for Tier-defined sites. That is a method for asking questions, not proof of CyberWorld's allocation.

Testing matters because transfer events reveal dependencies that diagrams miss. The evidence package should include recent generator-load tests, UPS battery or energy-storage tests, automatic-transfer results and the exceptions discovered. It should state whether customer equipment participated or whether only facility plant was tested. A test that excludes tenant racks cannot demonstrate tenant recovery.

Cooling is an availability system, not an efficiency footnote

Singapore's climate makes heat rejection a primary operating constraint. The Infocomm Media Development Authority has noted that cooling can account for up to 40 per cent of a typical data centre's energy use. Higher server density compresses the time available to respond to a cooling interruption because racks continue turning electrical power into heat even when chilled-water flow or air movement is impaired.

The consequence is not theoretical. In October 2023, Channel NewsAsia reported that a chilled-water issue during a planned upgrade at an Equinix facility in Singapore raised temperatures in some halls and affected customer operations, including banking, payment and mobile services. The report did not identify SG1, so it would be wrong to attach the event to this location. Its relevance is systemic: cooling maintenance can propagate from a facility subsystem into nationally visible digital services.

CyberWorld's assurance should therefore cover both capacity and response. What inlet-temperature and humidity ranges are monitored at its cabinets? Are alarms available to CyberWorld directly or only through the facility operator? Which team has authority to shed load, shut down equipment or move traffic? How long can the racks remain within acceptable conditions after a loss of active cooling? Are both production and management systems inside the same thermal zone?

The ASHRAE data-centre guidance gives equipment-environment ranges and design context, but a standard does not replace measurements from the actual cabinets. A useful customer record would include sensor placement, recent temperature trends, alarm thresholds, escalation times and a cooling-failure exercise. If the deployment spans SG1 and SG3, it should also show that both sites do not depend on a shared customer-side control system that can fail first.

Carrier diversity has four separate meanings

The public network evidence supports some connectivity, but not yet resilience. AS136565 has visible paths through AS6939 and AS33229, while AS136569 has no visible neighbours. A buyer should divide carrier diversity into four tests.

The first is contractual diversity: are there two independent providers with current agreements and support channels? The second is routing diversity: do both paths carry the routes needed to keep the service reachable, and can the remaining path accept the full production announcement? The third is physical diversity: do the circuits use separate building entrances, meet-me rooms, risers and outside-plant routes? The fourth is capacity diversity: can one path carry the critical traffic load after the other is removed?

Passing one test does not pass the others. Two BGP sessions can ride one cross-connect provider. Two carrier brands can lease the same metro fibre. Two ports can terminate on one router. A backup link can be technically live but commercially capped or too small for peak demand. Route collectors see internet paths; they do not see shared ducts, patch fields or emergency support entitlements.

PeeringDB shows hundreds of networks and multiple exchanges at Equinix SG1, which makes the building a rich interconnection location. Yet CyberWorld's own profile lists facility presence without a public exchange attachment. That is not a defect: a network can buy private transit or cross-connects without joining an exchange. It does mean the building's carrier density should not be mistaken for the tenant's purchased diversity.

The most persuasive test is controlled withdrawal. CyberWorld should be able to remove each upstream or edge router in turn, show route convergence from several external vantage points, measure packet loss and latency, and demonstrate that the surviving path is not congested. The customer should see the same test during a realistic traffic period, not only in an empty maintenance window.

The Thai network provides corroboration, not a recovery site

CyberWorld's Thai network AS136566, named CyberWorld-BKK2, is another positive operating signal. Its APNIC record uses the same Thai company identity, and the RIPE routing view showed one IPv4 /24 and two IPv6 /48 prefixes on 12 July 2026. Public path observations included AS55423 and AS45642 as adjacent networks.

This shows that the company identity is attached to an active Thai routing surface as well as the active Singapore sibling ASN. It does not show that Bangkok backs up Singapore. Disaster recovery requires workload replication, adequate compute and storage, compatible network policy, tested failover and a recovery objective. Two active ASNs in different countries can still carry unrelated services.

The Bangkok address trail also needs care. Historical APNIC contact information refers to 90 CyberWorld Tower on Ratchadaphisek Road. Public colocation directories and an AIS/CSL data-centre brochure identify a facility at CyberWorld Tower operated by CS LoxInfo, now CSL/AIS. Similar names and addresses are not proof that CyberWorld Data Center Co.,Ltd. owns that facility or that its BKK2 network is located there today.

