Summary
- The strongest identity evidence is the APNIC registration for AS18244, which names
STIDCas Stateline Internet Data Center Co., Ltd. Shanghai. Matching Stateline registrations cover the portable IPv4 allocations120.64.0.0/14,211.148.0.0/20,211.148.16.0/24and103.253.204.0/22, plus2403:7700::/32in IPv6. Those allocations are substantial administrative assets. They do not identify a data-centre building, rack count, utility service, cooling plant, generator reserve or customer contract. - AS18244 is not a broadly active public origin. RIPE's overview marks it unannounced, and its announced-prefix view contains no route visible to the service's minimum threshold. RouteViews also returns no current prefixes. RIPE does retain a low-visibility observation of
211.148.28.0/24under AS18244 in January 2026, so the ASN should be described as historically and intermittently used, not erased from the operating history. - Current reachability sits elsewhere.
120.65.0.0/16, one quarter of Stateline's registered/14, was visible to all 325 reporting RIPE IPv4 peers through China Unicom's AS4837 on 12 July 2026.103.253.204.0/22was visible to 324 of 325 peers through 21Vianet's AS17428. The Stateline-labelled211.148allocations were covered by a larger China Telecom AS4812 route, while the Stateline IPv6/32had no visible route. This is evidence of address use and carrier dependence, not proof of customer failover between independent paths. - The repeatedly published contact address, Room 1802 at CITIC Square, 1168 West Nanjing Road, is not a disclosed data hall. Shanghai's official visitor site describes CITIC Plaza as a grade-A office and commercial building. Public evidence does not identify the building that contains Stateline-operated racks, or establish whether Stateline owns a facility, leases wholesale space, supplies addresses to another operator, or retains legacy resources used by associated parties.
- Shanghai's rules make the missing physical facts material. The city's data-centre development policy ties network and power access to completed approvals, energy review, acceptance and machine-room safety assessment. It also calls for power, cooling, fire, lightning, flood and earthquake protection and regular emergency exercises. A credible capacity claim from Stateline would therefore need site-level evidence: operator and licence, energised rack count, dual-feed topology, UPS and generator test results, fuel runtime, cooling redundancy, fibre entrances, carrier contracts, recent maintenance history and demonstrated customer recovery.
The address estate is large, but the operating claim is small
Stateline is not a name inferred from an isolated hosting advert. It has a durable footprint in the internet-number system. AS18244 is registered as STIDC in China and carries the full company name. The entity was registered in 2008, although the public routing history reaches back to 2002, and it retains a contact at Room 1802, CITIC Square, 1168 West Nanjing Road in Shanghai. The same broad contact surface appears across several portable address allocations.
The largest is 120.64.0.0/14, a range of 262,144 IPv4 addresses registered as STIDC. A second allocation, 211.148.0.0/20, adds 4,096 addresses. The adjacent 211.148.16.0/24 adds another 256, and 103.253.204.0/22 adds 1,024. The IPv6 allocation 2403:7700::/32 is far larger in address terms, although IPv6 size should never be translated into a server count. Together, the records show that the Stateline name has been entrusted with meaningful numbering resources over many years.
That is the point at which a superficial assessment can go wrong. Addresses are not racks. A /14 does not reveal how many servers are installed, how many cabinets have passed electrical commissioning, how much cooling remains after the hottest design day, or how many customers can be restarted after a hall is lost. Address space can be delegated to customers, announced by a carrier, retained for future use, routed in only one portion, or administered under an arrangement in which another company controls the building and edge routers.
The allocation dates also resist a simple growth narrative. The largest IPv4 block dates from 2008; the IPv6 /32 from 2012; the 103.253.204.0/22 block from 2013; and the 211.148 registrations in their present form from 2014. Later changes to contact or maintenance fields show that records were updated, not that new server halls opened. The proper conclusion is therefore bounded: Stateline has long-standing internet-number assets and some of them remain in use. The public evidence does not quantify a current Stateline-operated data-centre estate.
A Shanghai office address is not a disclosed machine room
Every Stateline address record points back to the same central-city address. The official Shanghai tourism description of CITIC Plaza identifies 1168 West Nanjing Road as a commercial building built to grade-A office standards. A current CITIC procurement notice describes the property as a complex of shopping centre, offices and underground parking. That makes Room 1802 a plausible corporate or administrative address. It does not make it the location of the power-dense infrastructure implied by the company's English name.
