Summary

  • AS24910 is still an assigned internet number, but it is not an operating public network. RIPE's current AS overview marks it unannounced, and the routing-status observation shows no IPv4 or IPv6 routes, no visible address space and no collector peers seeing it on 10 July 2026. The last route attributed to AS24910 was seen in December 2019.
  • The identity is institutional and London-specific, not regional-ISP evidence. The RIPE registration calls the network HIGHBRIDGELDN and describes it as HIGHBRIDGE LONDON, while the linked organisation record names JPMorgan Chase & Co. Highbridge's current firm page describes an alternative investment manager headquartered in New York with a research presence in London; it does not offer broadband, fibre access or fixed-wireless service.
  • The clearest London block, 91.213.81.0/24, remains registered as HCM-LDN-NET but no longer follows AS24910. RIPEstat's prefix view shows Amazon's AS14618 as the current origin, and routing history places the handoff after AS24910 disappeared. Amazon's own documentation explains how customer-owned addresses can be advertised through AWS, but the route table does not identify the workload or prove that this particular block uses any one AWS product.
  • Corporate and network status changed on different clocks. Companies House says Highbridge Capital Management (UK), Ltd. was dissolved in September 2024 after a members' voluntary winding-up, yet Highbridge still lists the same Tudor Street location as its London research presence. Neither the dissolved company nor AS24910 should be used as evidence that Highbridge's London activity ended.
  • The final network evidence grade is Negative for the assigned regional-ISP thesis and Medium for a historical enterprise-edge analysis. Public evidence supports a former London routing edge, one observed logical upstream near the end of its life, a later Amazon-originated /24 and continuing institutional activity. It does not establish retail customers, access plant, towers, poles, subscriber equipment, field-repair crews, physical route diversity, site power design, application placement or current dependence on AS24910.

The category error is the infrastructure finding

HIGHBRIDGELDN looks like a network brand until its parts are read literally. HIGHBRIDGE identifies an investment firm. LDN identifies London. The autonomous system registration supplies no service catalogue, coverage area or customer proposition. RIPE's aut-num record for AS24910 contains the short description HIGHBRIDGE LONDON, links the number to JPMorgan Chase & Co's RIPE organisation and records an assignment created in May 2002. It does not describe a neighbourhood broadband network.

The company evidence points in the same direction. Highbridge's current overview says the firm was founded in 1992 and offers credit- and volatility-focused investment strategies to sophisticated investors. It places headquarters in New York and a research presence at 5 Tudor Street in London. The firm's website terms identify Highbridge Capital Management, LLC as the operator of the site. Neither page markets internet access, customer installations, speed tiers, a serviceability checker, fibre construction, radio coverage or telecommunications support.

That absence is not a claim that no third party could ever have bought a communications service from a related JPMorgan company. It is a boundary on what AS24910 can support. An autonomous system is a routing identity: a network under one routing policy can announce addresses to other networks. Banks, asset managers, universities, manufacturers and retailers hold AS numbers because they need control over enterprise connectivity. The number does not turn them into broadband carriers any more than a company-owned switchboard turns them into a public telephone operator.

The difference changes the physical analysis. A regional ISP usually owns or leases an access network that reaches outside customers through poles, ducts, towers, street cabinets, building entries and customer-premises equipment. Its failure population is the homes and businesses attached to that plant. An enterprise edge sits at the boundary between an institution's systems and one or more carriers. Its physical chain may include office or data-centre routers, managed circuits, cross-connects, firewalls, local power and staff with authority to change routes.

The directly affected users are people and systems that rely on those paths, not subscribers in a public service territory.

AS24910 belongs in the second category. Near the end of its visible life, route collectors saw a single adjacent network, Equinix's London managed-IP service. The route history does not expose a residential access layer behind it. The current public state is even clearer: there is no AS24910 route to buy, use or repair. A current broadband bill cannot depend on an autonomous system that is not announcing any public address space.

The appropriate category is therefore institutional, and the appropriate region for this network-specific report is the UK. JPMorgan Chase is a US company with a global footprint, but the AS name, description, London address block and former upstream context all point to a London edge. A US country value in the linked RIPE organisation record identifies the resource holder's registration context. It is not a service map.

