Summary

  • Compnet's public record points to a Zelenograd-centred business access operator with real routed address space, business internet and telephony services, public-sector contracts and some industrial customer evidence. That is a workable niche, but it is not a scale story by itself.
  • The 2025 financial record is the warning sign. Public profiles report revenue around 47 million rubles and profit of only about 50,000 rubles, after much stronger reported profits in prior years. Until that margin break is explained, traffic growth should be treated as a cost test rather than a victory.
  • The network looks small and transit-sensitive: AS24680 announces four IPv4 prefixes, no visible IPv6 prefix, and current routing sources show a narrower observed upstream/peer picture than the RIPE policy object. Compnet's best strategy is selective B2B depth, not a price war with national and local alternatives.

Start with one customer circuit

The cleanest way to understand Compnet is not to begin with an autonomous system number or a list of prefixes. Begin with one office circuit in Zelenograd. A small manufacturer, laboratory supplier, trading company or public institution wants fixed internet, a telephone line, a static address, remote access, maybe a simple IT maintenance arrangement, and someone local enough to answer when the connection fails. Compnet can plausibly sell that package. The official site presents the company as a provider of business internet, telephone communication and IT maintenance, and the contact block separates subscriber service from technical support.

That is the surface of a local business operator.

The economic question is what remains after that customer is connected. Monthly revenue is not contribution margin. From the invoice, Compnet has to pay or absorb the access build, port, router or customer premises equipment, billing, support, compliance documentation, IP address administration, upstream transit, possible backhaul, and the labour cost of any visit that cannot be solved remotely. If the customer is already in a building or industrial area where Compnet has equipment and spare capacity, the incremental margin can be attractive.

If the circuit needs construction, custom equipment, repeated support visits or a heavy service-level promise, the same customer can look profitable in revenue and weak in cash.

That distinction is essential because the company is small by public financial evidence. Business registries report 2025 revenue around 47 million rubles. That is enough to support a niche operator, not enough to hide many bad circuits. A few support-heavy accounts, a compliance remediation cycle, a vendor bill, a customer loss or an upstream repricing can move the result. The public financial data show exactly that kind of fragility: revenue did not collapse in 2025, but reported profit nearly disappeared.

Compnet's history also argues for a careful circuit-level reading. Older provider profiles describe a Moscow telecommunications company on the market since 1991, offering traditional and IP telephony, internet access, leased channels and telecom equipment installation. Current Russian company records show LLC COMPNET registered in 2001 and state-registered in 2003 under OGRN 1037739121155 and INN 7735104182. RIPE still identifies the member and routed network as JS Company Compnet. The naming looks like a registry legacy that has lived through changes in legal presentation, website design and service mix.

The durable fact is not the corporate label. It is that Compnet has been a Zelenograd communications operator for a long time and still runs AS24680.

The right question, then, is not whether Compnet has traffic. It does. The right question is whether the next unit of traffic comes with enough paid service around it. For an access provider, traffic can be a cost more easily than an asset. A customer who doubles usage without paying for a higher-grade service consumes more backhaul and support capacity. A customer who buys a managed office circuit, static addressing, voice and maintenance creates a more defensible bundle. Contribution margin comes from the bundle, not from packets counted in the abstract.

Identity is real, but the public record is uneven

The identity trail is stronger than it first appears. RIPE's member page lists JS Company Compnet in Zelenograd, with a Moscow postal code, a technical contact email at compnet.ru and Russian service area. The RIPE organisation object gives ORG-JCS1-RIPE, country RU, reg-nr 1037739121155 and org-type LIR. The aut-num object for AS24680 names COMPNET-AS and describes it as intended for multihome routing of CompNet IP address space. That is not a marketing claim. It is resource-holder infrastructure evidence.

The live company site, however, speaks in the language of LLC COMPNET. Its personal-data offer identifies the operator as LLC COMPNET with the same OGRN and INN, a Zelenograd legal address on Yunosti Street, and Alexey G. Yushmanov as director. Business registries cross-check the same registration number, INN, principal activity in wired telecommunications, private ownership and small-company profile. TBank, RBC Companies, Companium, Querycom, SPARK-Interfax, RUPep and Zachestny Biznes all point to the same legal core, though they differ in rounding and the amount of public data exposed.