This distinction is not pedantic. A customer assessing colocation risk needs the address where its equipment will stand, not only the address where notices may be received. The rack location determines the utility substation, flood exposure, fire compartment, seismic design, fuel-delivery route, fibre ducts, carrier meet-me room, security perimeter and travel time for replacement parts. Even a contract signed by the correct legal company cannot answer those engineering questions without a site schedule.
The absence of a disclosed site also obscures the operator boundary. Stateline could own a building, lease a private suite, rent cabinets from a wholesale provider, supply internet addresses to a related operator, or preserve legacy allocations while another company delivers the customer service. Each arrangement can be legitimate, but each assigns failure response differently. The party that invoices for a rack may not control the medium-voltage switchgear. The holder of the address allocation may not be able to dispatch the carrier technician. The company with a facility badge may not have authority to change the public route.
A useful disclosure would name the operating company at each building, the property owner, the licence holder, the party responsible for power and cooling, and the party responsible for cross-connects and external routes. It would give the exact service address under controlled access, not replace it with a central office address. It would also explain whether the STIDC resources serve those buildings or are used in facilities run by other companies. Until that map is available, the physical asset base must be treated as unverified rather than presumed from the name.
AS18244 has moved from a visible origin to an intermittent identifier
The historical routing record shows that AS18244 once did more than sit in a registry. RIPE's routing history for the ASN records 211.148.0.0/19 under Stateline from 2002, followed by more-specific routes. It later saw parts of 21Vianet-labelled space, including 103.253.204.0/22 and several 211.148 prefixes, briefly under AS18244 in 2021. It also saw 2403:a200:a100::/48, a slice of 21Vianet's IPv6 allocation, under AS18244 in early 2023. These observations show a real historical routing role and sustained technical proximity to networks now originated elsewhere.
The present is different. RIPE's AS overview says the ASN was unannounced at the 12 July 2026 observation point. Its announced-prefix response is empty because no current route clears the service's visibility threshold. RouteViews finds no current prefix originated by AS18244, CIDR Report calls it unannounced, and CAIDA marks the ASN unseen, with no measured provider, peer or customer degree.
There is one important wrinkle. RIPE's status retains a last-seen route for 211.148.28.0/24 under AS18244 in January 2026, at low visibility. Its neighbour view identifies AS17621, China Unicom's Shanghai network, as the one observed neighbour in the latest low-visibility sample. That prevents the stronger statement that the ASN has had no recent operating use. It supports a more precise description: AS18244 has no broadly visible current origin, but it has appeared intermittently and remains capable of being used.
An assigned ASN is an operational option. It can support a distinct routing policy if sessions, equipment, address authority and carrier agreements are in place. It is not by itself proof of an independent edge. APNIC explains that an organisation seeking an ASN should be multihomed with a distinct routing policy or able to meet that condition shortly after assignment, but the assignment criteria do not guarantee that those conditions remain visible years later. Customers need the current topology, not only the continued existence of the number.
One quarter of the largest block is live through China Unicom
The most substantial current routing fact is easy to state. Stateline's 120.64.0.0/14 contains four /16 blocks. At the observation point, only 120.65.0.0/16 appeared as a more-specific public route. It was originated by AS4837 and visible to all 325 reporting RIPE IPv4 peers. RouteViews independently saw the same /16 through China Unicom, with broad collector visibility.
The AS4837 registration identifies the origin as the China Unicom China169 backbone. That means 65,536 addresses inside Stateline's allocation have a current global path through a major national carrier. It is a stronger sign of active use than a registration alone. It is also only one quarter of the allocated block. RIPE found no route for the covering /14, and no current more-specific route for the other three quarters in the same status view.
The gap between allocated and routed space should not be interpreted as empty racks. Address planning does not map neatly onto hardware occupancy. A provider can have powered servers behind private addresses, conserve public IPv4, reserve ranges, or use a carrier's addresses. Conversely, a routed /16 can support customer access, access networks, enterprise lines or other services without proving any particular colocation footprint. The route gives reachability and an origin network; it does not give a building or workload inventory.
The origin also lacks a positive cryptographic statement in the observed validation view. RIPE's RPKI response for AS4837 and 120.65.0.0/16 classifies it as unknown because no validating route-origin authorisation was found. RouteViews expresses the same state as not-found. Unknown does not mean invalid, and it is not evidence of hijacking. It means that this public security mechanism does not provide a signed confirmation of the intended origin.