Assigned does not mean alive

AS24910 illustrates one of the internet registry's most important distinctions. A number can remain assigned after its routes disappear. The registration preserves who may administer the resource; the global route table shows whether the resource is currently being used as an origin visible to public collectors. Those are related but different states.

At the observation point on 10 July 2026, RIPEstat's AS overview labels AS24910 unannounced. The more detailed routing-status result reports zero IPv4 prefixes, zero IPv4 addresses, zero IPv6 prefixes and no RIPE RIS peer seeing the AS in either address family. The announced-prefix result is an empty list, as is the current neighbour result.

That is strong negative evidence of public operation. It does not prove every router once configured with AS24910 has been destroyed or that the number does not appear in a private laboratory. It does mean the AS was not functioning as a globally visible origin at the measured time. An outside network had no AS24910-originated route through which to reach a public prefix.

The current registration is nevertheless maintained. The AS record was last modified in October 2022 and uses a JPMorgan maintainer. Its linked JPMorgan RIPE organisation was modified in May 2026. That recent organisation-level change is evidence that JPMorgan's broader RIPE presence remains administered. It is not evidence that AS24910 returned to service. One maintained parent record can cover many live and retired resources.

The routing-consistency view makes the residue visible. It finds three prefixes in internet routing registries that are not in BGP under AS24910: 91.213.81.0/24 in RIPE, 198.235.60.0/23 in ALTDB and 203.193.19.0/24 in RADB. None is an active AS24910 announcement. Historic route or policy entries can outlive the routing configuration they once described.

This distinction also prevents a capacity error. AS24910's status cannot be measured by counting entries attached to its registration. Installed addressing, registered policy and usable public reach are not the same. For the present AS24910 origin, usable public routing capacity is zero observed prefixes. The wider institution may have enormous connectivity elsewhere, and the London office may be fully connected through provider-assigned addresses, another JPMorgan AS, secure access services or cloud infrastructure. None of those possibilities revives AS24910.

A route history spanning several institutional blocks

The network was once real and long-lived. RIPEstat's routing-status history marker first sees AS24910 originating 199.245.114.0/23 in May 2003 and last sees it with 91.213.81.0/24 in December 2019. The full AS routing history contains six observed route forms across that period: 199.245.114.0/23, the more-specific 199.245.114.0/24, 203.193.19.0/24, 204.86.14.0/24, 206.17.152.0/23 and 91.213.81.0/24.

They were not six simultaneous London access networks. The timelines overlap in different combinations, and the address registrations carry different legal and geographic histories. ARIN's record for 199.245.114.0/23 calls it a direct allocation to Arc Four dating from 1994. The current record does not establish why AS24910 originated it, what services used it or whether Arc Four had a corporate connection to Highbridge at the time. BGP tells us who announced a route, not who owned every machine behind it.

The other North American records are more directly institutional. ARIN's 204.86.14.0 allocation names JPMorgan Chase & Co. and covers 204.86.14.0 through 204.86.15.255. RIPEstat observed AS24910 originating the first /24 from 2011 into 2015. ARIN's 206.17.152.0/23 record is a carrier assignment to Highbridge Capital Management, Inc. at a New York address, while RIPEstat observed AS24910 originating it from 2006 into 2016.

The Asian route is the clearest warning against reading today's registration backwards. The current APNIC record covering 203.193.19.0/24 was registered in 2020 and carries a Hong Kong country code with contacts unrelated to Highbridge. AS24910's route history for that /24 ends in 2007. The current registration therefore cannot tell us who controlled the range during the older announcement. Historical BGP and present-day address records answer different questions.

Taken together, these routes make AS24910 look more like an institutional routing policy used for selected Highbridge or JPMorgan-connected blocks than a local access provider. The network could originate a London-specific block and, at different times, address space associated with New York or wider corporate allocations. That pattern is consistent with an enterprise edge or shared routing arrangement. It is not proof of one physical wide-area design, one operations team or one building.

