Arelion’s AS1299: Tier-1 reach demands constant renewal

Summary

• Arelion is the private operator formerly called Telia Carrier; Twelve99 retains a technical and historical identity, while AS1299 is its global routing identity.
• AS1299 reaches the whole Internet through customer and settlement-free peering routes, but being Tier-1 does not guarantee shorter paths, better support, or immunity from incidents.
• IP transit and dedicated access coexist with Ethernet, wavelengths, managed optical networks, cloud and exchange connectivity, DDoS mitigation, mobile, voice, and AI Direct.
• Accumulated routes, fibre, points of presence, peers, and knowledge create advantage; continuously funding capacity, security, and expansion remains the burden, with no public standalone accounts.

Four names describe four distinct layers

The business was born within the Telia group and for years operated as Telia International Carrier and later Telia Carrier. Polhem Infra completed the acquisition of Telia Carrier on 1 June 2021 and transferred control away from Telia Company. The operator adopted the name Arelion in January 2022. Those dates separate corporate history: references to Telia ownership are correct before completion and erroneous as a description of the current company.

Twelve99 did not disappear with the rebranding. The name remains visible on the technical domain twelve99.net, including the public looking glass, and continues to be associated with AS1299. That persistence is useful for engineers because autonomous system numbers, hostnames, route filters, customer configurations, and operational references often need more stability than a commercial brand. It does not create a second company. Twelve99 is technical continuity; Arelion is the current corporate and commercial identity.

AS1299 is another layer. It is the autonomous system identity through which the backbone originates, receives, selects, and announces routes. An ASN is not a legal entity nor a physical cable. It identifies a routing domain whose policy is expressed in routers, points of presence, and interconnections. Arelion operates that domain, while the underlying paths may use fibre under various ownership and contract models.

The four-part distinction — former brand, current company, technical name, and routing system — avoids several common mistakes. It prevents historical continuity from becoming a false ownership claim, avoids treating an identifier as a subsidiary, and forces network figures to be attributed to the company that publishes them. It also exposes the central idea of Arelion’s story: corporate control changed faster than the operational identity of the network.

A Tier-1 network is a relationship, not a badge

The expression “Tier-1 backbone” compresses several facts into a single label. At the routing layer, a Tier-1 autonomous system can reach the entire Internet through routes learned from its customers and through settlement-free peering with other large networks. It does not need to buy upstream transit to obtain general global reach. That position matters because it eliminates a class of supplier dependency and allows selling full-table transit to other networks.

It must also be maintained: peers can change their policies, traffic can shift, and the operator must retain sufficient capacity and direct relationships for the settlement-free exchange to remain mutually acceptable.

None of that definition certifies latency, support quality, DDoS protection, enterprise access, or financial strength. BGP chooses paths according to policies and attributes, not a universal shortest-distance rule. A network can be Tier-1 and still depend on colocation operators, submarine systems, access carriers, router and optics suppliers, and customers whose traffic sustains its economic relevance. The label describes routing independence under a specific set of interconnection relationships. It should open analysis, not close it.

Arelion’s current profile makes that difference clear. The commercial proposition is not only that AS1299 has global reach. The company claims it can turn that reach into IP transit, private packet transport, optical services, access to clouds and exchange points, network security, and operational support. Buyers therefore encounter Tier-1 status through a contract whose value depends on demarcation, geography, capacity, routing policy, and the ability to repair faults.

From the lineage of a national incumbent to an international operator

Arelion places the organic development of its backbone in 1993. The network grew within Telia’s lineage, but its operational logic was international. A national operator serves consumers, mobile, and businesses in a domestic market. An international carrier must connect other networks across countries, place routers in neutral facilities, acquire long-haul routes, maintain optical capacity, and negotiate interconnection with organisations that may also be competitors.

That difference helped Telia Carrier separate from its former parent. By 2020, the carrier had customers and infrastructure relationships far beyond Telia’s domestic business. Telia Company agreed to sell it to Polhem Infra in October 2020; the acquisition completed the following June. The closed transaction, not the announcement, marks the change of control. Polhem Infra’s account also noted a continued strategic network relationship with Telia, showing that shareholder independence did not eliminate commercial interdependence.

The 2022 rebranding gave the independent operator a name separate from its former owner. The underlying asset was not a newly assembled network. It was a mature network whose fibre paths, PoPs, customer connections, and peering relationships had accumulated over decades. This matters because direct connectivity cannot be reproduced by simply installing a control plane. A new entrant can lease capacity and open ports quickly in certain markets, but it cannot immediately recreate the history of direct routes, operational trust, and fault recovery of a established global network.

Independence also changed the investment question. Inside Telia, the carrier competed for capital in a diversified group. Under Polhem Infra, it became a focused infrastructure investment backed by Swedish pension capital. That structure can favour long-duration investments, but it does not eliminate return requirements or make capital unlimited. It changes who decides, how the asset is framed, and what evidence outsiders can see.

The backbone is a chain of control, not a homogeneous asset

Arelion reports more than 80,000 kilometres of fibre, more than 350 points of presence, and service in 129 countries. It also declares more than 2,000 customers and around 450 access partners. These figures describe distinct layers and must remain attributed to the company. The kilometres refer to physical reach; the PoPs to service and interconnection locations; the countries to commercial availability; and the partners extend delivery beyond direct facilities. Adding them together would not produce a useful metric.

The physical chain can include fibre owned by Arelion, controlled through long-term agreements or leased as capacity; optical systems installed on that fibre; routers and switches; colocation space and power; cross-connects with customers and peers; capacity on submarine systems; connections to clouds; and local circuits provided by other carriers. Arelion can control the service and routing policy while sharing physical failure domains with suppliers. “Global backbone” describes a system assembled through various forms of control; it does not prove that a single company owns every trench, cable, and building.

The distinction is operational. Two logical routes can appear diverse on a diagram while traversing the same conduit or submarine cable. Two services can be sold separately and share a router, optical system, or power supply. Conversely, leased infrastructure can be highly resilient if the contract, monitoring, and physical diversity are well designed. Ownership alone does not answer the reliability question; the useful evidence is the failure domain and the provider’s ability to observe and restore it.

Arelion’s commercial platform therefore begins with the control of a chain. It must know which segment is part of its backbone, which depends on a facility, which is supplied by a partner, which end is controlled by a cloud, and where the SLA changes. The customer sees an order and a support relationship, but an incident can traverse several organisations. Part of the operator’s value consists in making that chain manageable without pretending it is a single one.