For a buyer, the practical questions are contractual. Does Stateline instruct China Unicom to announce the prefix, or does another party do so? Who can withdraw, deaggregate or filter it during an incident? What service-level commitment covers that action? Which physical handoff carries the route, and does a backup path exist outside the same carrier and duct? The public route proves that traffic can be directed to China Unicom. It does not prove that Stateline can keep customer services reachable after that path is removed.
The other live block is carried by 21Vianet
Stateline's 103.253.204.0/22 is also clearly active. RIPE sees the complete /22 originated by AS17428 and visible to 324 of 325 reporting IPv4 peers. RouteViews sees the same origin across its current collectors. AS17428 is registered as CHINA-ABITCOOL, 21Vianet Co., Ltd.
This route has a clearer origin-authorisation position than the China Unicom route. RIPE's validator finds a valid authorisation for AS17428, with more-specifics permitted down to /24. That makes the intended public origin less ambiguous. It still does not establish the commercial agreement behind the route or the physical location of any server using an address from the block.
The historical proximity is notable. AS18244's contact entity has long used a 21viamail.com address. Stateline's ASN has previously originated parts of address space now associated with AS17428. Current route records for several neighbouring 211.148 prefixes also identify 21Vianet as origin. These facts support the existence of an operating connection over time. They do not, without a corporate filing or agreement, prove that Stateline is owned by 21Vianet, that the two names are interchangeable, or that every Stateline resource terminates in a 21Vianet facility.
That boundary matters in an outage. If Stateline sells a service delivered from another operator's site, customers need to know whether Stateline is a reseller, network tenant, resource holder, subcontractor or historical name. They need to know which entity owes the uptime commitment and which entity controls the remedial action. A carrier-operated route can be highly reliable, but it adds an organisational handoff. The customer should see the escalation path across that handoff before an incident, not discover it while a prefix is unreachable.
The route is therefore both positive and limiting evidence. It shows that a Stateline-registered asset is in sustained public use and that route security is configured for the current origin. It does not show an independently operated Stateline edge, a Stateline-owned building, or failover from AS17428 to AS18244. It makes the case for asking a more exact question: what operational responsibility remains with the named Shanghai company today?
China Telecom's covering route does not identify the last mile
The 211.148 registrations present a third arrangement. Stateline is the named holder of 211.148.0.0/20 and 211.148.16.0/24. Yet RIPE's status for the /20 finds no exact route and instead sees it covered by 211.148.0.0/18, originated by AS4812. The same is true for the adjacent /24 at the observation point. AS4812 is China Telecom's Shanghai network.
A less-specific route means the global internet has a path toward the larger block. It does not prove that China Telecom's internal network has a working route to every address in the Stateline ranges. Nor does it identify the customer circuit, facility or router to which traffic would be delivered. The absence of an exact public route can be a normal part of aggregation, but it reduces what an outside observer can verify about that segment.
The /24 has a useful recent history. RIPE records it under AS4812 from 2005 and last saw that exact route in March 2026. It is now reached, if at all, through the larger aggregate. This change could be routine route consolidation. It could also reflect a service change. Without an operator explanation, neither interpretation should be presented as fact.
The three current origin surfaces together are revealing. China Unicom originates 120.65.0.0/16; 21Vianet originates 103.253.204.0/22; and China Telecom originates a larger aggregate that covers the 211.148 space. That is not the same as three-carrier resilience for one rack. Each address range follows its own route. A server numbered only from the China Unicom block does not automatically fail over to the 21Vianet block. DNS, security policy, customer allow-lists, certificates, session state and application configuration may all bind a service to one address.
Genuine network recovery requires a demonstrated design. It might use the same portable prefix through a second authorised origin, a controlled change of origin, provider-independent addressing across separate links, application-level distribution, or pre-provisioned alternative addresses. Whatever the method, the customer needs measured failover time and evidence that the physical paths do not share the same entrance, duct, meet-me cabinet or upstream aggregation point.
Three carrier names are not three independent failure domains
Carrier diversity is often reduced to a list of logos. The Stateline evidence shows why that is inadequate. Three major network names appear around the registered resources, but none of the public views establishes two simultaneous external paths for a single Stateline-originated prefix. AS18244 has no broadly visible current prefix and no public PeeringDB network entry. PeeringDB participation is voluntary, so absence there is not proof of no interconnection. It simply adds no independent evidence of facilities, exchanges, ports or an open peering policy.