The history also resists a simple capacity total. A /23 contains 512 IPv4 addresses and a /24 contains 256, but overlapping routes cannot be added twice. More importantly, address count is not throughput. A /24 behind a 10Gbps handoff can have more bandwidth than several larger blocks behind a constrained circuit; a large allocation can contain mostly unused addresses; a route can be visible while every application behind it is unavailable. Public routing history establishes reachability and origin, not port speed, traffic, utilisation or service quality.

The London /24 is the continuity clue

The strongest link between the AS name and London is 91.213.81.0/24. RIPE's address record names the block HCM-LDN-NET, assigns country code GB, links it to JPMorgan's RIPE organisation and gives it provider-independent status. It was created in July 2009, shortly before the route appeared under AS24910. The associated RIPE route object says HIGHBRIDGE LONDON and names AS24910 as origin.

Provider-independent space matters because it can survive a carrier change. A business that uses addresses assigned by an access provider normally has to renumber when it leaves that provider. A holder of portable space can authorize another network to announce the same range, subject to registry and routing controls. That continuity can preserve public endpoints, firewall allowlists and counterparties' configurations while the underlying infrastructure changes.

The public record shows exactly that kind of origin change, although not its business purpose. RIPEstat's prefix routing history around the transition sees AS24910 through 12 December 2019. It then sees Amazon's AS14618 originating the exact /24 from 21 January 2020. The current prefix overview still names AS14618 in July 2026, and the current routing status reports the route visible to all 327 listed IPv4 RIS peers.

The route-policy layer has not become equally tidy. RIPEstat's prefix consistency result finds the live Amazon AS14618 route, a registered but not observed Amazon AS16509 route and the old registered but not observed AS24910 route. The original RIPE route object was not removed. There is also no validating route-origin authorization in RIPEstat's RPKI check for AS14618 and the /24, which returns unknown rather than valid or invalid.

That does not make the current route illegitimate. Unknown means the public validator found no applicable cryptographic authorization; it does not mean the origin failed one. Amazon can also verify address control through mechanisms beyond what a current RPKI result reveals. The important operational point is narrower: route visibility, registry allocation and policy authorization are three layers, and their records do not all update at the same time.

The /24 is therefore evidence of address continuity across an infrastructure transition. It is not proof that the old London office edge was lifted intact into the cloud. The same public addresses could support web endpoints, remote-access gateways, security services, email infrastructure, market-facing applications or systems not readily classified from outside. They could be used in one AWS region, through a global service or only as stable public identities. The route table does not answer which.

Amazon origin is a strong signal with a strict limit

Amazon documents a mechanism that fits the visible routing pattern. Its EC2 bring-your-own-IP guidance says an organization can bring a publicly routable address range to AWS, continue to control the range and have AWS advertise it. The minimum IPv4 range is a /24, exactly the size of HCM-LDN-NET. Amazon's authorization guidance identifies AS16509 and AS14618 as Amazon origins that customers may authorize.

Its advertising procedure advises a customer to stop announcing the range from the old location before AWS announces it, because simultaneous announcements can produce unpredictable routing. The 91.213.81.0/24 history shows a withdrawal under AS24910, a gap and then an AS14618 origin. That sequence is consistent with a controlled bring-your-own-address transition.

Consistency is not confirmation. No public AWS account record links the /24 to Highbridge, no service configuration identifies the application, and the current RIPE allocation links the space to JPMorgan Chase rather than exposing which affiliate or team uses it. Amazon also offers more than one service capable of advertising customer addresses. The evidence cannot distinguish an EC2 address pool from another AWS architecture simply by looking at the origin AS.

The safe conclusion has three parts. First, control of the public route moved away from AS24910. Second, Amazon's network now supplies global reachability for the /24. Third, the registered address holder retained a continuity interest strong enough that the range remained identifiable as JPMorgan-linked space instead of being replaced with ordinary Amazon addresses. The application and operating team remain unknown.

