A renewal meeting in Helmond prices two products

Somewhere on a Brabant industrial estate this quarter, a transport or food-processing firm of perhaps eighty desks is renewing its IT contract, and the decision on the table is not really about hours of support. Its managed-services provider has offered two columns. In one, the firm's twenty virtual machines stay where they are, in a datacenter at Schootense Dreef 26 in Helmond, reached over the provider's own fiber, backed up nightly to a second hall in Veghel twenty kilometres away. In the other column, the same workloads move to Microsoft's West Europe region, where the published price sheet lists a modest two-core, eight-gigabyte Linux machine at €0.1009 per hour — about €74 a month before storage, licences or traffic — and the Windows version of the same machine at €0.1817 an hour, roughly €133. The provider will manage either column. But the columns are not the same business. In the first, the provider keeps the rack margin, the connectivity margin, the backup margin and the labour. In the second, it keeps the labour and a reseller's sliver of someone else's cloud.

Read the small print of the two columns and the asymmetry deepens. The Azure column's compute line is only the opening bid: managed disks are metered separately, outbound traffic is metered, backup vaults are metered, and the Windows licence either rides the hourly rate at nearly double the Linux price or has to be brought along under a hybrid-use arrangement that itself requires software assurance the firm may not have. The house-cloud column hides different arithmetic: the price per virtual machine is not published anywhere — it emerges from a negotiation in a meeting room in Helmond — but it arrives flat, bundled with the backup, the circuit and the phone number of an engineer whose surname the finance director knows. One column is a tariff; the other is a relationship with a tariff inside it. Most of the Dutch mid-market has spent two decades quietly preferring the second, and the interesting question is not whether that preference exists but what it is worth per month, per rack and per year of contract.

That renewal decision, repeated across a few hundred contracts, is the entire economic question hanging over e-Quest IT Diensten, the Helmond company whose datacenter sits at the end of that first column. e-Quest is the purest available Dutch specimen of the attach-rate business: an IT services firm that built or bought every layer underneath its own labour — trenches, glass, racks, a routed network, even a solar field on a former landfill — so that each hour of support it sells drags infrastructure revenue in behind it. The claim this article tests is simple: the attach is where the margin durability lives, the labour is where the relationship lives, and the model's ceiling is set by how long Dutch SMEs will pay local-iron prices for sovereignty they could approximate, more cheaply at list price, in Microsoft's Amsterdam-region halls.

One address, seven registrations

Before the economics, the identity, because at Schootense Dreef 26 the registry is crowded. The Dutch chamber of commerce records a cluster of private companies under the e-Quest name, and reading them in sequence tells the corporate story more honestly than the marketing does. The oldest operating entity, e-Quest IT Diensten B.V., KVK 17102349, was registered on 19 February 1998 and traded for years as E-Quest Datacenter Diensten B.V. before taking its current name around February 2018. Its sibling, e-Quest IT Projecten B.V., KVK 17121883, dates from 18 February 2000, carries the trade name e-Quest Automatisering, and emerged in its present shape from a June 2018 demerger that also spun out an EQ Auto B.V. Above them sits e-Quest Groep B.V. (KVK 71977023, formed in 2018) and, at the top, Patrique Dankers Beheer B.V., the personal holding of the founder-owner whose name it carries. A dissolved E-Quest glasvezel diensten B.V. from 1996 lingers in the record as a fossil of an earlier structure.

The register that matters most for this piece, though, is the telecom one. The Autoriteit Consument en Markt's public register of telecom providers, downloaded as a spreadsheet on 3 July 2026, lists five group companies. e-Quest IT Diensten B.V. has been registered since 18 April 2018 as a provider of both a public electronic communications network and a public service. e-Quest IT Projecten B.V. has carried the same double registration since 14 February 2011. Breedband Helmond B.V. (KVK 17133171), the group's retail internet brand, has been a registered network operator since 3 November 2006. e-Quest Glasvezelnetwerk B.V. was added on 12 February 2020, and Glasvezel Helmond B.V. — the joint venture with the municipality, of which more later — on 23 November 2018. Whatever the group's brochures say, the regulator's ledger says this is not an IT consultancy with a server room. It is a small telecom operator with a consultancy attached, and it has been one on paper for two decades.