Logical and physical diversity must be tested separately. Two carrier contracts can enter the same building through one duct. Separate ducts can converge in one street chamber. Different circuits can terminate on the same line card or depend on the same metropolitan aggregation site. A route may have two upstream names while both are ultimately carried over the same local access fibre. Conversely, a single carrier can sometimes deliver genuinely diverse paths through separate infrastructure. Names alone settle none of those questions.
The relevant evidence begins with site drawings. A customer should see the two building entrances, the path to separate meet-me rooms, the cross-connect identifiers, the routers and power feeds at each demarcation, and the maintenance responsibility at every handoff. It should see route tables during normal operation and during a controlled withdrawal. It should also know whether denial-of-service mitigation is applied by the origin carrier, by Stateline, or by another service, and how quickly filters can be changed outside business hours.
RFC 4271 explains how BGP exchanges reachability between autonomous systems, but BGP does not guarantee that the physical cable under two logical paths is independent. RFC 7454 describes operational safeguards such as filtering and maximum-prefix controls. Those practices are relevant to a resilient edge, but no public evidence shows Stateline's current filter policy, router pair, session protection or change discipline.
The public route pattern should therefore be treated as a dependency map, not a redundancy certificate. It proves that Stateline-labelled resources can be reached through major carriers. It also shows that present reachability depends on origins other than AS18244. Until a site-level topology and failover test are disclosed, the customer should assume that each live prefix has a single observable public origin and price the service accordingly.
Installed, usable and recoverable capacity are different numbers
Data-centre capacity is often expressed as racks, megawatts or floor area. Each number can be technically true while overstating what a new customer can safely use. A building can have room for cabinets before utility power is energised. A transformer can be installed before final acceptance. A hall can be commissioned while network cross-connects are pending. A rack can be powered but unsuitable for the density of the customer's equipment. A sales allocation can consume the spare power needed for recovery from another hall.
Stateline's public footprint contains no current rack count, megawatt figure, occupancy rate, commissioned hall list or service catalogue that can be tied to the exact Shanghai legal name. The absence should not be converted into a zero. It means that any marketed figure requires a site schedule and a definition. The useful categories are planned, constructed, energised, commissioned, occupied, technically free, contractually available and recoverable capacity.
Recoverable capacity is the hardest and most valuable category. Suppose two halls each operate at 70 per cent of their power limit. They may appear to have 60 per cent of one hall's combined load free across the pair, but they cannot necessarily absorb the loss of either hall if cooling zones, network ports or rack density are mismatched. A spare cabinet without the right power whips, cross-connects, security zone or customer hardware is not immediate recovery capacity. A backup site without current data is not a recovery site.
Shanghai's policy gives a useful local reference point by defining a standard rack as one with 6 kW of IT load. The same 2022 implementation opinion aimed for about 280,000 standard racks citywide by 2025 and an average installation rate above 85 per cent. Those citywide ambitions say nothing about Stateline's inventory, but they illustrate why rack counts need power assumptions. A nominal 100-rack room at 3 kW per rack is a different asset from one commissioned for 6 kW or higher density.
A credible offer would state the contracted IT load, the maximum rack density, the sustained cooling capacity at design conditions, the committed cross-connects, and the reserve kept out of sale for failure recovery. It would identify any customer concentration on one power train or cooling zone. Without those facts, an address portfolio can make an operator look larger than the physical and recoverable capacity available to the buyer.
Power access is a permitting question before it is a redundancy claim
In Shanghai, power is not an invisible commodity that appears when a room is fitted with racks. The city's 2022 data-centre policy explicitly coordinates scale, location and energy use. It says compliant projects need the relevant approvals, energy review, acceptance and machine-room safety assessment before network or power access is supplied. Power companies are expected to report connection demand from data centres and other high-energy users for review.
The city's thresholds have also tightened. The Shanghai carbon-peaking plan set a PUE ceiling of 1.3 for new data centres and an ambition of no more than 1.4 after renovation of existing sites. The 2024-2027 green transition plan lowered the stated control target for new facilities to below 1.25. These are not performance figures for Stateline. They show the regulatory environment in which any new or expanded Shanghai capacity has to be justified and operated.