That shift changes the likely failure surface. Under the old origin, public reach depended on AS24910's routing equipment and upstream path. Under the current origin, public route propagation depends on Amazon, while the application depends on whichever AWS service, region, load-balancing design, security policy and back-end system the holder selected. Users still depend on their own local access and on the internet between them and Amazon. The route can remain visible while an application fails, and an application can remain healthy in one location while a user loses local connectivity.

The address continuity can hide this architectural change from counterparties. A bank, exchange, investor or service provider may keep the same address allowlist while the physical path behind it changes completely. That is often the point of portable addressing, but it creates an assurance obligation: operational owners need an accurate record of which team controls the address, which team controls the cloud service, which security controls bind it and which party can withdraw or reroute it during an incident.

The company timeline and the network timeline diverge

Highbridge's corporate history explains why the RIPE organisation names JPMorgan Chase. JPMorgan's 2004 annual report says it formed a strategic partnership with Highbridge and acquired a majority interest in December 2004. It described Highbridge as a New York-based hedge-fund manager with offices in New York, London and Hong Kong. JPMorgan's 2007 annual report records an additional purchase that took its interest to 77.5% in January 2008. The 2009 shareholder letter says JPMorgan assumed 100% ownership in mid-2009.

The relationship remains current. A 2025 SEC prospectus describes Highbridge as an indirect subsidiary of JPMorgan Chase, and Highbridge's own page says its J.P. Morgan partnership began in 2004. A 2026 SEC filing shows Highbridge Capital Management, LLC continuing to act as an investment adviser. The active firm is not the retired AS and should not be described as defunct merely because its old routing identity is dormant.

The UK company has a separate clock. Companies House records Highbridge Capital Management (UK), Ltd. as a fund-management company incorporated in November 1997 and dissolved on 3 September 2024. Its filing history shows a confirmation statement in November 2020, a declaration of solvency and special resolution to wind up in March 2021, liquidation reports and the final dissolution. The official Gazette notice identifies a members' voluntary liquidation and lists 5 Tudor Street as the principal trading address.

A members' voluntary liquidation is not evidence of an insolvency-driven network collapse. The declaration of solvency points in the opposite direction. Nor did the legal timetable begin at the exact moment the route changed. AS24910's final London announcement disappeared in December 2019; director changes and accounting-period filings followed in 2020; the winding-up resolution came in March 2021; dissolution came in 2024. The sequence may reflect a broader reorganization, but the public documents do not state a causal connection.

The current Highbridge page still lists 5 Tudor Street as its London research location. That is the same address the Gazette identified as the dissolved UK company's principal trading address. Continued activity at an address can be carried by another legal entity, a branch, an affiliate, shared premises or a different operating arrangement. The public site does not specify which UK legal person employs the London researchers or signs their telecommunications contracts.

This boundary is consequential during an outage. The company that holds an ASN registration may not own the router. The company named on a lease may not contract with the carrier. A parent technology group may control route policy while a local facilities team controls power and building access. A cloud team may control an application while a separate security team controls the portable addresses. Corporate names can remain familiar while authority moves.

What the old physical edge probably required

Public evidence does not identify an AS24910 router address, rack, carrier cross-connect or power circuit. It nevertheless supports a bounded reconstruction of the components required for the old route to work. A public /24 originated from an enterprise AS needs routing equipment or a managed routing service that presents the origin to an upstream. It needs a path from the office or application environment to that edge. It needs power, configuration, security policy, address management and people authorized to make changes.

Near the end of 2019, RIPEstat's historical AS-neighbour observation saw only AS21371 immediately to the left of AS24910. RIPE's record for AS21371 describes Equinix Connect in the London metro and says that network had been integrated into Equinix AS15830. That supports a logical handoff to an Equinix managed-IP network. It does not prove that Highbridge's equipment was in an Equinix data centre; managed connectivity could extend to customer premises.

There are at least three plausible physical arrangements. The border router could have been in the Tudor Street office, with a carrier tail to an Equinix aggregation point. It could have been in a colocation facility, with a private circuit back to the office. Or the routing function could have been substantially managed by the provider, with Highbridge controlling policy and addresses but less of the physical edge. All three fit the public BGP observation. No public evidence selects one.