Fibre provides the path; optics provide usable capacity

Fibre length is a geographical measure, not a capacity figure. A strand can carry a varying number of wavelengths, and each wavelength can be upgraded with new coherent optics and line systems. Therefore, the same physical route can support much more traffic after an equipment refresh without opening another trench. Arelion’s programme of 400G and higher capacities belongs to this optical layer, where transponders, coherent modules, amplifiers, and spectrum engineering turn glass into active transport.

A wavelength service delivers a dedicated optical channel over a validated route. It is suitable for predictable, high-volume movements such as data centre interconnection, replication, or carrier aggregation. The customer gets a clearer capacity boundary than with public IP transit, although the wavelength still depends on fibre, optical equipment, and route protection. “Dedicated” refers to the channel; it should not be extended to the idea that the customer owns a physically isolated cable.

Arelion also sells dark fibre and a Managed Optical Fiber Network. The managed offering shifts more design, equipment, and operations responsibility to the operator. For a large enterprise, cloud provider, or another carrier it can avoid building an optical organisation for each route. The trade-off is dependence on Arelion’s supported geography, vendor choices, restoration process, and lifecycle.

Optical services also explain why a mature network cannot be treated as a finished asset. Traffic grows, interfaces move to higher speeds, and customers expect new paths to clouds and data centres. Each upgrade consumes capital before demand is certain. Building too little can create congestion; building too much can tie up optics, ports, and commitments. A carrier’s financial discipline is hidden in capacity planning long before the customer notices a speed change.

Points of presence turn long-haul capacity into a market

A point of presence is where network capacity becomes accessible for interconnection. It can hold routers, optical systems, switches, ports, and cross-connects within a data centre or carrier hotel. A dense PoP footprint reduces the distance between Arelion and customers, peers, clouds, and exchange points. It also exposes the operation to power, cooling, building access, and cross-connect processes.

The stated total of more than 350 PoPs is a scale figure, not proof that every product and speed is available in each. A location may offer IP transit but not a particular wavelength route; Ethernet at 400 Gb/s requires validation; a cloud connection depends on the on-ramp; and a remote site may need a partner tail. The useful buying question is “What service is available at this demarcation?” rather than “Is Arelion in this country?”.

PoP density also affects connectivity. Direct interconnection reduces the number of intermediary networks, and more direct customers and peers can improve route choice. Arelion claims it can reach 95% of end users in the United States and Europe within one network hop, and describes AS1299 as the most connected backbone by its metric. These claims must remain attributed because they depend on data, definition, and date. One hop does not equal minimum latency for every flow, and connectivity is not traffic share.

The packet layer carries multiple products over shared infrastructure

On top of optics, routers and packet systems turn capacity into routed and private services. AS1299’s BGP plane exchanges Internet routes. MPLS and segment routing features support traffic engineering and private transport. Ethernet presents Layer 2 connectivity; Smart IP-VPN a managed WAN; Cloud Connect and IX Connect extend paths to specific ecosystems. The products share parts of the underlay and expose different service boundaries.

The Ethernet portfolio includes point-to-point EVPL and multipoint ELAN from 10 Mb/s up to 400 Gb/s, depending on route and availability. The documentation describes segment routing and Flex-Algo for certain low-latency paths within the MPLS backbone. These controls can improve determinism within the provider’s domain, but they do not eliminate an access tail or external endpoint outside its full control.

SLA figures show that boundary. Arelion publishes 99.999% for MPLS backbone services, 99.99% for basic PoP-to-PoP Ethernet, and 99.5% for an end-to-end service with a managed NID and access tail. They are not interchangeable. The lower commitment for the broader service reflects additional equipment and dependencies. Actual terms depend on the contract and the route.

A customer can also run several services over a single port through logical separation. This reduces cross-connects and interfaces but concentrates services into the same physical failure domain. A port failure can affect several distinct circuits. Multiplexing efficiency and concentrated risk must be explained together.

IP transit is the defining commercial expression of Tier-1 status

An IP transit customer typically operates an autonomous system, announces prefixes via BGP, and receives the Internet’s routes from Arelion. The carrier carries outbound traffic to other networks and returns traffic to the customer’s prefixes. The product is sold as reach, capacity, and service, while the underlying value comes from AS1299’s relationships with customers and peers.

The transaction appears simple because BGP hides the physical path behind announcements. In operation, Arelion must maintain full tables, filter invalid or unauthorised announcements, balance traffic, provision capacity, protect sessions, manage communities, and recover from failures across many PoPs. A multihomed customer gains redundancy and choice, but also a more complex routing problem. It must understand preferences, inbound engineering, and failure propagation in each provider.

Tier-1 economics does not mean every interconnection is free. Settlement-free peering eliminates payments for a defined exchange with networks that meet conditions. Arelion still pays for fibre, facilities, equipment, power, access, staff, maintenance, and other relationships. It may also buy local services or partner access when that is more efficient than building. The absence of upstream transit is a cost element, not a zero-cost network.

The routing position must also be maintained. If the balance, geography, or policy shifts materially, a peer may demand different terms or end the relationship. The advantage is accumulated and continuously defended. The carrier needs enough customers, direct reach, capacity, and credibility for other large networks to keep the settlement-free exchange.

Dedicated Internet access shifts more routing responsibility to the provider

Dedicated access serves enterprises that want managed connectivity without running a full BGP relationship or global table. The customer may receive a default route or managed edge, while Arelion takes more responsibility for the design towards the Internet. The physical access and quality can resemble transit in certain places, but the operational contract is different.

The difference matters during failures and changes. A transit customer can use its own ASN, prefixes, communities, and multihoming. A DIA customer typically has less control and depends more on the provider’s edge routing and support. Neither model is inherently superior. Transit serves organisations with expertise and a need for policy; DIA serves buyers who prefer a managed demarcation.

SecureConnect combines access or transit with automated DDoS protection. The bundle reduces purchase and configuration boundaries for customers who would otherwise contract transport and mitigation separately. The protection still has a defined scope. It does not protect credentials, endpoints, application logic, or attacks outside the chosen connectivity service.

BGP communities turn routing policy into part of the product

BGP communities are tags added to route announcements. Arelion interprets customer tags and applies defined actions such as changing preference, limiting propagation, adding AS prepends, or activating blackholing. They give the customer influence over the carrier without direct access to its routers.

It is a powerful form of delegated control. An operator can make a prefix less attractive in a region, limit where it is announced, or sacrifice reach to protect the rest during an attack. The semantics are specific to AS1299; a value from another carrier may mean something else. Customers need current documentation, controlled changes, and verification with routing tools.