That unevenness matters. A buyer of a large circuit would want the current licence extract, current ownership records, current contract authority and current financial statements, not only RIPE objects and registry aggregators. But for an economic profile, the identity is sufficiently coherent: the routed AS, the company website and the Russian legal registry point to one operator centred on Zelenograd.

The ownership surface is simple. Public records identify a small private company with 10,000 rubles of charter capital and two owners, Oleg Ivashchenko at 51% and Ekaterina Kurnosova at 49%. The director is reported as Alexey Yushmanov. There is no evidence in the public materials of a large strategic owner, national carrier parent or deep balance-sheet sponsor. That means the operating company has to stand on its own cash generation. It cannot easily subsidize a broad network expansion from group capital unless there are private arrangements not visible in the record.

The service surface is also narrower than the word "regional" can imply. Compnet is not presented as a mass national broadband network. Its current site says business internet, telephone communication, IT maintenance and other business technology services. Zelenograd24 lists it under internet and telephony. Older ISP profiles put it in Moscow and Zelenograd, with business and institutional customer categories. The most concrete IP assignment found in RIPE search points to "Zelenograd South Industry" inside the 80.243.10.0/25 range.

That is a useful clue: Compnet's durable centre of gravity appears to be office, industrial and institutional connectivity around Zelenograd, not dispersed household access across Russia.

This is a more attractive market than undifferentiated residential broadband if the operator has local depth. Business customers can value static addressing, predictable support, voice continuity, direct contact, simple IT help and a provider that knows the building or industrial site. They may also churn less often than households if service is stable. The drawback is that each business account can carry special requirements. A small residential ISP can standardize around mass-market plans.

A business operator accumulates exceptions: one customer needs a static routed block, another needs voice, another needs weekend support, another needs documents for procurement, another needs a VPN or firewall change, and every one of them expects the provider to remember the history.

That is why the legal and operating record matters so much. A business-access operator can be high-margin when it is disciplined. It can become support-heavy when it says yes too often.

The 2025 margin break is the central fact

The most important public number is not the size of Compnet's address space. It is the profit collapse reported for 2025. TBank reports revenue of 47.15 million rubles and profit of 50,000 rubles. RBC Companies reports revenue of 47.157 million rubles, cost of sales of 46.121 million rubles and profit of 50,000 rubles. Companium reports revenue of 47.2 million rubles and net profit of 50,000 rubles, compared with reported revenue of 49.3 million rubles and net profit of 14.8 million rubles in 2024.

There are several possible explanations, and the public record does not prove which one is right. Costs could have risen. A large contract could have ended. Revenue mix could have shifted toward lower-margin resale. The company could have absorbed equipment or compliance expense. Accounting classification under a simplified tax regime can also make year-to-year comparisons less intuitive than they look. The responsible conclusion is not to invent a cause. It is to treat the margin break as a change-of-view fact until original filings and management explanation are available.

Even without a cause, the implication is clear. A company with roughly 47 million rubles of annual revenue and near-zero profit has little room for unpriced support labour, bad debt, installation overruns or traffic that requires more paid transit. If RBC's cost-of-sales figure is read literally, gross profit is only a little over 1 million rubles, about 2.2% of revenue. That number may not map cleanly to network gross margin, but it captures the pressure. Compnet cannot afford growth that arrives with an equal or larger rise in costs.

The contrast with prior years is striking. Companium's public table shows net profits around 14-15 million rubles in 2022, 2023 and 2024, against revenue in the 49-57 million ruble range. If those figures are comparable, Compnet had a much stronger earnings base before 2025. If they are not comparable, the public data still leave investors and customers without a reliable margin story. In either case, the article's judgement stays the same: the burden is on contribution margin, not traffic.

One useful scale calculation is revenue per announced IPv4 address. Compnet's four announced prefixes total 5,888 IPv4 addresses, or 23 /24 equivalents. Dividing 2025 revenue of about 47.157 million rubles by 5,888 addresses gives roughly 8,000 rubles of annual revenue per announced address. That is not a billing metric; many addresses may be unused, infrastructure-assigned, customer-assigned, or tied to services that are not priced by address. But it helps calibrate scale. Compnet is not monetizing an enormous network. It is operating a small address base where each business customer matters.