Each arrangement has a different failure chain. An office-based edge is exposed to building power, local cabling, office access restrictions and the path between the building and carrier network. A colocated edge adds the facility's power and cross-connect but may separate public routing from an office outage. A provider-managed edge shifts more restoration work to the carrier but can make service status and change authority dependent on the contract and escalation path.

The London location itself does not reveal route diversity. Two circuits entering the same building can share the same street duct. Two ports can terminate on one provider chassis. A managed service can use a resilient core while the customer tail remains singular. Conversely, one visible upstream AS can operate multiple routers and physically diverse paths. Public BGP collapses those details into an AS path; it cannot audit ducts, entrances, optical equipment or power feeds.

There is likewise no basis for putting poles, towers or household equipment into the picture. A central London institutional office is more likely to use commercial building risers, ducts, carrier fibre, Ethernet handoffs and enterprise routers than rural towers or residential drops. Even that is a reasoned expectation, not a disclosed asset list. The only subject-specific visual and analytical claim the evidence safely supports is a compact London enterprise edge dependent on a managed upstream and authorized institutional support.

One observed upstream is not a resilience design

The last observable topology was logically concentrated. The historical neighbour result counts one unique adjacent AS, AS21371, and no right-side neighbour. The same endpoint shows that neighbour in June and December 2019 before the route disappears. There is no evidence in that period of a second directly visible transit AS.

That supports a single-upstream conclusion at the AS-path level, but not a single-cable conclusion. Equinix could have delivered multiple circuits, multiple devices or a managed service with internal resilience while still presenting one AS number. It could also have delivered one customer tail whose failure removed all public reach. Contract, circuit and facility records would be needed to tell the difference.

The distinction between upstream diversity and physical independence matters because enterprises sometimes buy two services that converge below the level visible in commercial descriptions. Two access circuits may share a local exchange. Two carrier names may lease the same last-mile duct. Two BGP sessions may run over one optical bearer. A second router may share one power distribution unit. Route diversity, facility diversity and power diversity are separate properties and need separate evidence.

The public history provides no failover test, outage report, service-level commitment, busy-hour measurement or spare-capacity figure. The visibility of the /24 through AS21371 shows that the route worked from many collectors; it does not show how much traffic the handoff could carry or whether a backup could absorb a full load. The route's disappearance in 2019 appears to be a retirement because the prefix later returns through Amazon, but no public notice identifies a failure or emergency restoration.

For the old edge, a serious resilience review would have asked whether the London office and public routing site were the same place; whether a second carrier used a different building entrance; whether edge routers, firewalls and switches had independent power; whether configuration backups were current; whether the /24 could be announced elsewhere; and who could authorize an emergency change outside market hours. None can be answered from the surviving route record.

For the current prefix, the questions change. Amazon supplies the origin, so AS24910 upstream diversity is irrelevant to that route. Resilience now depends on the selected AWS service and deployment: regional or global design, availability-zone separation, application replication, health checks, security policy, capacity reservations and the path from users or partner institutions to Amazon. The public origin alone confirms none of them.

Installed address space is not usable service capacity

Infrastructure reporting often turns visible numbers into capacity claims too quickly. AS24910 is an especially good corrective because nearly every available number requires a different interpretation.

The ASN itself is one identifier, not one network site. The London /24 contains 256 IPv4 addresses, not 256 active servers. The historic /23s contain 512 addresses each, but some belonged to provider or corporate allocations whose relation to the AS changed over time. Six route forms in history do not mean six independent connections, because they appeared in different periods and one /24 was more specific than a covering /23. Zero current prefixes means no current public origin under AS24910, not zero current connectivity for Highbridge.

Port and traffic capacity are almost completely absent. There is no public AS24910 interconnection record stating port speeds, traffic range or facility presence. BGP visibility shows control-plane propagation, not the number of gigabits available. A prefix seen by every listed collector can still sit behind a congested handoff. A lightly visible prefix may carry little traffic or be intentionally restricted. Without interface counters, contract capacity, flow records or service monitoring, throughput claims would be invented.