Communities can also cause failures. A wrong tag can withdraw reach or send traffic through an undesired path. A remote-triggered black hole deliberately discards traffic to a destination to prevent congestion. It is useful in an emergency because it trades one target’s availability for overall stability. It is not mitigation without cost.

Looking glasses expose a portion of routing reality

The Twelve99 looking glass allows inspection of routes, pings, and traceroutes from selected nodes. It answers practical questions: how AS1299 sees a prefix, which path it chooses from a point, and where latency appears. It also preserves Twelve99 as a technical identity after the rebranding.

A looking glass is evidence from a single viewpoint, not an audit of the entire network. It does not show every alternative, historical incident, internal preference, or physical path. BGP policy varies by location, and traceroute can hide equipment or respond differently under load. It is most useful combined with customer measurements, collectors, tickets, and provider data.

Arelion also publishes monthly IP performance metrics. They help track measures chosen by the provider, but methodology and scope remain theirs. Publishing data improves transparency; independent verification still requires definitions and external observation.

RPKI reduces a risk without fixing BGP completely

A Route Origin Authorisation allows a prefix holder to indicate which AS may originate it. Validation classifies the announcement as valid, invalid, or not found. Filtering invalid origins reduces certain hijacks and errors. Arelion’s material presents RPKI as part of routing security.

The mechanism validates origin authority on covered prefixes. It does not validate the full AS path, does not guarantee correct exports, and does not prevent every leak. A valid origin can still be propagated through an undesired path, and errors can exist in filters, entities, or configuration. RPKI reduces the trust problem; it does not replace BGP policy, monitoring, or response.

For a Tier-1, the boundary matters because security and availability can conflict. Aggressive filtering can block legitimate traffic if registries are wrong; weak filtering can propagate invalid announcements. The operator needs gradual deployment, communication, exceptions, and current data. Saying “RPKI secures BGP” conceals the remaining operational judgement.

IX Connect sells access to interconnection, not full reach

Internet exchange points provide places to establish peering. IX Connect transports the customer from an eligible location to an exchange port, enabling remote participation without building local presence. It can reduce cost and time when reaching multiple peers, particularly when entering a market.

The transport does not create the peering policy. The customer still needs membership, compatible ports, bilateral or route-server agreements, and its own filters. A remote path adds another dependency versus a physically placed router. IX Connect is access to a route market, not a substitute for the customer’s strategy.

The product shows how a Tier-1 cooperates and competes with IXs. Arelion peers on them and sells transport to them. A customer can replace some transit with direct peering and still buy Arelion capacity to reach the exchange or cover the rest. The boundaries between transit, peering, and transport are commercial decisions over the same physical network.

Cloud Connect brings the network to a cloud-controlled demarcation

Arelion lists private connectivity with AWS, Microsoft Azure, Google Cloud, Oracle, and IBM. Cloud Connect transports traffic to supported on-ramps and reduces reliance on public routes for that leg. It can offer more predictable routing, capacity, and security than sending everything over general transit.

The service terminates at a shared boundary. The cloud controls the virtual interface, regions, quotas, and internal fabric. The customer controls the account, routes, and workloads. Arelion controls the sold path. A failure can be on any side, and configurations must align. Calling the whole “private” is misleading if it implies ownership of all components or the absence of shared infrastructure.

Cloud connectivity also changes competition. Hyperscalers operate private networks and carry more traffic between regions. Arelion’s opportunity lies in linking enterprises, data centres, multiple clouds, and networks outside a single domain. Its limit is the same: it does not control what happens inside the cloud or replace its fabric.

DDoS defence is already part of capacity planning

Volumetric attacks consume links, routers, and scrubbing capacity. A global carrier sees enough traffic to detect shifts and divert attacks before congestion reaches the access. Arelion’s service combines detection, diversion, and cleaning, and returns permitted traffic.

In the 15 July 2026 report, Arelion said that Aisuru represented roughly a third of observed attack traffic and that the largest attack reached 6.1 Tb/s. These are the provider’s observations on its network and methodology. They show the scale visible to a major backbone, not a universal, worldwide census.

The 6.1 Tb/s should not be turned into a mitigation guarantee. An attack can be observed in a distributed fashion without a single customer receiving the full volume, and commitments depend on architecture and contract. Effectiveness also depends on detection time, convergence, scrubbing location, clean capacity, and false positives. Application-layer attacks can pass volumetric filters because their packets look legitimate.

Remote-triggered blackholing is the emergency extreme. A community can make traffic to an attacked prefix be discarded. It protects shared capacity and makes the destination unreachable. Scrubbing seeks to preserve service; blackholing accepts an outage. A mature product needs both, clear triggers, and defined customer authority.

SecureConnect integrates mitigation into the connectivity purchase. It can improve adoption by being active before the incident. The strategic test is not the name but the onboarding rate, response evidence, protected scope, and the ability to scale with peaks.

AI Direct bundles WAN transport for distributed AI systems

AI Direct is the connectivity portfolio for moving data between AI clusters, data centres, and clouds. It combines Ethernet, wavelengths, Internet access, managed optical networks, and security under an AI infrastructure proposition. It does not provide GPUs, storage, or training software. Its role begins when data leaves a facility or domain.

The boundary is economically important. Training inside a campus depends on a high‑performance local fabric that a WAN carrier cannot replace. WAN demand appears when datasets, checkpoints, replicas, inference, or complete loads move between sites. Arelion’s advantage is reach; its service connects to the internal fabric, it does not become it.

In May 2026 it added 400G EVPL to AI Direct. The announcement establishes a high‑capacity packet option on qualified routes, not universal availability. In April it had announced Danish network investments and landing stations as part of a Nordic AI corridor. Earlier additions included Oklahoma City, while the Mexico channel expanded distribution. The facts show an active strategy, though utilisation, customer concentration, and per‑project investment were not disclosed.

Evidence that AI Direct exceeds a label would be named customers, activated ports, recurring traffic, and per‑route commitments. “AI highway” expresses intent; it does not reveal revenue, utilisation, or new capacity attributable to AI. The evidence supports a credible portfolio and an investment direction, not that AI has already transformed the finances.

Enterprise services monetise the backbone beyond transit

Transit prices face long‑term pressure as capacity and alternatives improve. Arelion can sell the same routes with more control, demarcation, and support. Ethernet, Smart IP‑VPN, Cloud Connect, IX Connect, managed optics, and DDoS increase the problems solved per customer.