For a customer, the first power question is therefore not whether a brochure says dual feed. It is which utility connections are approved and energised for the named building, at what capacity, and under which operator. Two cables from one substation can fail together. Two substations can still share a higher-voltage dependency. A second feed that cannot carry the full critical load during maintenance is not equivalent to a fully redundant supply.
The proof is an electrical single-line diagram reconciled with commissioning records. It should identify utility sources, transformers, switchboards, automatic transfer logic, UPS modules, battery autonomy, generator sets and the loads excluded during emergency operation. It should show the largest failure that the system can tolerate while maintenance is under way. It should also state whether cooling pumps, controls, carrier rooms, security systems and fuel-transfer equipment remain powered, because keeping servers energised while heat rejection or access control fails does not preserve service.
Stateline's address records do not answer any of this. They cannot. The gap does not prove that adequate electrical systems are absent. It means the marketed capacity cannot be evaluated until power evidence is attached to the exact facility and contracting entity.
Generator runtime is a chain of fuel, controls and cooling
Backup generation is often presented as a runtime number, but runtime is conditional. A generator may have enough fuel for several hours at one load and much less at another. The usable fuel can differ from tank nameplate capacity. Refuelling depends on road access, supplier priority, pumps, staff and permission to operate during an extended emergency. A generator that has passed a no-load start is not proven under a sustained facility load.
The transfer sequence creates its own failure path. Utility loss must be detected; UPS batteries must bridge the gap; generators must start and stabilise; switchgear must transfer; parallel units must share load; and cooling equipment must return without unstable controls. A weak battery string, failed starter, closed fuel valve, control error or protection setting can defeat a design that looks redundant on paper. Common maintenance can also expose both sides of an A/B system at once.
The buyer should ask for recent integrated systems tests, not only equipment certificates. The evidence should include a utility-loss simulation, full-load or load-bank results, battery condition, actual transfer times, generator fuel burn, on-site fuel quantity and refuelling agreements. It should include tests during maintenance states, because an N+1 label says little if the spare is unavailable when another unit is being serviced.
Power restoration is only useful if cooling survives. Chillers, pumps, cooling towers, computer-room air units and controls can draw large starting currents. Some sites shed part of the cooling load during generator operation. That may be acceptable if thermal conditions stay within limits and service can continue, but it must be measured at the installed IT density. High-density racks consume the thermal buffer more quickly than lightly loaded rooms.
The relevance is not theoretical. Uptime Institute's 2024 global survey reported on-site power distribution as the leading cause of impactful outages, accounting for 54 per cent in its respondent sample. That global finding cannot predict Stateline's performance. It explains why evidence of actual transfer and sustained operation deserves more weight than a component list.
Cooling headroom determines whether electrical capacity is sellable
Every watt consumed by computing equipment becomes heat that must be removed. A facility can have spare breaker capacity and still be unable to accept another rack because the cooling path is saturated. The relevant limit may sit in a chiller, pump, heat exchanger, cooling tower, air path, containment arrangement or local rack inlet. It may appear only during humid summer conditions or while one unit is out for maintenance.
PUE is useful but incomplete. Shanghai defines it as total data-centre electricity divided by IT-equipment electricity. A lower figure generally indicates less overhead for power conversion and cooling. It does not by itself establish reliability. An efficient design can have a shared component that creates a common failure, while a less efficient design can have more reserve. The customer needs both measured annual efficiency and the capacity of each cooling fault domain.
The public Stateline evidence provides neither a measured PUE nor a cooling design. No exact hall is identified, so even local weather and water dependencies cannot be assigned. A buyer should therefore ask for the cooling topology, rated and measured capacity, maintenance configuration, maximum inlet temperatures, leak detection, water storage where relevant, and alarms. It should see trend data during peak ambient conditions and during the loss of the largest cooling component.
Increasing rack density changes the answer. A hall designed around modest loads may host a few high-density cabinets only with rear-door heat exchangers, direct liquid cooling or reduced capacity elsewhere. A quoted rack count can remain unchanged while the number of sellable racks falls. Conversely, a well-designed lower-density room can have substantial operational reserve. The marketing unit must be linked to kilowatts and heat rejection.