The routing system agrees, and adds a wrinkle the corporate register does not show. Autonomous system 42707 is registered in the RIPE database as EQuest-AS, described as e-Quest IT Diensten, and on 3 July 2026 it announced 27 IPv4 prefixes and two IPv6 /32s to the global table. Its PeeringDB entry still carries "Breedband Helmond B.V." as an alternative name — a fossil of the years when the network's public face was the consumer fiber brand rather than the IT company, and a reminder that in this group the ISP came first and the cloud grew on top of it, not the other way round. In December 2025 the company took the further step of becoming its own RIPE local internet registry under the organisation name e-Quest IT-diensten B.V. — note the third spelling variant of the same firm across three registries, each honest, none identical. LIR membership is a mildly expensive piece of administrative sovereignty that small resellers do not bother with, and a useful signal that the network is treated as a permanent asset rather than a legacy. One loose end is worth recording: the group's public financials are effectively dark. The KVK sells extracts behind a paywall; the group's companies file only the abbreviated balance sheets Dutch law demands of small entities (the latest, for financial year 2025, was deposited on 4 May 2026 per the aggregated registry record); and no revenue figure for any e-Quest entity appears anywhere in the public record. Every margin statement in this article therefore has to be built from tariffs and physical facts, not from accounts, and is flagged accordingly.

What the group actually sells, and who signs

The service catalogue has been remarkably stable for fifteen years; only the labels have moved upmarket. The archived 2010 homepage already shows the three pillars — office automation, fiber, datacenter — with shared hosting, colocation, backup and anti-spam filtering underneath, and Microsoft, HP and Cisco partner badges on top. By the 2013 capture the datacenter column had grown virtual private servers, a hosted workspace product and hosted Exchange; the fiber column had dark fiber, point-to-point circuits and voice. The current site sells the same stack under 2026 vocabulary: datacenter services, cloud, storage, cybersecurity, fiber connectivity and the modern workplace, wrapped in a data-sovereignty pitch — customer data stays in the company's own Dutch datacenters, inside European rules, with the Dutch data-protection regime and the EU's NIS2 security directive doing the emotional work that "uptime" did a decade ago. The Dutch Data Center Association's member page describes the firm's evolution since 2000 as regional IT provider to managed-security provider, which is the industry's polite way of saying the same customers now buy compliance as well as continuity.

Who pays is visible at the edges of the record. The company's reference list names the Elkerliek hospital in Helmond — which takes fiber circuits, colocation, cloud and office automation from the group and is quoted as "bijzonder tevreden" with stability and performance — alongside SPIE, the logistics firm Van den Broek, the food business Van Rijsingen, the workwear group Lavans and a string of other regional mid-market names. A hospital is the perfect attach customer: regulated, risk-averse, local, and legally obliged under NIS2 to know exactly where its data sits and who touches it. It is also the reference that does the most selling, because the implicit syllogism — if it is good enough for the hospital's patient flows, it is good enough for your invoices — closes deals across every business park in the Peel region. The rest of the roster is classic Brabant mid-market — firms large enough to need real infrastructure, small enough not to staff it themselves. Note the contract shape the references describe: never one product, always a braid. Elkerliek's listing alone spans four of the six service lines, which means four renewal dates, four switching costs and four reasons the account survives any single bad quarter. On the labour side the group reports "70 colleagues" on its own about-page and around 90 via the Dutch Cloud Community and trade interviews, spread across Helmond, Veghel and a Venlo office; the spread itself is a small signal, discussed later.

There is also a third, quieter customer class: other providers. The routing data shows two downstream autonomous systems buying transit from AS42707. One is Intention B.V., an IT firm in Uden; the other is Esmero Holding, behind a Dutch hosting brand that has run for over twenty-five years. Both are, functionally, smaller versions of e-Quest buying wholesale from it. When a firm's competitors become its customers, the infrastructure has crossed from cost centre to product — which is precisely the attach thesis in miniature.

The iron under the invoice

The physical layer accreted in datable steps, each one visible in a different corner of the record. The company entered the datacenter business by buying a small facility in 2006, according to a 2023 trade interview with its business unit manager. In 2012 it built its own hall in Helmond — 110 racks by an early sponsor-page description, 120 by the count in a 2014 supplier announcement in which the integrator PQR delivered a FlexPod stack of Cisco servers and NetApp storage to carry the house cloud's Exchange, SQL and virtual-desktop workloads. The NetApp relationship survives in the vendor's current partner listing. In early 2022 came the second site: two halls and 120 racks in Veghel, sold as a twin-datacenter proposition — synchronous mirroring across the group's own glass between the towns, which converts the second site from redundancy expense into a billable continuity product.