A buyer considering a Thailand-to-Singapore recovery design should ask for exact site names and operators, not a city pair. It should confirm who holds each cabinet contract, where replication terminates, which prefixes are announced at each site, how DNS or anycast changes occur, and whether staff can access both locations during the same regional incident. Until that is shown, the Thai network is corroboration of company activity, not verified failover capacity for SG1.

Thailand's data-centre rules show what serious capacity looks like

The Thai policy context provides a useful benchmark even if the Singapore presence is housed in Equinix. Thailand's 2025 investment-promotion guide sets demanding conditions for promoted data-centre projects: at least 2 MW of IT load, multiple domestic and international telecommunications systems, concurrently maintainable service, continuous-rated generation, backup power and cooling, independent distribution paths, fire protection and round-the-clock security.

Those conditions are not proof that CyberWorld sought or received investment promotion. They should not be imposed retrospectively on every small network service. They are useful because they translate the phrase "data centre" into measurable infrastructure. The policy distinguishes a facility project from a company name by requiring power, connectivity, maintainability and safety together.

The current market raises the value of that discipline. Thailand's Board of Investment reported a surge of large data-centre applications in 2025 and 2026. A 2025 BOI announcement described approved projects with disclosed IT-load figures, while a 2025 Asian Development Bank financing announcement identified a 25.6 MW colocation project in Samut Prakan and a designed power-usage-effectiveness target. Large projects now make capacity claims concrete enough to finance and regulate.

That environment can leave smaller providers at a narrative disadvantage, but the answer is not to borrow scale from a category label. A small deployment can be useful, resilient and commercially attractive. It should disclose its actual unit of sale: cabinets, kilowatts, servers, transit, managed service or remote hands. It should also identify the facility operator and the limits of its own responsibility.

For SG1 CyberWorld, the best improvement would be precision rather than grandeur. A clear statement that CyberWorld operates network equipment in named third-party facilities, with current routing and support terms, would be more credible than an unqualified suggestion of owned data-centre capacity.

Power availability must be confirmed, not inferred from a map

Power constraints are becoming part of site selection and service lead time across Thailand. The Metropolitan Electricity Authority's load-capacity map explicitly warns that its displayed capacity is preliminary, can change, and does not guarantee supply or account for required network construction and upgrades. That disclaimer captures the difference between apparent and deliverable capacity.

The same principle applies inside a colocation building. Published building capacity does not guarantee that a particular customer can add another cabinet on the requested date. The available power may be reserved, stranded behind distribution constraints, incompatible with the requested redundancy, or contingent on construction. A tenant can have a signed commercial relationship without an energised circuit.

For any new CyberWorld commitment, customers should request an order-specific ready-for-service record. It should identify cabinet, A and B feed ratings, breaker sizes, usable continuous load, commissioning date and any restrictions. If the service is not yet installed, the schedule should distinguish facility approval, cross-connect order, equipment delivery, configuration, burn-in and customer acceptance. Calling all of those steps "capacity" hides construction and activation risk.

The Energy Policy and Planning Office's direct-power pilot discussion also illustrates how data-centre demand increasingly interacts with generation and grid policy. Such national measures do not tell us anything specific about CyberWorld's power contract. They show why a provider must name the utility and facility boundaries behind a service rather than treating electricity as an unlimited input.

Maintenance is where advertised redundancy is consumed

Many severe incidents occur not in steady state but during maintenance. A redundant component has already been removed, an alarm is suppressed, a valve or route is in an unusual position, and an unexpected fault lands on the reduced system. The October 2023 Singapore cooling incident is a vivid example of the general pattern: planned work can expose a common dependency and reduce the time available for recovery.

CyberWorld customers need to know who controls each maintenance window. Equinix may schedule building work. A carrier may schedule fibre or router work. CyberWorld may update its own routers and servers. If those calendars are not coordinated, two individually acceptable windows can overlap and remove the intended fallback.

Evidence should include a change calendar, customer-notification threshold, freeze periods for critical business dates, and an escalation route for emergency deferral. It should identify which maintenance states have been tested at full customer load. It should also state whether the provider can see facility alarms and carrier notices directly, or receives them only after an intermediary forwards them.

The technical test is straightforward: take each maintainable component or path out of service and show that the promised load remains supported. Then identify the failures that would still cause interruption in that state. Honest providers do not need to promise impossibility. They need to tell customers which risks remain and how long restoration should take.

For AS136569, maintenance history could also explain the lack of routing, but no public statement does so. A route that is intentionally dormant should be labelled as such. A route intended for standby should be exercised and monitored. Without that explanation, customers cannot distinguish a deliberate cold reserve from an abandoned identifier.

Fire and flood risk cross organisational boundaries

Equinix publishes fire detection and suppression details for SG1 and says the facility is elevated above the stated 100-year base flood level. These are relevant design signals. They do not eliminate risk to a tenant's service or equipment.

Fire protection has layers: early detection, suppression, compartmentation, emergency power isolation, evacuation, re-entry and restoration. A small smoke event can interrupt service if staff cannot access the affected zone, if equipment is shut down protectively, or if residue and water damage require inspection. A tenant should know whether its racks span fire zones and whether critical spares are stored outside the same compartment.

Flood resilience also extends beyond the building floor. A site can remain dry while utility substations, fuel deliveries, access roads or terrestrial fibre routes are affected. The relevant question is not only whether water reaches the rack. It is whether the facility can continue receiving power, fuel, people and connectivity during the event.

The public record does not disclose CyberWorld's site-specific fire or flood plan. A customer should request the facility emergency interface, equipment insurance responsibility, spare-location plan and recovery sequence. It should also check that any second site does not share the same regional hazard or supplier chain.

These questions are not an attempt to turn a small network operator into a building owner. They recognise the opposite: when the provider rents critical infrastructure, recovery depends on clear handoffs between organisations. The customer needs to know who makes each decision and who communicates it.

Route security is one layer, and the dormant ASN has none to validate

For an active prefix, Route Origin Authorisation can reduce the risk that networks reject or misinterpret the intended origin. The active AS136565 prefixes are reported as RPKI-valid by public routing aggregators, which is a positive control-plane signal. The AS136569 record has no visible prefixes, so there is no current origin pair to validate in the public table.

This difference reinforces the need to contract against the network actually used. A customer should receive the production origin ASN, prefix list, expected upstreams and contact for route incidents. It should monitor for withdrawals, unexpected origins and changes in adjacency. If CyberWorld uses provider-assigned space, the provider ASN and routing responsibility should be stated instead.

RPKI cannot prove facility resilience. A perfectly authorised route can lead to an unreachable server. It does not protect against power loss, cooling failure, a bad access-control list, a severed cross-connect or limited public evidence transit capacity. Its value is narrower: it makes the intended route origin more explicit and gives other networks a basis for rejecting invalid announcements.

For SG1 CyberWorld, the public routing evidence therefore produces two different conclusions. The active sibling network demonstrates sustained Singapore reachability and some routing hygiene. The named AS136569 offers no present route surface to assess. Those conclusions should remain separate until CyberWorld publishes or demonstrates a service mapping between them.

Who is affected when the chain breaks

A small visible network can support consequential services. The public route table does not disclose whether CyberWorld carries enterprise applications, websites, managed servers, VPN endpoints, reseller infrastructure or internal systems. It therefore cannot quantify the number of affected users. It can identify the likely impact mechanisms.

If facility power fails beyond ride-through and generator support, customer equipment becomes unreachable and may shut down abruptly. If cooling fails, thermal protection may throttle or stop systems even while electrical power remains available. If a cross-connect or carrier path fails, servers can remain healthy but isolated. If the management path shares the production failure, engineers may lose the console needed to restore service. If support escalation stalls between CyberWorld, Equinix and a carrier, a technically simple repair can become a prolonged outage.

Downstream effects depend on the workload. A hosted identity service can lock users out of other systems. A DNS or VPN endpoint can make healthy applications appear down. A reseller can propagate one facility fault to many customers that have never heard of the underlying operator. Backup and replication jobs can fail silently before a later incident reveals the missing recovery point.

The 2023 Singapore cooling event showed how a shared facility issue could become visible through banking, payment and mobile-service disruption. There is no evidence that CyberWorld was involved in that event. The lesson is about concentration: customers must design their own service placement so that a facility incident does not become a business-wide incident.