Public contracts reinforce the point. Companium reports seven government procurement contracts totalling 22.3 million rubles, including 44-FZ and 223-FZ work, and lists customers such as Rosoboronexport, GBU KRZ, AO NII Submicron, FSUE SKB Radel and KP KRZ. TBank reports four 44-FZ contracts, including local telephone service and internet-access items. These are valuable signals because they fit the business-access thesis. They also raise concentration questions.

Contract totals cannot be compared directly with one year's revenue without knowing timing, scope and recognition, but a 22.3 million ruble public-contract history is meaningful beside a 47 million ruble revenue base.

The strong version of the company is a sticky, relationship-based operator serving Zelenograd institutions and businesses with access, voice and IT support. The weak version is a small operator whose public-sector and business circuits are not large enough to absorb compliance, support, transit and equipment costs. The 2025 result makes the weak version too plausible to ignore.

Routing shows a small, real network

AS24680 is active. RIPEstat's overview identifies the holder as COMPNET-AS JS Company Compnet and shows announced status. The announced-prefixes endpoint shows 80.243.0.0/20, 185.211.224.0/22, 193.164.200.0/23 and 213.5.74.0/24. BGP.tools, IPinfo, Ipregistry, IPIP and 2ip all converge on the same broad scale: four IPv4 prefixes, 5,888 IPv4 addresses and no visible IPv6 prefix. That is a real network, but a small one.

The prefix composition is informative. The 80.243.0.0/20 block is the largest at 4,096 addresses and appears as the older core range. RIPE search for 80.243.0.1 shows an administrative/server assignment, COMPNET-TECH-0. Search for 80.243.10.1 shows a 128-address assigned PA block, COMPNET-INET-NIIMP, described as Zelenograd South Industry. The 185.211.224.0/22 allocation dates from 2017 in the registry record and is routed by AS24680. The 193.164.200.0/23 range is assigned PI and routed by AS24680. The 213.5.74.0/24 range is also routed by AS24680 but is identified as RU-S-TERRA in registry data and by several third-party prefix pages.

That looks like customer or hosted-address evidence rather than evidence of broad Compnet geography.

This address plan fits a small business operator that has accumulated blocks and customer assignments over time. It does not fit a consumer broadband network with hundreds of thousands of users. It also does not show an IPv6 transition story. No public routing source used here shows originated IPv6. In 2026 that is not fatal for a small Russian business ISP, because IPv4 still carries the commercial workload. But no IPv6 is a strategic weakness. It limits modern network posture, reduces future flexibility and leaves the company more dependent on scarce IPv4 management.

Transit and peering evidence is mixed in a way that should make a reader cautious. The RIPE aut-num policy object lists imports from AS9002, AS25478, AS50384, AS50817 and several smaller ASNs, and exports to the same groups. That policy object describes a multi-homing intention. Current observed sources are narrower. RIPEstat neighbours returned AS12722 in the check used here. BGP.tools reports one upstream, AS12722 RECONN LLC, and two peers, AS12722 and AS35598 INETCOM CARRIER LLC. IPinfo and Ipregistry also show one upstream. CAIDA's AS Rank API infers a broader degree, with providers and peers, but no customers.

The safest reading is that the registered policy is broader than the most visible current routing observations.

That matters for economics. If Compnet is effectively dependent on one current upstream for much of its reachable traffic, then transit price, path quality and outage exposure matter more than the policy object suggests. If the broader policy relationships are active but not visible in every dataset, then the network is more resilient than the simplest current observation. Either way, the operator does not appear to have a large peering fabric that can change its cost structure. It is not a downstream-heavy transit provider. It is an access and small hosting network buying reach.

The traffic profile also supports the access reading. IPinfo labels the activity pattern as a consumer or eyeball-like day/night rhythm, but the hosted-domain count is small and the official product surface is business-heavy. That does not mean the customers are households. Office networks also sleep at night. What matters is that the network is not showing the shape of a major hosting platform, content network or wholesale transit AS. Its value must come from local customers and service relationships.

The absence of downstream networks is particularly important. A network with paying downstream ISPs can turn routing into wholesale revenue. Compnet's public data do not show that. Its announced prefixes and customer assignments are useful, but they are not proof of transit revenue. Therefore peering and transit should be analysed as cost and resilience tools, not as a separate profit engine.