Breadth does not mean uniform margin or buyer. A wholesale carrier may buy transit and wavelengths; a multinational managed access and IP‑VPN; a cloud, optics; a regional ISP, transit and IX. The organisation must sell common infrastructure through different contracts and support wrappers.

The broad offering also creates internal complexity. Capacity for private services, public transit, and mitigation is planned jointly. An optical change can affect several products. A local delay can stall an automated order. Reuse improves economics only if isolation, changes, and capacity accounting are correct.

Mobile, IoT, voice, and messaging services widen the catalogue further. The evidence is strongest for backbone and connectivity; they should be treated as current categories without inferring share. Their relevance is that Arelion is not a pure transit wholesaler, even though AS1299 remains its defining identity.

Access partners extend reach and change the SLA boundary

Arelion states around 450 access partners. They can connect a customer building to a PoP when the network does not reach directly. The model extends coverage without building every last mile and gives the customer a single commercial relationship for a multi‑provider path.

The access tail is often the least standard part. Timescales, repair, bandwidth, jumbo frames, NID, and regulation vary. Arelion may manage the tail, but the physical repair belongs to another carrier. The lower end‑to‑end availability versus PoP‑to‑PoP reflects that exposure.

Global reach must be read as delivery capability, not a map of ownership. Diligence includes naming the local provider, diversity, escalation rights, and shared ducts. A global contract simplifies governance, but it does not change local physics.

Operations turn assets into a usable service

A backbone only has value if it can be provisioned, observed, and repaired. Arelion’s model includes MyArelion, management, and NOC support. These systems turn ports, routes, tickets, maintenance, and performance into a customer experience.

Automation accelerates changes, but it does not turn everything into an API. Fibre repair requires crews; a cloud may refuse configuration; a peer can change policy; a data centre may demand a cross‑connect; a local carrier may miss dates. Automation quality includes representing exceptions and responsibilities, not just accepting normal orders.

Arelion reports experience awards and high NPS. They can indicate a service culture, but methodology and sample are not independently audited in the evidence. Awards and surveys supplement operational detail; they do not replace it.

The business model reuses a base through several demarcations

Arelion sells capacity, routes, service guarantees, and operational responsibility. Transit monetises reach; Ethernet and IP‑VPN private paths; wavelengths and managed optics capacity and engineering; cloud and IX access to ecosystems; DDoS visibility and protection. Each product starts at a distinct demarcation even though they share the backbone.

Reuse can improve utilisation. A route and a PoP generate more revenue with several products. Nevertheless, isolation and margin are needed. If everything is planned on the same optimistic assumption, an attack or traffic shift reveals contention. Accounting must separate sold ports, normal traffic, protected traffic, and failure scenarios.

Commercial leverage reduces coordination for the customer. A multinational could separately contract local carriers, IXs, clouds, transit, optics, and security. Arelion integrates part of the chain. The customer pays for bits, fewer contracts, defined support, and transferred responsibility. The operator only earns a premium if the integrated service works better than the fragmented alternative.

Pension‑backed ownership changes the horizon, not the economics

Polhem Infra owns Arelion and, according to its current description, is jointly owned by the Third AP Fund and the Fourth AP Fund of Sweden. This connects the carrier with national pension capital through an infrastructure vehicle, suitable for routes and interconnection built over years.

Patient capital can accept fibre rights, optical upgrades, and PoPs with multi‑year returns and favour stable cash flows. It is still subject to return duties. The pension affiliation does not mean demand‑free funding, nor does it disclose the owner’s mix of debt, cash, and equity.

The governance shift from 2021 was more visible than the technical one. AS1299 kept routing while control, funding, and priorities moved from Telia to Polhem. Daniel Kurgan was appointed CEO in October 2023 and Charles Gill CFO in March 2024. The current page shows a broad leadership, not a complete statutory register.

Executives should not be attributed every architecture decision. The network is the accumulated work of engineering, operations, procurement, sales, and partners over decades. Leadership sets capital, risk, product, and disclosure; daily quality depends on distributed expertise and institutional memory.

Private ownership leaves core financial questions open

The evidence does not establish Arelion’s current revenue, profit, debt, capex, valuation, traffic, contract duration, or customer concentration. This is material because a global network must fund equipment, routes, colocation, access, and security. Announcements show activity, not return.

The historical transaction value would describe 2020–2021, not 2026. Carrying it forward ignores changes in traffic, assets, capital, and market. Nor is a connectivity ranking a proxy for revenue or margin. A network can be well connected in a market where the per‑bit price is falling.

The opacity changes the assessment. Public BGP and looking glasses show reach; news shows activity; customers show some deployments. Nothing reveals whether capital is generating a return, whether one customer dominates, or whether leverage limits the next upgrade.

The limited conclusion is that Arelion appears active, global, and in continuous investment. The evidence does not allow an estimate of profitability or a claim that pension capital guarantees the next cycle.

The economics of renewal lies between physical ownership and contractual control

The investment question is more complex than deciding whether Arelion should own or lease a particular route. A Tier‑1 carrier needs to hold several forms of control at the same time. Fibre ownership brings direct authority over a physical path, but it still leaves dependencies on permits, power, repair crews, landing stations, and technology suppliers. Long‑term fibre rights can resemble ownership in operational planning while locking in multi‑year contractual commitments. Buying capacity on shorter contracts preserves flexibility, though it exposes the carrier to repricing, supplier availability, and renegotiation.

The right structure depends on the route, the market, and the consequences of a failure.

This matters because the network’s economics are set before the service is sold. Arelion may have to deploy optical systems, routing capacity, cross‑connects, and headroom before a new customer commits traffic. Expansion into a PoP can require facility contracts and equipment even when the business case depends on future demand. A high‑capacity AI corridor can look attractive at market scale while each individual buyer remains unsure where their workloads will sit over the next few years. Capital therefore arrives ahead of some of the revenue evidence that is supposed to justify it.

The same temporal problem exists for resilience. A backup route has value precisely when the primary fails, so part of its capacity may appear underutilised in normal conditions. Finance can see idle headroom where operations see protection. The network must decide how much apparently free capacity is economically justified by the contracts and by the failure scenarios it protects. Too little reserve can turn a fibre cut or an equipment failure into customer loss; too much can lower return if pricing does not recognise the value of availability.

Arelion’s broad portfolio complicates planning further because the same physical and packet infrastructure supports products with different demand profiles. Transit can be variable and price‑sensitive. Ethernet and wavelengths can carry large committed flows. DDoS mitigation needs capacity that may stay lightly used until an attack arrives. Cloud connections depend on external locations and platforms. A planner cannot therefore simply add traffic averages. They must model correlated peaks, reroutes during failures, maintenance, and the possibility that several services migrate simultaneously to the same surviving path.