CyberWorld's responsibility is to disclose its part of the chain and test it. The customer's responsibility is to decide whether one site, one provider or one management domain is acceptable for the workload.

What would raise the evidence grade

The Weak grade can be improved with a compact set of current evidence. First, CyberWorld should state whether AS136569 is planned, private, standby, migrated or retired. If it is not the production network for SG1 services, the company should identify the actual ASN and prefixes used. That single clarification would remove the largest ambiguity.

Second, it should document the facility boundary. A current colocation record or operator confirmation should show CyberWorld's presence at SG1 and SG3, without disclosing sensitive cabinet details publicly. The statement should distinguish facility operator, contracting entity and service provider. It should say whether both locations host production equipment or whether one is only an interconnection point.

Third, it should publish service-level quantities that customers can verify: energised cabinets or committed kilowatts, current route capacity, upstream roles, support coverage and installation lead time. These need not reveal customer names. They should separate installed, sold, available and degraded-state capacity.

Fourth, customers under confidentiality should receive one-line power and network diagrams. The diagrams should show A/B rack power, device power supplies, edge routers, cross-connects, carrier entrances, management access and site-to-site replication. Each shared component should be named. The provider should accompany them with recent failure-test results.

Fifth, CyberWorld should demonstrate recovery. Useful evidence includes a carrier-withdrawal exercise, generator or facility-maintenance participation, cooling-alarm drill, configuration restore, spare replacement and workload failover between sites. The record should give measured interruption and recovery times, not only a pass label.

None of these requests requires CyberWorld to pretend it owns Equinix. A transparent tenant can offer a strong service by combining a good host facility with disciplined network and operational controls. The aim is to prove that combination.

A practical customer test

A prospective customer can evaluate the service in four stages. The first stage is identity. Confirm the contracting company, registration number, invoice entity, support contacts and the relationship between the SG1 service name and CyberWorld Data Center Co.,Ltd. Confirm that the contract names the facility operator and jurisdiction where relevant.

The second stage is placement. Obtain the street-level facility identity, cabinet allocation, active ASN, prefix, upstreams and any second site. Verify that the production route is visible from independent networks. Check that the recovery site contains actual reserved capacity rather than a right to order it later.

The third stage is removal. Ask the provider to remove one power path, one edge router and one carrier in controlled tests. Observe traffic and management access. Confirm that surviving links have enough capacity and that alarms reach both provider and customer. Review the result after the environment has returned to normal.

The fourth stage is restoration. Restore a configuration or workload from a known backup, replace a failed component with an onsite spare, and practise escalation through the facility and carrier. Measure from customer-visible failure to service restoration. A service that cannot be restored in a test should not receive an untested recovery promise in the contract.

Customers should repeat these tests after material changes. Facility presence from 2022, a route seen in 2026 and a corporate filing from 2024 describe different clocks. Current assurance requires them to converge on the same service today.

The verdict is useful precisely because it is narrow

SG1 CyberWorld Data Center Co.,Ltd. should not be dismissed as a fictional infrastructure name. The Thai company is active in public corporate records. The organisation has long-held APNIC registrations. AS136566 has a live Thai route surface, and the related AS136565 has a small but visible Singapore network with declared presence at Equinix SG1 and SG3. Those are real signals.

The same evidence does not prove the proposition implied by this profile's name. AS136569 was not visibly announcing address space on 12 July 2026. No public record reviewed here establishes CyberWorld ownership of a Singapore data-centre building, a quantified SG1 power allocation, a rack count, customer capacity, generator runtime entitlement, tested facility failover or an active route under the named ASN. Equinix's SG1 specifications belong to Equinix's facility layer; CyberWorld must still prove its tenant and service layer.

That is why the final evidence grade is Weak, not Negative. The adjacent active network, matching organisation handle, Singapore facility declarations and Thai operating record make a plausible service footprint. They also expose exactly what remains unverified. A Negative grade would imply evidence against operation; the present evidence instead shows an unresolved mapping between company, ASN, facility and service.

The commercial conclusion is simple. Treat SG1 as a hypothesis that can be tested, not as capacity that has already been demonstrated. Ask which ASN is live, which cabinets are powered, which carriers are independent, which building operator controls the plant, and what happened in the last failover exercise. If CyberWorld can answer those questions with current records and measured results, a small network presence may be entirely serviceable. Until then, the name is more precise than the proof.