The fiber layer came before either hall mattered much. By 2013, a sponsor profile from the local football club — often the most candid corporate biography a Brabant family firm ever publishes — described an own fiber network already lit across the industrial estates of Helmond, Deurne, Laarbeek, Gemert, Asten and Someren, feeding a 110-rack datacenter. That geography is the attach map in physical form: business parks, not living rooms, chosen a decade before the municipal network existed, in exactly the places where a fiber circuit converts into a managed-services conversation.

The network wrapped around those halls is small but seriously built. PeeringDB records AS42707 with two 10-gigabit ports at AMS-IX, two at NL-ix, and one each at Frys-IX and Speed-IX — 60 gigabits of public peering capacity — plus presence in interconnection facilities in Eindhoven, Hilversum and the NIKHEF building in Amsterdam, one of the oldest interconnection sites in the country. Self-reported traffic is five to ten gigabits, mostly inbound, which is what an eyeball-and-hosting network of a few tens of thousands of endpoints should look like; the six-fold headroom between traffic and port capacity is bought deliberately, because for a provider whose pitch is continuity, congestion at an exchange is a reputational event, and exchange ports are the cheapest insurance in networking. Transit comes from Cogent and the Open Peering aggregation service, with Hurricane Electric among the peers. None of this is remarkable to a network engineer, and that is the point: it is the standard kit of a real, independent, mid-size Dutch operator, of which perhaps a few dozen remain after a decade of consolidation.

The most distinctive asset is the power arrangement. Since 10 May 2019 the group has operated 1,667 solar panels on a capped former landfill beside the Helmond site, generating roughly 475,000 kWh a year — around 60 per cent of its Helmond electricity need, by its own published figure. The state-aid record shows a €597,303 SDE+ renewable-energy operating subsidy awarded to e-Quest IT Diensten B.V. in February 2018 — a figure visible in the aggregated registry record and single-sourced there — which is the standard Dutch instrument that makes such a field bankable. The cooling design is adiabatic — outside air and tap water, one kilowatt of electrical draw per 10,000 cubic metres of cooled air per hour, no refrigerants — and the offices heat on server exhaust rather than gas. For a facility of this size these are not gestures; they are a structural hedge on the input that dominates every datacenter's cost line.

The arithmetic of the attach

Now assemble the unit economics, keeping evidence and inference strictly apart. The evidence: Microsoft's published West Europe tariff prices the two-core, eight-gigabyte D2s v5 at €0.1009 an hour for Linux and €0.1817 for Windows — €73.66 and €132.64 a month at 730 hours, compute only, before disks, egress or backup. At the commodity end of the Dutch market, TransIP's published price list sells a four-core, eight-gigabyte virtual server with 100 GB of NVMe storage and unlimited traffic for €50 a month. That pair of published tariffs is the observable spread this whole business lives inside: on comparable memory, Dutch independent iron at list undercuts Azure's Linux list price by about a third (€50 against €73.66) while including the storage and traffic Azure bills separately, and undercuts the Windows price by more than 60 per cent before licensing arithmetic. e-Quest does not publish its own cloud prices — a diligence gap flagged here explicitly — but it sells into a market where those two poles are a phone call away, so its achievable price sits between them: above TransIP, because it bundles management, locality and a named engineer; below or near Azure list, because that is the alternative column on the renewal sheet.

The fiber side is fully observable. The group's retail brand publishes €45 a month for 500 megabits and €50 for a gigabit, with a €19 activation fee and a fee schedule that monetises every touch: €5 monthly for an extra media box, €39 for a relocation, €39 for a missed appointment, €1.50 for the privilege of a paper bill. A late-joining household pays €198 for a fiber connection on the open city network — a fee that prices the marginal dig, not the network, and tells you the passive layer was financed on rollout-window participation rather than per-connection revenue. The current promotion, up to six months of the basic package free on a two-year commitment, values customer acquisition at as much as €270 of forgone revenue per household, which on a €45 tariff implies the brand expects retail relationships to last well past the contract's second year. These are thin-margin utility prices — national fiber retail sits in the same €45-55 band — and they are not where the money is. Their function is different: every connected business premise is a pre-qualified lead for the columns that do carry margin.