The AI Direct expansion in 2026 adds another layer of uncertainty. Arelion has a reasonable reason to position high‑capacity Ethernet and optical services around distributed AI infrastructure, because moving data between facilities generates real WAN demand. However, AI‑related demand does not remove traditional carrier investment discipline. The route still has to link physical facilities, ports must be activated, the customer must pay for capacity, and traffic must last long enough to justify equipment and committed rights. A popular workload category does not change the accounting of unused capacity.

Vendor strategy also defines renewal. Router and optical platforms are not interchangeable without cost. A major upgrade can create years of operational familiarity, spares planning, software dependencies, and manufacturer support. Open optical systems and coherent modules can reduce some lock‑in by separating elements of the optical layer, but interoperability still requires engineering and testing. Optionality only has value if the operator retains the capabilities and processes to use it.

Facilities create a similar form of commitment. A PoP gains value as customers, peers, and clouds connect to it, but that density makes it harder to move. Cross‑connects, customer equipment, maintenance processes, and commercial relationships accumulate around the location. A data centre can gain bargaining power even when Arelion owns the routers sitting inside. The operational question is whether alternative locations, physical diversity, and sufficient commercial influence exist to prevent a building or a supplier from becoming a hidden strategic constraint.

That is why private financial opacity matters in an infrastructure profile. Without current figures for capex, debt, lease commitments, and cash flow, an outside observer does not know whether renewal is funded from operating cash, owner capital, or debt, nor how much headroom remains for the next hardware cycle. Technical announcements prove that investment is happening. They do not prove that the same pace can be maintained if prices worsen, demand slows, or resilience needs increase.

The strongest external evidence is cumulative: routes that are completed, not just announced; PoPs that attract customers; modern interfaces that become orderable; direct relationships that are maintained; and security systems that absorb larger peaks. None of those items proves profitability by itself. Together, they show whether capital is being turned into an operational system capable of keeping AS1299 economically relevant.

For Arelion, the renewal obligation is inseparable from its advantage. Part of that advantage comes from a network position accumulated over more than three decades. That position prevents customers from having to rebuild the same reach on their own. But it also creates a complex installed base that must be upgraded without interruption. The more valuable AS1299 is as a neutral global route, the greater the potential cost of deferring investment. Tier‑1 status therefore brings interconnection power and simultaneously imposes a permanent capital obligation.

Geography is a map of service, route, and dependency

The roots and ownership are Nordic, while the network covers Europe, North America, and Asia. The 129 countries describe commercial reach, not owned fibre systems or PoPs. A buyer needs three maps: direct infrastructure, partner delivery, and technical availability per service.

Europe is dense by history. North America is a core market, and Oklahoma City was presented around cloud and AI. The 2026 Danish investment affected routes and a landing station important for the Nordic corridor. In Mexico, routes and a channel expanded access. There is less public detail for Asia than for Europe and the US.

Submarine cables require separate treatment. Arelion may buy spectrum, install equipment, and operate a service without owning the entire cable. Cable faults, stations, and shared backhaul affect multiple carriers. Logical diversity must be verified at the cable and duct level.

Geopolitics also changes the economics: cross‑border regulation, sanctions, permits, cable security, and resilience policies. The dossier does not establish a current specific dispute, so these are structural constraints, not accusations.

Competition occurs in several layers at once

At Tier‑1 and global transit, Arelion competes with NTT, Lumen’s Level 3 lineage, GTT, Tata Communications, Cogent, and Sparkle. Comparison demands dated data on routes, geography, latency, capacity, security, support, and price. A ranking does not decide everything.

At the physical and optical level, Zayo competes on fibre and wavelengths. In the cloud, Megaport and Equinix Fabric offer software‑defined interconnection. In the enterprise, regional carriers, SD‑WAN, and managed services assemble alternatives on top of the Internet. They are overlapping, not identical, categories.

Hyperscaler backbones pose another challenge. A cloud can internalise traffic and bundle transport with compute. Arelion serves where clouds, enterprises, data centres, and networks need linking outside a single domain. If more traffic stays within a cloud, some opportunities fall; if workloads distribute, neutral reach gains value.

IXs both complement and substitute some transit. A network can peer directly for large destinations and buy less transit; it still needs remaining reach, transport to the IX, and support. IX Connect and transit place Arelion on both sides.

The advantage is not that each component is uncopiable. Fibre, routers, on‑ramps, and DDoS can be bought. The set — routes, peers, customers, PoPs, optics, support, and reputation — takes years and requires continuous investment.

Connectivity creates resilience and contagion at the same time

A well‑connected network offers more paths and direct reach and can reroute flows. It also exposes more organisations to its mistakes. A leak, filter, or incident can propagate beyond direct customers because other networks depend on AS1299’s routes.

Physical concentration has the same double effect. Dense PoPs improve interconnection, but a single facility can affect several services. Logically diverse topologies share ducts. A DDoS platform protects, but a diversion error moves a lot of legitimate traffic. Scale amplifies both capacity and blast radius.

Resilience depends on less visible controls: maintenance, staged changes, filters, RPKI, community governance, optical protection, reserves, scrubbing, out‑of‑band access, communications, and rollback. Pages and tools show elements, not a complete track record or audit.

Customers also have responsibility. Multihoming, prefixes, policy, testing, and local diversity determine how much resilience reaches the edge. Buying Tier‑1 does not remove the need to survive the loss of that carrier.

Sustainability claims need denominator discipline

Growth consumes equipment, energy, colocation, construction, and maintenance. Coherent optics can increase capacity per watt and per space. Efficiency can coexist with higher total consumption if traffic and footprint grow more. Arelion publishes a framework, but there is no full audited network and supply‑chain footprint in the evidence.

Efficiency must be separated from absolute impact. An upgrade reduces energy per bit; a new route adds equipment and power. Leasing transfers part of the emissions to suppliers, it does not eliminate them. Without consistent scope, baseline, and traffic, net claims cannot be sustained.

Sustainability matters to customers and pension owners and affects where capacity can be deployed because of energy and cooling. The signal to watch is whether reporting becomes granular enough to connect expansion with measured consumption.

What public evidence cannot show

The scale figures are primarily Arelion’s assertions. The evidence pack contains no route‑by‑route audit of 80,000 km, no uniform PoP definition, and no complete inventory of ownership versus lease. There is also no comparable traffic series across competitors.