Shanghai encourages liquid cooling, high-voltage direct current, integrated energy monitoring and other efficient systems in new or upgraded facilities. Those policy directions show what the city wants operators to improve; they are not evidence that Stateline has installed any of them. The correct test is site-specific: what is operating now, what has passed acceptance, what load has been sustained, and how much reserve remains after one cooling component is unavailable?
Fire and water can cross the boundaries that diagrams keep separate
A resilient data hall is more than duplicated electrical equipment. Fire can disable both power paths if cables share a room or riser. Water can enter below raised floors, damage switchgear or trigger an emergency shutdown even when servers remain dry. Smoke, suppression discharge and safety procedures can make a room inaccessible. A local event can therefore outlast the technical fault that started it.
Shanghai's policy explicitly calls on data-centre operators to strengthen protection against fire, lightning, flood and earthquakes, improve power and cooling availability, maintain emergency plans and conduct exercises. A draft national standard for green public-sector data centres similarly calls for smoke and heat detection, very-early warning in critical areas, independent UPS support for fire systems, water-leak detection where flooding is possible and drainage measures. These are useful expectations, not proof of implementation at a Stateline site.
The undisclosed facility address is especially consequential here. Flood risk varies by site elevation, drainage, basement use and the location of utility equipment. Fire compartmentation depends on the actual building. A carrier room in one fire zone may be lost even when the customer hall is protected. Fuel tanks, battery rooms and loading areas create different hazards. None can be assessed from a room number in a commercial office tower.
Evidence should include fire-compartment drawings, detection and suppression test records, battery chemistry and separation, water-leak sensors, drainage paths, placement of switchgear above flood exposure, and the shutdown authority held by staff and emergency responders. It should also show where the two power paths and fibre paths share spaces. A design is only as independent as the point at which the duplicated systems converge.
Recovery after fire or water damage is also an inventory question. Replacement switchgear, UPS modules and chillers may have long lead times. A spare server cannot repair a destroyed distribution board. Customers need to know whether capacity exists in another commissioned fault domain and whether their data, network policy and access rights can move there. A statement that the operator has another site is incomplete without the migration path and tested recovery time.
Permitting and construction delay can strand announced capacity
New capacity passes through several states before a customer can safely occupy it: land and building approval, energy review, utility connection, construction, equipment installation, commissioning, acceptance, safety assessment, carrier delivery and customer fit-out. Delays at any stage can leave an announced rack unavailable. The distance between a planned megawatt and an energised, customer-ready megawatt can be measured in months or years.
Shanghai's rules are unusually direct about this sequence. Projects receiving energy support were expected to complete approvals, begin construction and enter operation within a stated period, with support at risk if they did not. Operators with projects that failed energy acceptance could be excluded from new project applications. The city also required adherence to commitments on location, design, function, energy measures, operator and ownership after commissioning.
These provisions make legal and operator identity part of capacity. If Stateline markets space in a building whose approvals, utility contract or safety assessment name another operator, the customer should know the arrangement. Subcontracting is not inherently weak, but the responsibility for delay, acceptance and remedial work must be clear. A letter of intent for power is not an energised feed. A carrier feasibility study is not a delivered cross-connect.
The public evidence identifies no current Stateline project, construction site or expansion announcement that can be tested against those milestones. It would be wrong to invent one. The practical response is to demand a readiness schedule for any offered space: exact building, hall, rack, contracted IT load, approval references, utility energisation, commissioning date, carrier order status and acceptance certificate. Deposits and migration plans should be tied to those milestones.
This is where a thin public footprint should change commercial terms. It does not automatically disqualify the provider. It just shifts the burden from brand recognition to documentary and physical verification. Capacity should be counted only when the customer can identify the room, power train, cooling zone, network handoff and responsible operator, and when those elements have been tested together.
Who is affected when the chain fails
The impact of a data-centre failure depends on what the customer placed in the room and what was kept elsewhere. A power or cooling event can stop web services, enterprise applications, payment systems, communications, monitoring and backups at once. A carrier failure can leave healthy servers inaccessible. A route error can isolate only the public addresses while private links continue. A support failure can extend a minor hardware fault because no authorised person reaches the rack.
Colocation customers bear risks that are easy to overlook. They may own the servers but rely on the operator for power, cooling, security, remote hands and cross-connect access. If the operator boundary is unclear, each supplier can wait for another to act. A customer using Stateline-registered addresses but housed in a third party's building may need one organisation to diagnose the rack, another to repair the circuit and a carrier to change the route.