Peering can help, but it will not rescue weak access economics

For a small ISP, peering is tempting because it appears to turn traffic growth into savings. If more customer traffic can be exchanged locally with content networks, carrier peers or route servers, paid transit falls and latency improves. That is true in principle. But peering has its own fixed costs: ports, cross-connects, router capacity, engineering time, monitoring, filtering, route policy maintenance and fault isolation. A provider has to reach enough traffic scale to justify the operational overhead.

Compnet's current public routing evidence does not prove that it has reached that threshold. The RIPE policy object lists several relationships, including large or familiar Russian-connected networks, but current third-party observations point most visibly to RECONN and INETCOM. BGP.tools shows four originated IPv4 prefixes, no IPv6, one upstream and two peers. Ipregistry says no direct peering agreements and no downstreams. That is not a condemnation. It is a reminder that a route policy line and an economic peering position are different things.

The contribution-margin test should be applied to interconnection the same way it is applied to a customer circuit. If a peering port reduces paid transit by more than it costs in port fees, equipment depreciation, cross-connects and engineering time, it improves margin. If it is added for prestige or because traffic is growing without higher revenue, it can become another fixed cost. For a company with 2025 profit reported at only 50,000 rubles, even modest fixed costs matter.

The most valuable interconnection move for Compnet would be narrow and measurable: reduce the cost or improve the resilience of traffic serving paying business customers. That could mean a better upstream contract, a second physical path, private interconnect with a critical partner, or a local exchange route that materially improves application performance for customers. It does not need to look like a national peering programme. The company should not chase broad routing complexity unless traffic volume and customer requirements pay for it.

Observed low-latency pings from Moscow-based measurement points show that parts of the network are close and reachable, which is what one would expect for a Zelenograd/Moscow operator. Those measurements do not prove user experience across all customers. They do show that Compnet is not a dormant AS. It has live addresses, reachable routers and business-useable locality. The open question is utilization: how many of those addresses are tied to revenue, what share of peak traffic is paid for by business-grade contracts, and how much headroom exists before another upstream or router upgrade is needed?

Peering is therefore a lever, not the thesis. The thesis is service density around customers that pay for reliability. If Compnet can use peering to protect those customers and lower unit cost, it should. If traffic growth mainly comes from low-price users or unmanaged hosting without service premiums, more packets will not improve the company.

Deployment advantage is local, not national

Compnet's strongest deployment advantage is likely local familiarity. The public address trail repeatedly points to Zelenograd. The current legal address is in the Savyolki municipal district. The old provider listing uses Zelenograd's south industrial zone. The RIPE contact address is another Zelenograd location. The specific 80.243.10.0/25 assignment to Zelenograd South Industry is a rare concrete clue inside the network record. This is the footprint of an operator embedded in a local business geography.

That kind of footprint can matter more than national scale for a particular customer. If Compnet already has access to a building, industrial site or business centre, the cost of another circuit can be low. If it knows the customer's voice setup and static addressing history, support can be faster. If it can combine internet, telephony and IT maintenance, the customer may prefer continuity over a cheaper generic offer. Local presence can also help in procurement where a customer values known contacts and quick escalation.

The same localism limits growth. A national operator can spread product development, security services, portals, customer care systems and procurement across a huge base. MTS Business markets Moscow-region internet with 24/7 support, 99.9% SLA language and national reach across 84 regions. Beeline Business presents mobile, office internet, telephony, cloud, security and managed IT services as a broad bundle. Rostelecom business channels present fibre access, technical support and a national network story. Local providers such as ZelTelecom and Micron-Media also compete in Zelenograd access.

JustConnect advertises 21 providers in the city and office internet from 1,000 rubles.

Compnet should not try to beat all of that on breadth. It cannot outbundle the largest carriers, and it should not compete with residential aggregators on the lowest monthly price. The defensible lane is narrower: business customers that want static addresses, voice continuity, direct support, existing local route knowledge, and possibly custom IT work. These customers do not always choose the cheapest provider. They choose the provider least likely to make their specific office problem worse.

That lane still requires discipline. Installation must be priced honestly. If a building needs new access work, the payback period should be clear. If a customer asks for special support, the monthly fee should reflect it. If a customer wants a routed block or unusual configuration, the operator should understand the opportunity cost of IPv4 space and engineering time. Small operators often lose margin because they treat every custom request as customer service rather than as a product with cost.

The deployment question also includes replacement capital. Public records show fixed assets around 3.6 million rubles in 2025 in one registry profile, but that does not reveal the true network replacement cycle. Routers, switches, optical modules, UPS units, customer premises devices and cabling all age. Even if Compnet's capex is modest compared with a fibre-builder, replacement still has to come from cash flow. Near-zero profit gives little comfort. A company can appear stable until a router refresh, compliance equipment purchase or building access project consumes a year's earnings.

Support labour is the hidden cost centre

Support is where small business ISPs either defend their margin or lose it quietly. Compnet's official contact block gives separate subscriber-service and technical-support phone numbers and emails. That is a positive sign because it shows an explicit support surface. It is also a reminder that support is not free. Someone has to answer, triage, document, escalate, dispatch, close and invoice.

Business customers create support complexity that ordinary consumer metrics can miss. They have staff lists, office moves, router changes, static IP requests, phone-number issues, payment documents, procurement forms and after-hours emergencies. The official personal-data offer is unusually informative here because it describes the subscriber relationship as a communications-service contract with user-list obligations. For legal entities, the provider may need to receive and protect data about representatives, employees and users. That is administrative work tied directly to the service.

Regulatory support is another cost. The December 2023 Moscow Arbitration Court decision shows that Roskomnadzor brought a case against LLC COMPNET over communications licence requirements connected with operational-search measures. The court decision states that COMPNET provided communications services under multiple licences and had not implemented required measures in the required way. The fine, after small-business treatment, was only 4,000 rubles. The fine amount is not the main issue. The main issue is that compliance itself can require technical systems, documentation, coordination and management time.

There is also an unresolved public-record inconsistency around licences. TBank's profile labels five communications licences as active, while its event history reports February 2025 changes saying several licensed activities were suspended. Querycom lists multiple communications licences. Without a current original extract, the exact status cannot be settled here. For the economic analysis, the inconsistency is a risk signal. A business operator selling internet and telephony cannot treat licence clarity as clerical. Licence uncertainty affects procurement, customer trust and the ability to maintain services.

Support economics should also be read through the customer-review record, but carefully. 2ip shows a small number of reviews and speed measurements, including positive office-use comments and an older complaint about interruptions. These are not statistically reliable. They are still useful as a reminder of what customers actually buy: working service, stable support and a provider that does not blame the upstream indefinitely. In a small local market, a few bad support experiences can damage word of mouth. A few good business relationships can persist for years.

The practical margin rule is simple. Compnet should know the support cost of each customer class. A simple static office circuit in an already-served building is different from a customer with voice, unusual routing, compliance documents and frequent handholding. If both are priced as ordinary access, the second account is subsidized by the first. That is how a company can maintain revenue while profit vanishes.

Customers are valuable, but concentration is probable

The public evidence points to a B2B and institutional customer base. The current site talks to businesses. Historical profiles mention state and commercial enterprises, banks, trading companies, business centres, scientific and manufacturing firms. Companium reports public contracts and top public customers. The RIPE address assignment to Zelenograd South Industry is consistent with industrial service. These are good markets if the contracts are sticky and priced correctly.

They also create concentration risk. A company with about 47 million rubles of revenue can be materially affected by a small number of customers. If one public contract rolls off, if one industrial customer switches to a national carrier, or if one business centre renegotiates, the effect can be visible. The public records do not disclose customer concentration by revenue, so the article cannot quantify it. But contract totals and the small revenue base make the risk plausible.

Public-sector contracts can be especially double-edged. They are credible and can reduce churn if the service is embedded. They can also carry procurement pressure, documentation burden, delayed payment and compliance obligations. TBank lists local telephone and internet-access contract items, while Companium lists larger 223-FZ and 44-FZ totals. That is valuable recurring demand only if the margin is adequate. A contract that keeps technicians busy but leaves little profit is not strategic.

The best customer for Compnet is not necessarily the largest by bandwidth. It is the customer whose revenue stack is broad and whose support profile is predictable: internet, static addressing, voice, maybe IT maintenance, maybe a managed device, all in a location where Compnet already has plant. The worst customer is a custom low-price account far from existing infrastructure, with high support expectations and little willingness to pay for service quality.

The company should also be careful with address-space monetization. IPv4 addresses are valuable, and Compnet has a modest but useful pool. Selling or leasing address-heavy services can create revenue, but it can also attract abuse, support tickets and reputation issues if customer vetting is weak. CleanTalk and IPinfo-style pages provide security-noise signals around prefixes, though not proof of systemic issues. For a small operator, abuse handling can become another hidden labour cost. A few bad customers can consume more NOC time than their monthly fees justify.

There is a constructive version of customer growth. Compnet can deepen existing business accounts, add managed Wi-Fi, voice, firewall or IT maintenance, and secure multi-year public or industrial contracts where the cost-to-serve is known. That is better than chasing raw subscriber count. In this niche, fewer better customers can produce more margin than many cheap ones.

Alternatives discipline every price

Compnet operates in a competitive access market. Zelenograd is not an isolated northern settlement with no substitutes. Local and national alternatives are visible. Yandex Maps lists multiple nearby providers. SpeedGEO's Zelenograd statistics show MGTS, MegaFon, Tele2, Lovitel, Beeline and Akado among measured providers. ZelTelecom markets fibre service in and around Zelenograd and says it has been building cable internet since 2009. JustConnect presents a crowded provider list and low entry prices. Zelenograd24 lists Compnet alongside other local internet-provider names.

For business customers, the largest alternatives are more threatening than residential price lists. MTS, Beeline and Rostelecom can offer business internet, mobile service, voice, cloud, VPN, DDoS protection, customer portals and national coverage. MTS markets 24/7 support and SLA language. MTS also announced online self-service ordering for high-speed business internet for small and medium businesses, reducing one of the advantages smaller operators historically had: easier buying. Beeline Business presents broad corporate communications and IT services. Rostelecom can lean on national backbone and public-sector familiarity.

Compnet's response cannot be "we also provide internet." That is not differentiated. The response has to be local specificity. It can know a building better, fix a static-routing issue faster, maintain old voice arrangements, or give a business customer a direct path to someone technical. It can serve customers that are too small to matter to a national account team but too complex for a commodity residential provider. That is a real niche.

But the niche imposes pricing discipline. If Compnet charges a premium, the customer must feel the difference in continuity and support. If it matches national or aggregator prices, it needs lower cost-to-serve. If it bundles IT maintenance, the labour must be priced. If it provides voice, compliance and numbering work must be accounted for. The market will not pay Compnet for being local unless local means easier operations for the customer.

Mobile substitution also matters. For some small offices, 4G or 5G backup is good enough for resilience. For point-of-sale terminals, kiosks or temporary sites, mobile business internet can substitute for fixed access. MTS and Beeline both market business mobile and internet bundles. Fixed fibre still matters for stable office operations, large files, predictable latency and voice integration, but the minimum acceptable service has risen. A small office that once needed a local fixed ISP may now have more workable alternatives.

The existence of alternatives is not fatal. It is what forces Compnet to be precise. The customers it should want are those for whom a generic national offer is inconvenient, slow, insufficiently tailored or poorly supported. The customers it should avoid are those who see connectivity as a commodity and will switch for a small discount.

Capex is probably selective, but the margin for mistakes is thin

There is no public evidence that Compnet is undertaking a large regional fibre build. Its scale, address space and financial record point to selective access and customer-driven deployment. That is sensible. A small operator should not build ahead of demand unless it has anchor contracts. Capex should follow committed customers, not hope.

The danger is that even selective capex can hurt a small income statement. A new building switch, optical handoff, router, UPS, fibre extension or set of customer premises devices may look modest, but the payback depends on retention and support. If a customer leaves after a short term, the payback breaks. If equipment needs replacement during a low-profit year, cash tightens. If compliance requires new systems, capex becomes unavoidable rather than growth-oriented.

Public fixed-asset figures around 3.6 million rubles suggest a modest tangible asset base, but they do not reveal leased infrastructure, fully depreciated equipment, customer-owned devices, rented facilities, or off-balance-sheet arrangements. Therefore the article cannot assert low capex. It can assert that capex visibility is poor and that replacement capital is an important unknown.