The ranking depends on the metric. The 95% one‑hop claim describes declared reach in the US and Europe, not a guarantee for every destination, protocol, or moment. Monthly metrics are useful but provider‑selected. Purchasing demands per‑route and per‑contract evidence.

Availability is conditional. Ethernet supports up to 400 Gb/s, but not everywhere. AI Direct is connectivity, not proof of revenue. Observing 6.1 Tb/s does not reveal contractual capacity. MEF/Mplify certification supports definitions, not every tail or incident.

The financial gap is larger. Revenue, margin, debt, capex, valuation, and concentration are not published in the evidence. It should not be filled with assumptions derived from ranking, customers, or owner.

Renewal remains continuous even though the routing identity appears stable

AS1299 can appear remarkably stable from the outside. The autonomous system number stays, the known peers remain visible, Twelve99 continues to work as a technical reference, and customers continue to receive a global table. That stability is valuable, but it can hide the volume of change needed underneath. A backbone can keep the same public identity while replacing routers, optical platforms, line cards, software versions, power arrangements, fibre routes, and facility connections. The real product is continuity achieved through controlled change.

Hardware generations create a first renewal clock. Router capacity must follow the growth of interface speeds, route tables, telemetry, and security features. A platform sufficient when 100G dominated can become a constraint when 400G and higher aggregate rates become normal. The replacement moment does not only depend on whether the gear still forwards packets. Spares, manufacturer support, power density, software compatibility, and the cost of operating parallel generations can make a technically functional platform economically obsolete.

Optical systems create a second clock. Coherent technology allows the capacity of an existing fibre pair to be multiplied, but upgrading is not a universal switch. Distance, fibre quality, spectrum, amplifiers, ROADMs, open‑line design, and operational qualification determine what each route can carry. That is why Arelion’s ability to offer higher rates depends on per‑corridor engineering, not only on a module’s nominal speed.

Software creates a third clock. Routing platforms receive security patches, protocol changes, new features, and fixes. A stable network can deliberately avoid the latest release until it is proven, but staying on old releases for too long increases support and security risks. Upgrades must be staged so that fixing a defect does not cause a larger availability problem. At a Tier‑1, the quality of change management is as important as code quality, because a policy or control‑plane error can affect traffic far beyond a single customer.

Internet practices create a fourth clock. RPKI adoption, filtering expectations, peering policies, BGP community conventions, and security procedures evolve without the ASN changing. A well‑run network ten years ago cannot assume its historical controls are still enough. Customers and peers gradually expect more validation, more policy transparency, and better response to leaks and hijacks. Keeping trust requires process renewal as well as capacity renewal.

Facilities have a slower but equally important cycle. A PoP can stay in the same building while power density, cooling, cross‑connect costs, and the peer mix change. A historically important location may no longer be the best for new high‑capacity hardware, even though moving existing connections is expensive. New data‑centre campuses can draw traffic away from traditional carrier hotels. The network must move closer to new centres without losing the interconnection density that makes the old ones valuable.

Customer geography also changes. Enterprise applications migrate to cloud regions; content platforms add edge; AI clusters appear around energy‑abundant markets; wholesale networks open new national nodes. Arelion could keep the same PoP count and lose relevance if those points stop being near the demand that matters. Conversely, a few strategic expansions can greatly raise the value of the existing backbone without dramatically transforming the global footprint.

Security capacity follows an especially uneven demand curve. Normal traffic can be planned against known peaks, but DDoS systems must withstand rare events by definition. The largest attacks observed in one year can become common later as botnets, access speeds, and techniques grow. Sizing scrubbing against the average leaves the network exposed; building unlimited unused capacity destroys economics. The renewal challenge is to keep enough distributed headroom and diversion flexibility for plausible attacks while using the infrastructure efficiently under normal conditions.

Access partners also have a lifecycle. A local carrier can improve, merge, change ownership, retire a product, or lose competitiveness. The best choice for an enterprise site three years ago may not be the best today. Global quality therefore depends on continuous supplier qualification and the ability to steer new orders and, where viable, existing services towards better alternatives. The partner count measures breadth; renewal is about deciding which ones still deserve traffic.

Contracts age even when the network does not change. Long‑term capacity rights can look attractive when signed and expensive versus the market years later. Short agreements preserve flexibility and increase exposure to renegotiation. Facility, fibre, and vendor contracts also expire on dates that may not match customer contracts. The operator must manage those contractual clocks alongside the physical ones to avoid a route becoming unattractive simply because several commitments renew at a bad time.

People create another, less visible renewal need. Global backbones depend on knowledge of policies, optical characteristics, historical incidents, vendor behaviour, and customer exceptions. Some can be codified in inventories and automation; some remains institutional. Retirements, reorganisations, or outsourcing can remove context that only becomes critical during an unusual failure. A stable network needs succession, documentation, and training as much as spares.

Automation does not eliminate these cycles. It can improve inventories, accelerate provisioning, and expose dependencies, but it raises the importance of correct data. A wrong identifier, an out‑of‑date route, or a badly recorded demarcation can propagate faster through an automatic workflow than a manual one. The more Arelion standardises operations, the more important it is to validate the data model that underpins that automation.

The commercial consequence is that renewal spending should not be judged only by visible expansion. Replacing a card before failure, adding reserve capacity, moving a service off a correlated route, or upgrading a security platform may not add a country or a PoP. Those investments protect existing revenue. An owner measuring growth by headlines can undervalue them; an operator treating every preventive upgrade as mandatory can oversize capital. The useful discipline is to link each decision to a failure domain, a capacity limit, a contractual obligation, or a measurable operational improvement.

That is why the question “Is AS1299 still Tier‑1?” is too narrow. The routing relationship can be maintained while other aspects of the service improve or degrade. The fuller test asks whether the network tracks traffic growth, whether physical diversity remains real, whether peers and customers keep trust, whether security capacity follows the threat environment, and whether capital arrives before technical debt shows up as incidents.

Arelion’s long history is both proof and burden. More than three decades of operation mean the system has passed through several generations of hardware, traffic, and ownership. They also mean it contains decisions made in different eras. Renewal consists of keeping what still creates value and replacing what no longer fits. The ASN can remain precisely because many of the layers beneath it do not.

Buyers must verify the service boundary, not just the Tier‑1 label

For a customer, useful diligence begins after confirming that AS1299 is Tier‑1. The next question is where Arelion’s direct control over the purchased service begins and ends. A transit port in an Arelion PoP, a managed enterprise circuit delivered by a partner, a wavelength on an optical route, and a Cloud Connect all expose different failure domains. The same brand and the same backbone can sit behind operationally different contracts.