The cost base can be bounded from published physical facts, with the inferences labelled as such. Sixty per cent of Helmond's electricity from a 475,000 kWh solar yield implies total Helmond consumption near 790,000 kWh a year — an average load around 90 kilowatts across a 110-plus-rack hall, its cooling and the adjoining offices. That is the first analytically interesting number in this piece, and it is an inference from two company-published figures, so treat it with care. If it is roughly right, the Helmond hall runs at a fraction of a kilowatt per rack on average — either many racks are lightly loaded, or a good share stand empty. Both readings say the same economic thing: the constraint on this business is not capacity but demand. A second hall of 120 racks in Veghel, opened into that demand picture in 2022, only sharpens it. The attach model's fixed costs — two buildings, 60 gigabits of exchange ports, transit contracts, the standing cost of round-the-clock cover — are already sunk. Each incremental SME that chooses the house-cloud column contributes at close to the full spread between its invoice and the marginal electricity it burns, and the solar field caps even that. This is why the attach is worth defending fiercely: the marginal economics of the next customer are extraordinary precisely because the average economics of the estate are, on the visible evidence, unspectacular.

What cannot be computed from public documents is the blended result — how much of group revenue is labour at consultancy margins versus infrastructure at utility-plus margins. No filing discloses it; the abbreviated balance sheets are silent; and the honest position is that the group's five-time appearance in the FD Gazellen growth rankings, noted on a regional employer profile, is the only public, third-party-validated indicator that the combined machine grows at all. That award requires sustained double-digit revenue growth, and it is the single strongest piece of soft evidence that the attach has been accretive rather than decorative.

Azure as the counterfactual, line by line

Return to the renewal meeting and move the line items one at a time, with the published numbers attached. Twenty Linux machines of the D2s v5 class cost €1,473 a month at Azure's West Europe list rate; the same memory footprint at TransIP's published Dutch commodity rate costs €1,000, storage and traffic included. If half the estate is Windows, the Azure column rises by roughly €59 per machine per month before any licence mitigation. Those are list prices, and real enterprises negotiate — but a sub-100-seat Brabant firm does not command hyperscaler discounts, so list is closer to its truth than to a multinational's. If the estate goes to Azure, the compute line goes to Microsoft at those rates; the MSP re-bills it, typically through the reseller channel, keeping a margin that Microsoft sets and has repeatedly trimmed. The storage, backup and disaster-recovery lines — in the house cloud, products running on NetApp shelves the MSP already owns — become Azure meters, again Microsoft's revenue. The connectivity line thins from a dedicated fiber circuit into the MSP's own datacenter to a generic internet or ExpressRoute path. The labour line survives — someone still patches, monitors and answers the phone — but it is now the only line the MSP truly owns, and labour is the line with the least pricing power, because every one of the hundreds of Dutch MSPs can supply it against the same Azure back end. The attach, in other words, is not a technical preference. It is the difference between owning four margin streams and owning one.

What holds the estate in the first column is a bundle of switching costs and a sovereignty premium, and both can be described precisely. The switching costs are the ordinary ones — re-platforming effort, new run-books, retraining — plus one that is specific to this configuration: the fiber. A customer whose premises are lit by the group's own glass, whose voice runs over it and whose backups traverse it, faces a coordinated exit across three utilities, not one cloud migration. The sovereignty premium is softer but currently strengthening. e-Quest's pitch — data "always within Dutch borders" in its own halls, under a Dutch legal entity, audited to ISO 27001, NIS2-ready — lands differently in 2026 than it did before Schrems II, the CLOUD Act debates and two years of European anxiety about American platform dependence. The premium is real but should not be over-read: Microsoft now operates an EU Data Boundary and Dutch-soil regions, so the gap the premium prices is legal-jurisdictional, not geographic. It survives exactly as long as a Brabant accountant, hospital compliance officer or works council believes a Helmond BV under Dutch law is meaningfully safer than a Redmond subsidiary's Amsterdam region. That belief is an asset with a decay rate, and nobody — including e-Quest — knows its half-life.