The routed network creates another form of capital: technical credibility. Routers, filters, monitoring, route objects, abuse handling and address management all require competence. The RIPE Database object was updated in July 2025, and organisation details were modified in May 2026, showing that the registry presence is maintained. That is positive. But active registry maintenance does not answer whether the company has enough engineering depth for growth.

The right capex strategy is boring and strict. Build where there is contract duration. Replace equipment before failure where the customer margin justifies it. Avoid custom lateral builds without installation fees or term commitments. Treat IPv4 space as scarce. Treat compliance equipment and documentation as part of the service cost. Do not count a new circuit as won until installation, support and replacement economics are attached to it.

The judgement

Compnet is economically defensible as a focused Zelenograd/Moscow business-access operator. It is not defensible as a traffic-growth story without proof of margin recovery. The company has real assets: AS24680, 5,888 routed IPv4 addresses, a long local history, business-internet and telephony positioning, public-contract evidence, and signs of industrial-area deployment. Those facts make it more substantial than a bare listing.

The weak facts are just as important. Public records show 2025 profit of only about 50,000 rubles on roughly 47 million rubles of revenue. Current routing observations show a small network with no visible IPv6 and limited observed upstream/peer breadth. Licence records contain an unresolved inconsistency, and a 2023 court decision confirms a compliance failure that, while small in fine amount, points to real operating obligations. The market around Zelenograd includes national carriers and local providers. Customers have alternatives.

The answer to the assignment's core question is conditional. Business and access revenue can outrun transit, deployment, support and replacement capital only if Compnet stays selective. A business circuit in an already-served building, sold with voice, static addressing and manageable support, can produce strong contribution. A custom low-price circuit with construction, handholding, compliance burden and heavy traffic can destroy margin. The company has to know the difference before it connects the customer.

Peering and transit should be managed for resilience and cost, but they are secondary. If Compnet can reduce paid transit or improve customer latency through a measured interconnection move, it should. But peering will not fix weak pricing, support-heavy accounts or a customer base that buys traffic without paying for service. The routing footprint is evidence of capability; the income statement is evidence that capability must be monetized carefully.

The most attractive path is depth, not breadth. Compnet should deepen sticky business and institutional accounts, price installation and special support explicitly, use its IPv4 resources carefully, clarify licence status, and prove that the 2025 margin break was temporary or explainable. It should avoid a residential price race and avoid traffic growth that arrives without higher-grade service revenue.

In short, Compnet's contribution margin is the company. Routed traffic is only the raw material.

What would change the view

Several facts would improve the judgement. The first would be original 2025 financial statements or management commentary showing that the near-zero profit was caused by a one-off expense rather than structural margin pressure. The second would be 2026 interim data showing revenue stability with margin recovery. The third would be customer-concentration evidence showing that no small group of contracts can destabilize revenue.

Network facts could also improve the view. Active multi-homing on physically diverse routes, measurable transit-cost reductions, private peering tied to customer traffic, IPv6 deployment, and documented peak-traffic headroom would make traffic growth less threatening. Evidence that AS24680's broader RIPE policy relationships are economically active, not merely registered, would support a stronger resilience case.

Customer and support facts would matter most. Low churn among business customers, multi-year public or industrial contracts, short mean time to repair, low repeat-ticket rates, and clear pricing for installation and custom support would show that Compnet's local model is disciplined. A larger managed-service share of revenue would also help, provided the labour is priced.

Several facts would worsen the view. If 2026 profit remains near zero, the business is not merely having a bad year; it is structurally squeezed. If licence suspensions are current and material, procurement and service continuity risk rises. If a major public or industrial customer is lost, concentration risk becomes visible. If traffic growth requires more paid transit while ARPU stays flat, the network will be busier but not healthier. If national carriers undercut business access in Compnet's core buildings, local advantage weakens.

The current public record leaves many operating metrics unavailable: EBITDA, capex, free cash flow, customer count, churn, traffic volume, transit cost and support load. Those gaps do not make the company unknowable. They define the investment question. Compnet can be a good local operator only if the missing numbers prove that each customer relationship carries its own cost.

Sources