The physical demarcation must be clear. The customer needs to know whether the handoff is inside an Arelion PoP, in a third‑party data centre, at its own site, or at a partner facility. If local access exists, it must know who owns it, where it enters the building, and whether a second line really uses another physical path. Two circuit IDs do not prove diversity. The relevant test is separation in duct, entrance, facility, power, and upstream route.

The routing boundary matters separately. A transit customer controls its ASN and can use communities, preferences, and multihoming. A DIA customer delegates more edge policy. An Ethernet customer may receive Layer 2 and not see internal routing. A Cloud Connect customer reaches an on‑ramp where another administrative domain begins. Procurement must document these differences because they determine who can act during a fault.

Capacity must be verified at the exact boundary being bought. The backbone may support 400G while a local port, exchange, on‑ramp, or access carrier does not. A buyer with AI or data‑centre traffic must separate core‑network capacity from the speed actually orderable at both ends. Route qualification, port availability, and protection options are more useful than a general statement that a provider “supports” a speed.

Latency claims also need per‑route definition. Arelion’s connectivity can reduce intermediary networks, but the path with fewer AS hops is not always the physically shortest or the lowest‑latency. BGP policy, fibre distance, metro access, and the destination’s location matter. Latency‑sensitive customers should measure from relevant locations and know whether the service uses traffic engineering, protected routes, or standard public routing.

SLAs demand the same discipline. The difference between backbone, PoP‑to‑PoP, and end‑to‑end commitments explains why one figure is not enough. The customer should ask which components are included, how maintenance is treated, what exclusions exist, how it is measured, and what the remedy is. A high percentage can still allow a material outage if the measurement window and compensation do not match the real application cost.

For DDoS, detection, diversion, scrubbing, and emergency blackholing should be separated. The carrier’s ability to observe multi‑terabit attacks is a visibility signal, but the customer needs to know its own protection mode, activation conditions, covered prefixes, clean‑traffic return, and escalation procedure. It must also know what falls outside, especially application‑layer attacks that demand different controls.

Cloud Connect needs a three‑party operational model. Arelion delivers the carrier path; the cloud provider controls the interface and internal fabric; the customer controls account and routing. Troubleshooting works best when identifiers and contacts are registered before the incident. A private connection is valuable because it narrows the path and increases control, but it does not turn three domains into one operator.

Partner‑delivered access deserves particular attention because it is one of the hardest parts to standardise globally. The buyer should ask whether Arelion monitors the local circuit, receives proactive alarms, controls supplier escalation, and can obtain route information. It should also know what happens when the local provider repeatedly misses repair times. The global contract has more value when the lead carrier has enough data and commercial leverage to change the outcome, not only to forward tickets.

Operational transparency should be evaluated before purchase. Looking glasses, performance reports, and MyArelion provide information, but the customer must know which metrics are available for its service, how often they are updated, and whether it can export evidence. In a dispute, common timestamps, route observations, and incident logs are worth more than a general status page.

Change control is another boundary. Customers should know what they can modify through the portal or communities, what requires provider approval, and what Arelion may initiate during maintenance or security response. Delegated control is only useful if the action, its blast radius, and the rollback are understood. This is especially important when several logical services share a physical port.

A multihomed customer should test failure instead of assuming redundancy because it has two providers. Preferences can keep traffic on a degraded path longer than expected; prefixes can be filtered incorrectly; inbound behaviour can differ from outbound. Periodic testing shows whether the architecture uses AS1299 and the alternative carrier as designed.

The same holds for optical diversity. A protected wavelength may use two fibre paths, but the buyer must verify how they are defined and where they converge again. Two links with diverse endpoints can offer a stronger boundary than two channels that share a facility. The application’s own architecture may need to span more than one data centre when the facility is a critical failure domain.

Contract renewal should re‑examine those assumptions. A service bought three years ago may now be delivered by a different partner, facility, or architecture. New clouds and data centres can offer better endpoints. The provider may have added direct routes that change the optimal design. Treating renewal as an administrative price negotiation misses the chance to update the resilience model.

For BTW readers, this diligence shows the practical meaning of Tier‑1. AS1299 gives Arelion a strong routing position, but customers buy services at concrete demarcations. The backbone’s value materialises when policy, physical infrastructure, partner access, security, and operations work together at those boundaries. The label removes one dependency; it does not remove the others.

The central judgment

Arelion shows what a Tier‑1 has become. AS1299’s settlement‑free reach remains the defining position, but the commercial product is a layered platform. Fibre and optics give path and capacity. PoPs give access. BGP, communities, and security give policy. Ethernet, cloud, IX, and managed services create demarcations. DDoS uses visibility as a security function. Operations and support turn everything into a buyable service.

The advantage is historical accumulation. Arelion inherited decades of routes, facilities, peers, and knowledge from Telia Carrier. Polhem Infra brought an infrastructure investor and a new brand without replacing the technical identity. Twelve99 remains visible because a network has a longer memory than marketing.

That history creates an obligation. Tier‑1 status does not freeze the network. Prices change, attacks grow, clouds internalise transport, AI creates corridors, optics accelerate, and customers demand automation. Arelion must renew the platform and protect the trust that gives AS1299 its value.

The evidence allows the statement that Arelion is active, globally relevant, and technically broad. It does not allow a conclusion on profitability, leverage, or the conversion of AI announcements into durable revenue. The concrete conclusion is that a modern Tier‑1 is a routing relationship wrapped in capital‑intensive fibre, optics, packets, security, and operations, and its quality is demonstrated repeatedly at the service boundary, not by a label granted once.

Indicators that will show whether the backbone strengthens

The next phase must be judged by conversions, not slogans. Arelion already has a mature routing identity, global figures, and a broad catalogue. What is missing is knowing whether new routes turn into active demand, whether higher‑value services offset transit pressure, and whether resilience keeps pace with connectivity. These indicators link technical activity, commercial results, and operations.

Peer and route stability

Tier‑1 status depends on maintained settlement‑free relationships. Material changes in visible adjacencies, prefix paths, or propagation would be an early signal of economic shift. Looking glasses and BGP observation show the change, not the commercial terms.

Activated capacity, not just announced

New 400G products, optical upgrades, and routes matter when ports are active, traffic grows, and customers commit. Watch for qualified availability, completed routes, new PoPs, and named customers. Announcements without activation would weaken the narrative.

Product mix beyond transit

Volume can grow while the per‑bit price drops. If SecureConnect, Cloud Connect, IX Connect, managed optics, and Ethernet add to existing accounts, that would indicate monetisation of integration. Without per‑product revenue, customer and service announcements are the available proxies.

Access partner performance

About 450 partners extend reach and variability. It is worth tracking differences between PoP and end‑to‑end, lead times, incidents, and qualification. The SLA gap already shows how the last mile changes risk.

DDoS scale and mitigation evidence

The 2026 report sets the observed environment. The next ones should be read for peaks, botnets, time, impact, and distribution, not a headline. A larger observed attack does not prove larger protection.

Capital and governance signals

Private ownership limits analysis. Follow management changes, completed routes, hiring, vendors, funding, and Polhem’s posture. Less deployment or more partners could reflect discipline, demand, or both.

Five scenarios supported by the evidence

AI WAN demand becomes a durable driver

Neoclouds, enterprises, and research distribute workloads. Arelion activates 400G EVPL and wavelengths in named corridors, and AI Direct becomes a measurable source of traffic and customers.

Traffic grows while margins narrow

AS1299 moves more bits, but the price drops faster. Arelion remains important and leans more on cloud, security, Ethernet, and optics.

Security convergence improves retention

Customers buy transit or DIA with mitigation, giving SecureConnect a larger role in renewals. Success would be seen in onboarding, references, and availability during attacks.

Hyperscalers internalise more transport

Clouds carry more traffic on their own networks and reduce wholesale demand. Arelion keeps value in neutral paths, but the economics become selective.

A correlated route, facility, or capital shock reveals concentration

A peering change, cut, outage, or delay affects several services. The outcome depends on diversity, reserves, transparency, and owner capital.

Professional implications by stakeholder

Operators should test policy, communities, and multihoming. Enterprises should map direct and partner segments. Cloud teams should define the boundary with the on‑ramp. Security should separate observed volume from contracted mitigation. Investors should link expansion with utilisation and returns. Regulators should examine facility, cable, and AS concentration without assuming that ownership is the only form of control.

Control, incentives, and decisions that will shape the next decade of AS1299

The leadership problem is not choosing between network and business. The network is the business, and every technical decision locks in capital, supplier, and customer obligations for years. The control map is distributed: Polhem controls ownership and investment; leadership allocates capital; engineering and operations change routes; peers determine relationships; facilities and partners control physical segments; customers control prefixes and demand; clouds control the other side of the on‑ramp. Strategy works when these authorities align.

Decision one: fund physical diversity before selling reach

Arelion must measure diversity by cable, duct, station, facility, and power, not just logical topology. A PoP can increase apparent reach and share a hidden domain. The irreversible risk is committing to infrastructure that does not deliver the resilience sold.

The decision is to turn diversity evidence into a capital and design criterion. It can increase cost or delay launches, but it reduces multi‑product failures and protects credibility. A carrier with several forms of physical control can be trusted if it demonstrates domain control.

Decision two: preserve interconnection neutrality while monetising reach

The advantage depends on peers and customers trusting AS1299. Product growth must not distort policies. Commercial pressure can favour aggressive preferences, large accounts, or under‑investment in less profitable but necessary routes.

Leadership must separate peer policy, transit, security, and product engineering. The second‑order effect is diagnostics and trust; the third is losing direct relationships that cannot be recovered by buying fibre.

Decision three: demand activation proof for AI

AI Direct gives a credible language, but it can become a name for ordinary circuits. It should be governed by per‑route availability, active ports, use cases, utilisation, protection, and renewal.

That way, capital does not follow enthusiasm without evidence. It also distinguishes datasets, inference, replication, and cloud. The irreversible risk is overbuilding capacity based on failed forecasts.

Decision four: turn security evidence into operational feedback

The DDoS report offers a valuable view. It should be connected with planning, architecture, scrubbing, false positives, and learning. Publishing only the peak rewards spectacle; publishing mechanisms and results improves the service.

A better loop would distinguish observed, mitigated, impact, and blackholing. It would help customers and the owner. Without that evidence, SecureConnect risks becoming a commodity label.

Decision five: make the partner reach governable

Partners are necessary, but responsibility fragments. Arelion should treat selection, inventory, incidents, diversity, and exit as architecture. The customer must know what is direct, managed, or coordinated.

The second‑order benefit of good governance is a wider market without equivalent build. The third‑order cost of poor governance is that a local provider damages a global reputation. Failures can be outside physical control but not outside evidence and escalation control.

Decision six: disclose enough economics to sustain trust

Private ownership gives flexibility and opacity. Listed‑company reporting is not needed to publish coherent investment, activation, resilience, and sustainability without revealing contracts.

More disclosure would make routes meaningful and prevent connectivity from being used as a financial proxy. It would also discipline capital. The risk of opacity is that counterparties apply harsher assumptions under stress.

Second‑order effects of successful execution

If Arelion turns patient capital into diversity, keeps trust, and sells higher‑value services, AS1299 becomes a neutral WAN platform for clouds, enterprises, carriers, and AI. It improves retention, predictability, and the security justification.

Success also benefits partners, data centres, clouds, IXs, and customers with an alternative to concentration on a single hyperscaler. The benefit is distributed, and no single actor controls everything.

Third‑order effects of failure

A bad policy or hidden concentration can affect several products. Customers reroute, peers review trust, security loses credibility, and the owner needs capital just as revenues are at risk. Connected infrastructure transmits reputation as fast as packets.

A prolonged constraint delays optics and PoPs, lengthens paths, and increases dependency. Once relationships and accounts migrate, recovering them takes years.

Irreversible risks

The hardest risks are loss of settlement‑free peering, correlated routes, under‑investment in one generation, dependency on one segment, and an incident that exposes weak controls. A brand can be repaired; a routing position or a long contract can persist.

Leadership must separate reversible experiments from commitments. A portal feature changes; a fibre right or an optical architecture binds for years. A promotion ends; a damaged peering relationship may not return.

The leadership judgment

The strategic asset is not the brand or the hostname alone. It is the system behind AS1299: interconnection, routes, optics, policy, security, operations, and trust. Polhem and leadership inherit that system. They must renew it without breaking the trust built up under the previous identity.

The decisive test is disciplined conversion. Routes must become diverse paths; ports, traffic; AI positioning, services; DDoS visibility, availability; partners, governable delivery; and private capital, timely upgrades without hiding risk. When those conversions work, Tier‑1 keeps economic meaning. When they fail, the label outlasts the advantage.