There is also a scale trigger in the counterfactual, and it defines where the attach flips from margin to millstone. The house cloud wins while workloads are steady-state — file servers, ERP, the hospital's document flows. The moment a customer needs what only hyperscale sells — burst compute, managed databases, the full Microsoft 365 gravitational field, AI services — the MSP must either bridge to Azure anyway (hybrid, the current diplomatic answer) or watch the workload leave. Every rack in Helmond and Veghel is a bet that the region's mid-market demand curve stays inside what 230-odd racks and a 10-gigabit traffic profile can serve. If Brabant's SMEs digitise faster than that envelope, the iron that made the margin becomes a fixed cost chasing a departing market — the millstone case. The visible utilisation inference above says that day is not close; it does not say it never comes.

The city network that feeds the funnel

The strangest and most instructive asset in the group is the one it shares with the town hall. Helmond, a working city of roughly ninety-five thousand people, was late to fiber — late enough that residents were still petitioning on the KPN customer forum about the absence. e-Quest had wired the profitable fringes itself: Breedband Helmond, ACM-registered as a network operator since November 2006, had built out the Brouwhuis, Rijpelberg and Stiphout districts and the industrial estates. In 2018 the municipality decided fiber was a utility "like sewerage", and its public brochure laid out the deal: the city would see fiber laid through every unserved street between 2019 and 2021, connections free for households joining during the rollout, with e-Quest as the private partner and the existing Breedband Helmond districts folded in. Glasvezel Helmond B.V., the joint vehicle, entered the ACM register that November.

The 2018 brochure repays close reading, because it quietly allocates every risk in the deal. Households connecting during their street's rollout window paid nothing; laggards would pay their own connection later — the €198 now on the tariff card. Existing e-Quest districts and the industrial estates would "possibly automatically" migrate into the open network, folding a private asset into a quasi-public one on terms never published. KPN's fiber island in the Brandevoort district was to be negotiated in, and the brochure frankly predicted that KPN and Ziggo would decline to retail over the open network at all, preferring their own copper and coax — a prediction that held, and that handed the open network's retail shelf to small national independents. Deployment proceeded neighbourhood by neighbourhood with a 25 per cent participation threshold, which converts civic enthusiasm directly into build-out sequencing.

The economics of that bargain deserve a closer look than the civic framing invites. The municipality got citywide coverage, including the countryside no commercial builder wanted, and an open network: five retail providers now sell over it, of which the group's own Breedband Helmond is merely one against TriNed, Freedom, PLINQ and FiberNL. e-Quest got something subtler than a retail monopoly — it declined to seek one, and could not have had one under the open-access terms. It got the operator position: co-ownership of the passive asset, the operations relationship, and a city in which every business park was now reachable on infrastructure it helps run. For the attach machine, the open network is not the product; it is the funnel's mouth. A firm that takes a €50-a-month gigabit line lands inside the group's addressable base, one upsell conversation away from a rack in Veghel and a managed firewall. The retail competition on the open network costs e-Quest consumer margin it never really banked on, while the wholesale-and-operations layer — unpriced in any public document, and flagged here as the largest single blank in this analysis — is where the network pays its keep.

What the market says when nobody is presenting

The unofficial record around the group is thin, local and worth reading closely precisely because nobody curates it. Breedband Helmond's Trustpilot profile holds a 3.5 score across just eleven reviews — an unsolicited profile, since the company does not invite reviews — split between decade-long customers praising the fact that a human in Helmond answers the phone, and angry accounts of billing for a 4K television box and outages that dragged. A single formal complaint sits on the Consumentenbond's complaint register under a title that translates as "poor service, empty promises, no solution offered". Eleven reviews and one complaint for a network passing tens of thousands of premises suggests either a small retail base or a quiet one; combined with the €45-50 tariffs and five-provider competition, the small-base reading fits better, and it supports the funnel interpretation — retail is bait, not the business.

The retail noise also has a historical shape worth noticing. Breedband Helmond's own origin story, told on its about-page, is that it was founded because incumbent providers kept breaking promises — "constant outages and broken commitments" — and Helmond deserved a network that stayed up. Twenty years later the sharpest complaints against it are the same genre of grievance, aimed the other way. That is not hypocrisy so much as gravity: every access provider eventually owns the outage narrative in its own service area, and a provider whose commercial identity is reliability pays a higher reputational price per incident than one that competes on price. For the business franchise this matters more than the star rating does, because the hospital, the logistics firm and the works council all live in the same small city as the eleven reviewers.