Power capacity is equally opaque. A former office edge might have had building generator support, a local UPS or only short battery runtime. A colocation edge might have benefited from facility redundancy while its office backhaul remained exposed. Amazon's current origin moves the public announcement into a large provider network, but it says nothing about the back-end application's power architecture or whether dependencies outside AWS have comparable protection.

Human capacity cannot be counted from a corporate workforce figure. Highbridge serves sophisticated investors and JPMorgan operates at global scale, but neither fact identifies the number of people who supported AS24910. Enterprise network operations may be centralized across a parent company, contracted to a carrier, shared with a facilities provider or divided among security, cloud and office teams. The historical RIPE contacts and maintainers show administrative authority, not shift coverage, repair skills or response time.

The only defensible capacity statement is layered. AS24910 currently provides no observed public routing capacity. The 91.213.81.0/24 block remains globally reachable through Amazon AS14618. The application capacity behind that route is unknown. Highbridge's current institutional business remains active, but its dependence on that /24 is unknown. Separating those statements is more informative than adding unrelated numbers into a false total.

Who would have felt an outage

No subscriber base can be derived from AS24910. When the network was active, the directly exposed population would have been the users and systems relying on addresses or paths originated by the AS. The London naming makes office staff and London-facing institutional systems plausible, while the historical North American blocks show that the routing policy was not necessarily confined to one floor or one city. Public routes do not identify users.

An office access failure could have affected staff without withdrawing the public /24 if the public edge sat elsewhere. A border or upstream failure could have withdrawn the /24 while staff retained other internet access. A firewall error could have made applications unavailable while the route remained globally visible. A DNS or identity failure could have produced a user-visible outage with no BGP change. These are different failure paths, even if all are reported colloquially as "the network being down."

For an alternative investment manager, timing can magnify impact. Researchers, portfolio managers, risk teams and operations staff depend on market data, communications, authentication and counterparties. Yet it would be wrong to claim that AS24910 carried trading, investor or risk systems merely because Highbridge performs those functions. The current SEC adviser record establishes the business category; it does not map systems to the old AS.

The current Amazon-originated /24 creates a different possible affected set. Stable public addresses are often retained for endpoints that counterparties allowlist or for services whose address identity should not change during migration. Amazon's BYOIP guidance explicitly supports continued control of customer-owned space. But without DNS, application or account evidence tying Highbridge systems to the block, even that common use should remain a hypothesis.

What can be said is that an AS24910 outage is no longer a current scenario. There is no public AS24910 route to fail. A failure of 91.213.81.0/24 today would involve the Amazon-originated path or the service behind it. A London office connectivity failure today could involve completely different addresses and carriers. The historical label should not be used as a live dependency map.

Recovery moved from a router question to a control question

The retirement appears orderly at the routing level. The old /24 was withdrawn, remained absent for several weeks and then appeared with Amazon as origin. There was no prolonged period in the examined history when AS24910 and AS14618 both visibly originated the exact prefix. That sequencing reduces the obvious risk of competing origins, though it does not reveal the internal cutover plan or application downtime.

Under the old arrangement, recovery from upstream loss might have required Equinix to restore a managed circuit, local staff to repair or replace edge equipment, or JPMorgan network staff to move the route. Recovery from a building outage might have required an alternate site and an ability to originate the /24 there. The public record shows no second observed upstream, so there is no basis for claiming automatic internet failover.

Under the current arrangement, public route recovery is largely an Amazon control-plane responsibility, but service recovery is not automatically solved. A portable /24 can be advertised globally while an application remains tied to one region, one load balancer, one firewall policy or one identity system. Resilience depends on how the address pool is attached and how health is measured. Public BGP sees the front door, not the rooms behind it.

The corporate transition adds another recovery dependency: accurate authority. AS24910 and the /24 remain linked to JPMorgan's RIPE organisation, the old UK company is dissolved, Highbridge Capital Management, LLC remains active, and the London research presence continues. During an incident, someone must know which legal and operating entity owns each contract, who can change registry data, who controls AWS, who can enter the London premises and who communicates with affected users.

Stale route objects are a small but revealing sign. The RIPE route object still names AS24910 while BGP uses AS14618 and other registries contain Amazon-origin alternatives. This does not itself interrupt traffic, but stale policy descriptions increase the work required to distinguish intended state from residue. A sound control environment would reconcile current origin authorization, address ownership, registry contacts, cloud ownership and emergency procedures.

The next decisive evidence would therefore be operational rather than promotional: a current architecture statement for the /24, an ownership and authorization matrix, physical circuit and building-entry records for the London office, failover tests, application health evidence, route-origin authorization status, and restoration objectives. None needs to be public to be useful internally, but without them an outside assessment must stop at the visible boundary.

What would change the verdict

The regional-ISP finding would require a fundamentally different class of evidence. A current offer of internet access to outside customers would be the starting point, followed by a serviceable geography, tariffs, installation terms and an identified access medium. Claims of fibre or fixed wireless would then need support from construction records, pole or duct rights, tower or radio information, customer equipment and operating notices. Nothing in the current Highbridge, JPMorgan, Companies House or routing material supplies that chain.

A live-enterprise-network finding for AS24910 has a simpler test: the AS would need to originate at least one route again, with collectors seeing it and registry policy aligned to the observed origin. That has not happened. A private use of the number would not change the public conclusion because an autonomous system that appears only inside a closed environment is not a current public edge.

The historical resilience grade could improve without reviving the AS. Circuit orders, topology diagrams, colocation invoices, building-entry records, power designs or provider service descriptions from the operating period could show whether the apparent Equinix concentration concealed multiple physical paths. Incident reports and change records could identify the actual failure and restoration sequence. Those records would also settle whether the border function sat at Tudor Street, in colocation or inside a managed service.

The current /24 assessment needs a different evidence set. An AWS architecture statement could identify whether 91.213.81.0/24 supports a regional address pool, a global front door or another service, while account ownership and route authorization could establish who can change it. Application health and dependency mapping would show whether the block remains important to Highbridge, another JPMorgan function or no public-facing workload at all. Until then, the Amazon origin is a network fact and the application story is open.

Corporate clarification would help too. Highbridge could identify the legal entity responsible for the continuing London research presence and the party that contracts for office connectivity. That would not transform the firm into a carrier, but it would close the gap created by the dissolution of the old UK company. It would let readers separate the historical company named in filings from the active organization using the same London address.

These are not minor missing details. They divide five different questions: whether the institution is active, whether the London office is active, whether AS24910 is active, whether the portable /24 is active and whether a particular application is resilient. The public evidence answers the first, second, third and fourth. It does not answer the fifth.

The infrastructure conclusion is a retirement, not a broadband story

AS24910 once gave Highbridge's London identity a distinct place in the global route table. Its registration, London /24, long route history and final Equinix adjacency support a real enterprise edge. The number also originated selected institutional address blocks beyond London at different times, showing that its role was not reducible to one office cable. What it never establishes is a retail access network.

The 2019-2020 transition is the most informative event in the record. The AS stopped originating its final London prefix. Amazon began originating that same portable block. The old UK fund-management company later entered a solvent voluntary liquidation and was dissolved, while Highbridge's business and London research presence continued. Address identity, route origin, legal entity and office activity all followed different timelines.

That is why the remaining uncertainty should not be filled with regional-ISP assumptions. There are no evidenced subscribers, access lines, poles, towers, field installers or local tariffs. There is no current AS24910 route. The historic upstream observation shows one logical provider, not a protected ring. The current Amazon origin shows a cloud-network boundary, not an application design. The London address shows presence, not the location of every system.

The final judgement is Negative for the regional-ISP category. For the narrower institutional story, the evidence is Medium: strong on identity, route withdrawal, historical adjacency, address continuity, corporate status and current origin; weak on physical topology, capacity, workload placement, power, staffing and recovery performance. HIGHBRIDGELDN matters precisely because it shows how an enterprise network can disappear from BGP while the institution, office and addresses continue in other forms.