End users experience the combined result, not the contractual divisions. A dual-powered server is still offline if both feeds come from failed upstream switchgear. A healthy storage system is unavailable if the only public prefix disappears. A second carrier is useless if DNS still points to the first address and no tested update path exists. An off-site backup helps only if it is current, reachable, restorable and supported by enough compute and network capacity.
The customer should define recovery at the service level. How long can the service tolerate loss of the hall, the route, the carrier meet-me room or the support team? Which applications restart first? What data can be lost? Which external dependencies must be changed? Who has credentials and authority? These answers determine whether facility redundancy is sufficient or whether the customer needs independent deployment across two operators or regions.
The thin evidence around Stateline raises the importance of that customer-owned design. The company may be capable of delivering reliable service through established carriers and wholesale facilities. Public routes alone cannot prove it. Buyers should avoid placing all recovery responsibility in an unverified promise. They should retain portable configurations, tested backups, current contact paths and an exit plan that does not depend on the failed site.
The evidence that would convert address control into operating confidence
Stateline can answer the central question without publishing sensitive floor plans. First, it can identify every facility in which it offers service, the legal operator at each site and the role Stateline performs. It can state whether it owns, leases or resells the capacity. It can tie each address block and origin ASN to the relevant building or service without exposing customer assignments.
Second, it can publish bounded capacity figures. For each site, the useful figures are commissioned IT load, occupied load, contractually available load and reserve retained for recovery. Rack counts should include the assumed kilowatts. Expansion figures should be separated into under construction, energised and accepted. A dated third-party assurance or customer audit can support the claims without revealing security-sensitive details.
Third, it can disclose the resilience design at the level buyers need: number and source of utility feeds, UPS topology, generator configuration, tested runtime, fuel arrangements, cooling fault domains, fire compartments, flood controls, fibre entrances, meet-me rooms, current origins and backup routing method. The evidence should include dates of the last integrated power test, cooling failover, route withdrawal and customer recovery exercise.
Fourth, it can explain AS18244. If the ASN is retained for contingency, migration or closed networks, that role can be stated. If China Unicom, China Telecom and 21Vianet originate Stateline resources under contract, the control and escalation path can be described. If the resources serve another operator, the relationship can be identified without collapsing the two legal entities. A current route policy and valid origin authorisations would make intended origins easier to verify.
Finally, it can show how customers leave. The service should define address portability, equipment removal, data return, cross-connect termination and assistance during operator failure. Recovery evidence is strongest when it includes a successful restoration or relocation, not only a promise that spare capacity exists.
None of these disclosures requires Stateline to reveal customer names, exact rack coordinates or exploitable security details. They require the company to connect its visible administrative assets to a current physical and operational surface. That connection is what the public evidence now lacks.
Verdict: valuable network assets, unverified data-centre capacity
Stateline Internet Data Center Co., Ltd. Shanghai has more substance than a dormant name. It controls long-standing portable internet resources. A full /16 inside its largest allocation is globally routed through China Unicom. Its 103.253.204.0/22 is broadly visible through 21Vianet with a valid route-origin authorisation. Its 211.148 space sits under a China Telecom aggregate. Its own AS18244 has a real history and a brief recent appearance, even though it is not a broadly visible current origin.
Those facts support a narrow positive conclusion: Stateline-labelled network assets remain connected to live infrastructure. They do not establish how much data-centre capacity the Shanghai company operates, where that capacity sits, who owns the power and cooling systems, or whether one customer workload can survive loss of a utility feed, carrier path, meet-me room or hall. The central office address cannot substitute for a facility address, and three origin networks across different prefixes cannot substitute for tested failover.
Shanghai's regulatory setting raises the standard of proof. Power access, network access, energy acceptance, machine-room safety, PUE, emergency planning and protection against fire and flood are linked parts of an operating facility. A credible provider should be able to show that the exact room offered to a customer has passed those tests and retains enough reserve to recover during maintenance and failure.
Until Stateline publishes or supplies that evidence, its operating status should be treated as constrained and partially verified. The address portfolio is real. Two parts of it are strongly reachable through other networks. The independent edge, physical site, commissioned capacity and customer recovery path remain unverified. For buyers, the right response is not to assume that nothing operates. It is to count only the capacity that can be located, powered, cooled, routed and recovered under a named operator with dated test evidence.