On the business side the signals point the other way. The FD Gazellen listings, five years running, are incompatible with a stagnant firm. The headcount told three ways — 70 on the company's own site, about 90 in its community and trade profiles, a 51-200 band on LinkedIn — reads less like obfuscation than like a company growing faster than its own web copy is updated, with the 2018 restructuring having scattered employment across group entities in a way that makes every public counter honest and none complete. Operational freshness shows in small places: the PeeringDB record's peering data was updated in late June 2026, the RIPE LIR membership is six months old, and the recruitment site advertises across all six service lines. What would settle the picture — churn, retail subscriber counts, the wholesale terms with the municipality — is exactly what never leaks from a private Brabant BV, and the honest reading of the silence is that nothing in the public exhaust contradicts the growth story, while nothing independently proves its scale either.

Where the attach could flip

The risks to this model divide cleanly into three prices moving against it. The first is the price of the counterfactual. Microsoft's list prices have been remarkably stable — the D2s v5 tariff above has held since late 2021 — but the hyperscalers compete increasingly on committed-use discounts, EU-sovereignty packaging and bundled security, each of which erodes a distinct plank of the local pitch. If Azure's effective Dutch mid-market price falls 20 per cent while its sovereignty story thickens, the renewal columns converge and the house cloud's premium has to be earned by labour alone. The second is the price of inputs. Forty per cent of the Helmond load still comes from the grid, Veghel cools conventionally, and transit and IX ports are contracted from third parties — Cogent and the Open Peering platform prominent among the visible upstreams — so the independence story has suppliers behind it like everyone else's. The third is the price of people. A 70-to-90-person firm running 24/7 infrastructure across three towns carries key-person concentration from the founder-owner down to the night-shift network engineer, in a Dutch labour market where every MSP and every hyperscaler is hiring the same profiles.

Competition arrives on every layer at once, which is the tax the full-stack model pays for its margins. At the connectivity layer, the open city network means TriNed, Freedom, PLINQ and FiberNL bid for every Helmond household that Breedband Helmond courts, while KPN and Ziggo defend the rest of the city on their own plant and Eurofiber prices the business circuits. At the cloud layer the poles are the ones already priced above — hyperscale list on one side, commodity Dutch VPS on the other — with dozens of Dutch hosting firms in between making the identical sovereignty argument from identical certifications. At the labour layer, the national MSP directory lists competitors by the hundred, and the nearest ones sit twenty minutes away in Eindhoven's far deeper talent pool. e-Quest's defence is not superiority on any single layer; it is that no local rival fields all the layers at once, and the ones that could — the national consolidators — have so far found Brabant's mid-market accounts too small to chase individually and too sticky to dislodge collectively.

Behind all three prices sits the consolidation question. The Dutch regional-datacenter and MSP landscape has spent a decade rolling up into national platforms, and a family-owned group with two ISO-certified halls, a city-network operations seat, its own address space and a hospital on the customer list is a textbook acquisition target. Nothing in the record suggests a sale is contemplated; everything in the record — the 2018 holding restructure, the clean entity separation of network, projects and services — suggests the structure is ready for one. Whether that is estate planning or exit planning is not knowable from outside, and it is the difference between this being a twenty-year hold and a three-year one.

Facts that would move the judgement

A handful of discoverable facts would materially reprice this analysis, and they are worth naming plainly. Publication of any revenue split between labour and infrastructure — even once, even rounded — would replace this article's central inference with arithmetic. The wholesale and operations contract between Glasvezel Helmond B.V. and the municipality, or any tender for its renewal, would price the group's most opaque asset. A departure or public procurement by the Elkerliek hospital would test whether the flagship attach relationship is priced on quality or on inertia. Evidence of rack utilisation — a colocation price list, a capacity announcement, a third hall, or conversely marketing that begins discounting — would resolve the demand question the solar arithmetic can only gesture at. A transfer of AS42707, its prefixes or the LIR membership would be the earliest public tell of a sale, visible in the registries weeks before any press release. And a Microsoft price action or EU sovereignty ruling that collapses the legal distance between a Helmond BV and an Amsterdam Azure region would attack the premium at its root. None of these has happened as of 3 July 2026; each has a public venue where it would first appear; and that, more than any conclusion, is the practical output of this piece — a short watch-list of registries for a company that files almost nothing.

Evidence register

The load-bearing sources for this article, and what each carries: