A monopoly with a hundred and fifty storefronts
On Wednesday this week, the wholesale price of a mid-tier broadband service in Australia changed for every provider in the country at the same instant, by the same amount, for the same reason. NBN Co's tariff schedule for the year beginning 1 July 2026 lifted the 50/20 Mbps input to $57.60 a month and set the popular 100/20 Mbps tier at exactly the same figure, both now flat-rate, both indexed to inflation at 3.63 per cent, both non-negotiable. No retailer bargained for those numbers. None could. The company that owns nearly every suburban access line in the country is a Commonwealth-owned monopoly, and its prices are set through a regulated undertaking, published to the cent, adjusted once a year.
Here is the paradox that makes Australian broadband interesting to an economist and maddening to everyone else. A country that spent tens of billions of public dollars renationalising its last mile — unwinding three decades of privatised copper by building over it — did not end up with one internet provider. It ended up with more than 150 retail brands, a number NBN Co itself cites in its pricing releases, all reselling the identical regulated input. The state built the pipe; the market was invited to fight over the mark-up.
The fight is not going well for the small combatants, and the reason is visible in two published numbers. The wholesale input is $57.60. The most closely watched independent retailer in the country, Aussie Broadband, lists its equivalent retail plan at $99 a month as of this week. Strip the 10 per cent goods and services tax that lives inside every Australian retail price and the provider keeps $90.00, of which $57.60 goes straight back to the state. The entire addressable economics of a consumer NBN reseller — network, staff, marketing, customer equipment, billing, bad debt, profit — must fit inside $32.40 per line per month. At the discount end of the market, where advertised prices start at $60.99, the ex-tax retention is $55.45: less than the wholesale input, a gap sustainable only for the length of the promotional clock.
How the country arrived here matters for everything that follows. When the government of the day decided in 2009 that a wholesale-only, state-owned fibre network would replace Telstra's privatised copper, it deliberately atomised the retail layer: the incumbent was paid to migrate its customers and decommission its lines, structural separation became the price of participation, and the barrier to entering broadband retail fell to a wholesale agreement and a brand name. No spectrum auctions, no trenching, no capital. Hundreds of firms took the invitation — resellers, regional specialists, energy companies cross-selling, IT shops adding a connectivity line item. The policy succeeded in exactly the way its designers intended and failed in exactly the way an economist would have predicted: entry was cheap, so entry happened; margin was regulated, so margin went to whoever could spread fixed costs over the most lines. The barrier to entry was lowered to a signature. The barrier to profit stayed where economics put it.
This essay is about a company that looked at that arithmetic and declined to play. emPOWER Data Services is the name on the routing records; blueAPACHE is the name on the door; Blue Apache Pty Ltd is the name on the corporate registry. It is a Melbourne business-to-business provider that has spent a quarter of a century selling private networks, cloud hosting, voice and around-the-clock management to organisations, not households — using the nationalised last mile as one commodity ingredient in a much more expensive dish. Working out where its margin sits, from primary documents, is the cleanest way to see what was actually left for private operators after Australia socialised the wires.
The company whose name lives in a routing registry
Start with the identity, because in this case the identity is itself instructive. Search for "emPOWER Data Services" in the places companies normally live — storefronts, price lists, review sites — and you find almost nothing. Search the infrastructure registries and it appears immediately. The Asia-Pacific address registry records autonomous system 17473 under the handle E2-CLOUD-AS-AP with the description "emPOWER Data Services", registered on 28 July 2011 to Blue Apache Pty Ltd of 383 Johnston Street, Abbotsford, Victoria. The technical contact is Chris Marshall — not a role account but the company's founder, his name written directly into the plumbing. The PeeringDB entry for the same network gives "blueAPACHE" as the alias and blueapache.com as the website, and was refreshed as recently as March 2026. A dormant LinkedIn shell still exists under the emPOWER Data name. That is roughly the extent of the brand's public life: emPOWER Data Services is what the network calls itself when it talks to other networks; blueAPACHE is what it calls itself when it talks to customers.
The corporate chain underneath is unusually clean for this column. The Australian Business Register, mirrored in public company-data services, shows Blue Apache Pty Ltd registered on 30 July 1998, ACN 083 664 224, ABN 82 083 664 224, GST-registered since the tax was introduced in July 2000, with former business names including Blue Apache Information Systems and, for six years in the 2000s, Blue Apache Financial Services — a trace of an early flirtation with a different market. The registered address has sat in the same inner-Melbourne postcode since 2010. The ABR's own search confirms the active registration; the full ASIC extract, with shareholdings and charges, sits behind the regulator's paid search and was not purchased for this piece, so the ownership register is reported here only to the depth the free record allows: a proprietary company limited by shares, founder-led, with no listed-market disclosure obligations and no insolvency traces in the public indexes searched.
One regulatory box could not be ticked either way. Australian law distinguishes between carriers, which own transmission infrastructure and need a licence, and carriage service providers, which sell services over other people's infrastructure and need only comply with a lighter registration regime. The communications regulator's register of carrier licences timed out on repeated queries while this piece was researched, and the ombudsman scheme's member directory offers only an interactive search that returned no verifiable public list; nothing in the public evidence suggests Blue Apache holds its own carrier licence, and its published architecture — services running over NBN Co, Vocus and other licensed carriers' transmission — is exactly what a carriage service provider looks like. The unresolved register query is stated here rather than papered over; it does not change the analysis, but a carrier licence appearing under this name would.
The routing history supplies the company's real biography. The oldest address block still announced by the network, 203.202.6.0/24, first appeared in global routing tables on 18 January 2001, which means this business has been operating internet infrastructure continuously since the dial-up era — before the NBN was policy, before Telstra's copper was declared a national problem, through the entire nationalisation experiment. Today the same system announces eleven IPv4 blocks totalling 3,328 addresses and a pair of IPv6 ranges, fully visible to every route collector that looks. Twenty-five years of uninterrupted routing is a modest but genuinely hard-to-fake operating credential; address space and continuity of announcement cannot be conjured for a brochure.
The spread the state sets, measured to the cent
To understand what blueAPACHE opted out of, price the game it declined. The wholesale regime it would have played in is one of the most exhaustively documented tariffs in world telecommunications. Under the original NBN model, a retailer paid twice: a per-line access charge and a separate per-megabit capacity toll called CVC, a two-part tariff that made a retailer's costs rise with its customers' usage and turned bandwidth dimensioning into a decade-long public brawl. The Australian Competition and Consumer Commission accepted a varied undertaking on 17 October 2023 that ended the brawl: capacity charges were phased out of the higher tiers, prices became CPI-indexed under a defined path, and — the number that defines the whole settlement — NBN Co's claim on its historical losses was capped. Of roughly $44 billion in accumulated unrecovered costs sitting in the company's regulatory account at 30 June 2023, only $12.5 billion may ever be recovered through future prices. The Commonwealth, in effect, formally forgave the difference. That write-down is the invisible subsidy inside every $57.60 invoice: the spread retailers now argue over exists only because the state agreed never to charge for most of what the network cost to build. NBN Co implemented the new deal in a fresh wholesale agreement from 1 December 2023, and the tariffs have marched on the CPI path since: $55.19 for 50/20 and $58.53 for flat-rate 100/20 from July 2025, then the convergence of both at $57.60 this week, with the last capacity-tolled tiers becoming flat-rate at the same moment.
It is worth pausing on what the old two-part tariff did to retailer behaviour, because the scars explain the market's present shape. Capacity was bought in aggregate, per interconnection point, and every retailer faced the same nightly temptation: buy less headroom than your customers' evening streaming required and pocket the difference, at the cost of congested speeds that only showed up in someone else's speed-test data. For most of a decade, Australian broadband quality was a public sport of dimensioning forensics — regulators publishing typical-evening-speed league tables, retailers gaming the tolls, the wholesaler blamed for both. The flat-rate settlement traded all that discretion away. A retailer now cannot economise on capacity because there is no capacity charge to economise on; the product became genuinely identical across brands, and with it the last technical basis for differentiation at the consumer layer disappeared. What the reform gave retailers in cost certainty it took away in any remaining ability to be better rather than merely cheaper.
Set against those inputs, the retail market's structure follows almost mechanically. The competition regulator's wholesale market indicators for the December 2025 quarter count roughly 8.8 million wholesale lines. Telstra's group takes about 3.2 million of them, TPG's about 1.6 million, Optus just over a million, with Vocus, Aussie Broadband and Superloop clustered between about 660,000 and 880,000 each. Every other provider in the country — the long tail that makes the "150 brands" statistic true — squeezes into roughly 665,000 lines, seven-and-a-half per cent of the market. That tail share has barely moved in six years; the ACCC was celebrating its rise from 6.8 to 7.1 per cent back in 2019. The reason is fixed cost. A retailer buying direct must interconnect at up to 121 points spread across the continent, backhaul traffic from each, run a support operation and fund customer acquisition in a market where the product is, by regulation, identical. Those costs divide by two million subscribers or by twenty thousand; the $32.40 gross spread does not care. Small retailers survive by buying aggregated access from larger intermediaries — surrendering another slice of the spread for the privilege — or by finding customers who are not buying the commodity at all.
Two further currents in the same quarterly data sharpen the squeeze. Nearly a third of all lines — about 2.8 million — now sit on tiers above 100 Mbps, and fibre-to-the-premises connections passed three million as the copper-replacement upgrades accelerate, which means the average household is consuming a better product whose wholesale price is higher and whose retail price the discounters dare not fully pass through. And the tail's structural handicap has a second layer that the headline brand count conceals: most of the 150-odd retailers do not connect to the monopoly at all. They buy pre-aggregated access from a handful of wholesale intermediaries, surrendering another margin slice for relief from the 121-point interconnection problem — which is why the competition regulator counts "access seeker groups" rather than brands, and why the visible market of dozens collapses, at the network layer, to roughly seven buyers of consequence. A brand is a marketing decision. An interconnection footprint is a balance-sheet decision. The spread only rewards the second.
The wholesale monopoly's own accounts complete the picture from the other side. NBN Co's FY25 results show $5.7 billion of revenue at an average residential take of $50 per service per month, earning $4.2 billion of EBITDA against 8.63 million connected premises — a 74 per cent margin business at the wholesale layer, before the depreciation on its vast build. The state, having forgiven itself the capital, now runs the profitable toll booth; the private sector runs the low-margin shopfronts. That is the precise shape of "what is left after nationalisation": a regulated spread wide enough to sustain perhaps half a dozen scale players and a rounding-error tail, and — crucially for this company — everything the monopoly does not sell.
The wrap: where the boutique margin actually sits
What the monopoly does not sell is service. NBN Co is forbidden by its wholesale-only mandate from touching the retail relationship; more importantly, its product is connectivity to a premises, not the management of a hundred-site corporate network with voice, security, cloud on-ramps and a help desk that answers at 3 a.m. That unsold layer is blueAPACHE's entire business, marketed under the emPOWER product family: a private MPLS wide-area network, internet access, hosted voice, a private cloud spread across three east-coast data centres, Microsoft ExpressRoute circuits and managed security, sold to mid-market organisations on multi-year terms. The company describes the network core as Cisco-based, fully owned and operated, and puts its investment in the platform at more than $6 million — a figure worth flagging as the company's own, single-sourced and unaudited, but usefully calibrated: this is a boutique's network, two orders of magnitude below a carrier's capital account, positioned as a control layer over other people's transmission.
Now assemble the unit economics from the primary documents, keeping evidence and inference separate. The evidence side. When blueAPACHE delivers a serious business site today, the tail is typically NBN Co's Enterprise Ethernet — the symmetric, service-levelled fibre product the monopoly sells alongside its residential tiers. The published wholesale price list, current since 1 May 2025, prices a 100 Mbps symmetric circuit at $215 a month in its basic traffic class, plus $150 a month for the underlying interface in central business districts or $250 in the outer zones: call it $365 to $465 of wholesale input before a single engineer is paid. A gigabit in the same class is $500 for the circuit; in the premium, fully committed traffic class it is $760; ten gigabits runs to $5,000. A guaranteed four-hour, around-the-clock repair regime adds $75 a month per interface. Geographically diverse routing for a gigabit adds up to $2,975. Every one of those numbers is on the public record, which means every competitor's floor price is too — and so is the striking comparison with the residential product: the household 100/20 service that costs a retailer $57.60 becomes, in enterprise dress with symmetric speeds and a service-level regime, a $365-to-$465 input. The nationalised network itself charges a six-to-eight-fold premium for business-grade determinism over nominal consumer bandwidth. That published gap is the headline metric of this piece, and it is the space in which every Australian B2B boutique lives: the wrap begins where the flat rate ends.
What does the wrap sell for? The company does not publish a rate card, but it has published a transaction. In January 2015 it announced a three-year, $2 million agreement with the Victorian and Tasmanian Synod of the Uniting Church, covering more than 80 services across the church's facilities — including the Uniting AgeWell care network — on its emPOWER MPLS platform, with hosted voice for roughly 2,000 end-points layered on top. Divide the contract by its term and the deal cleared about $55,600 a month; divide by the service count and the blended figure is a little under $700 per service per month, in 2015 dollars, with voice seats folded in. The blend is the point. Where a consumer retailer's revenue per line was then and remains now well under $100 with a state-priced input consuming most of it, a managed B2B contract cleared roughly seven times as much per connected service — against wholesale inputs that today's published tariffs place at $215 to $500 for the underlying circuits. The inference side, flagged as such: assuming the mix of that contract resembles the company's current book, the gross margin on the network layer of a managed WAN plausibly sits in the 50 to 70 per cent range before labour, versus the sub-40 per cent gross and single-digit net available on consumer resale. The company's continued existence, growth citations and staffing posture are all consistent with that inference; its actual accounts, as a proprietary company, are not public, and this analysis says so plainly.
The arithmetic becomes concrete if the published tariffs are assembled into a client no confidentiality clause protects: an imagined twenty-site mid-market organisation of exactly the shape this company's public wins describe. Two city hub sites take gigabit symmetric circuits at $500 plus the $150 metropolitan interface — $1,300 a month. Six regional sites take 100 Mbps symmetric at $215 plus the $250 outer-zone interface — $2,790. Twelve small sites ride the flat-rate 100/20 tier at $57.60 — $691. Four-hour restoration cover on the eight fibre sites adds $600; geographically diverse routing on the primary hub adds $1,275. Total wholesale input: roughly $6,660 a month, every dollar of it traceable to the two price schedules cited above. What the managed overlay on top of that sells for is the inference step, and it is flagged as one: the single observed transaction in the company's record — the church contract's blend of just under $700 per service per month with voice included — and the general shape of Australian managed-WAN deals suggest such a client bills somewhere between $12,000 and $18,000 a month once management, security, voice and cloud are layered on. On those figures the network layer alone clears between 45 and 63 per cent gross before a salary is paid — two to five times the gross available to a consumer reseller on the same national infrastructure, on one-fifth the customer count a consumer operation would need to produce the same cash. The evidence carries the input floor and one revenue observation; the band around the rest is stated as estimate, because it is one.
Two further contract traces corroborate the model without pricing it. In August 2020 the company won the managed-services transformation for Berry Street, one of Victoria's largest family-services charities. It holds a registered supplier profile on the New South Wales government procurement platform. Not-for-profits, care providers and mid-sized institutions recur across its public record for a reason that is itself economic: such organisations are large enough to need real networks, too small to run carrier procurement in-house, and sticky once their voice, WAN and cloud land with one intermediary. The revenue logic is subscription-shaped — multi-year terms, monthly recurring charges, modules added over time — and who pays is precisely the segment the scale players find expensive to serve well.
A five-gigabit network is a choice, not a shortfall
The technical record lets us check the story the company tells against the story the infrastructure tells, and they match — instructively. On PeeringDB, the network self-reports 1 to 5 Gbps of traffic with balanced flows, an open peering policy, and presence at five public exchanges — Sydney, Melbourne and Brisbane domestically, a 200 Mbps port in Singapore — plus interconnection facilities across Equinix and NEXTDC sites on the east coast, one in Tullamarine, one in Singapore and, curiously, one in Slough outside London. Its transit comes from several providers at once: Vocus's national backbone, Swoop's wholesale network, Virtutel, and Hurricane Electric for global IPv6 reach, with a New Zealand regional operator among its adjacencies and at least a couple of small networks — a broadband minnow called Dinkum Broadband among them — sitting behind it as customers of its own. The NEXTDC ecosystem listing confirms the data-centre tenancy from the landlord's side.
Read economically, the numbers are small on purpose. One to five gigabits is less traffic than a single mid-sized consumer retailer pushes through one exchange port at dinnertime; a network built to win the spread game would be embarrassed by it. A network built to sell determinism is not. Business traffic is thin — databases, voice, transactions, remote desktops — and what customers pay for is not volume but control: private paths between their sites and the company's three cloud rooms, direct on-ramps to Microsoft, multiple transit exits so no single upstream failure strands the base, exchange peering so that latency-sensitive traffic short-cuts the transit market entirely. The balanced traffic ratio is itself a tell. Consumer networks are wildly download-skewed; a balanced profile is what a hosting-and-WAN operator looks like, with as much traffic leaving its cloud as entering. The multi-homing across Vocus, Swoop and Virtutel is the supplier-side mirror of the product promise: transit in Australia is a competitive, falling-price input, so the boutique multi-sources it and keeps the whip hand — the exact inverse of its position against NBN Co, where there is one seller and the price rises by CPI every July.
The address holdings tell a quieter version of the same story. Three thousand three hundred IPv4 addresses is a rounding error for a consumer ISP but a comfortable estate for a B2B operator that numbers customer WANs privately and only exposes services publicly — and because the oldest blocks date to the beginning of the century, they were acquired when addresses were free, not at today's scarcity prices. The open peering policy posted on its exchange records fits the same economic logic: a business network with balanced traffic gains more from free interconnection than it loses, since every peer reached directly is transit not purchased and latency not incurred. Even the oddities are legible. The New Zealand adjacency matches the company's stated trans-Tasman clients; the rack in Slough is what serving an Australian client's London office looks like when you rent presence by the unit instead of building it — the entire international footprint is a set of rooms in other people's buildings, entered and exited on commercial notice periods.
The Slough and Singapore entries, plus the company's claims of clients across New Zealand, Asia and North America, sketch the follow-the-sun support and offshore reach of an operator serving Australian mid-market firms with international sites — a modest global footprint rented room by room. Again: control layer, not capital account.
Costs, suppliers and the direction of dependence
A boutique provider's cost base has three storeys, and the documents illuminate each unevenly. The ground floor is the regulated input, and here dependence runs entirely one way. Every tail the company delivers over the national network is priced by a monopoly on a published, CPI-indexed path; the company can substitute between NBN Co's tiers, or occasionally route around them with carrier ethernet from Vocus or fixed wireless, but for the standard Australian business premises there is no second last mile. The input price will rise every July by roughly inflation, forever, under the settlement the regulator blessed in 2023 — a predictable tax on the wrap. Predictability, note, is not neutrality: every CPI rise transfers a sliver of the wrap's margin to the state unless repriced onward, and mid-contract repricing is exactly what multi-year B2B terms make awkward.
The first floor is the competitive infrastructure it rents — transit, exchange ports, data-centre racks, Cisco iron, Microsoft licensing. These prices fall or hold in real terms, are multi-sourced, and constitute the healthy part of the cost structure. The top floor is people, and it is the expensive one: a 24/7 service desk the company says is staffed in-house to at least level two, certified engineers across every vendor stack it manages, and the ISO 27001 apparatus it was accredited for — the credential that unlocks care-sector and government-adjacent buyers. Service businesses of this shape typically spend more on salaries than on all network inputs combined; the company's persistent recruitment of Melbourne network and portfolio engineers, visible in its live vacancies, is the observable trace of that.
The wedge between those storeys is where the next few years of margin will be decided. The company's revenue is contracted in multi-year terms; its dominant cost is Australian professional labour, which has been rising faster than headline inflation in the technology trades; and its regulated input rises by exactly CPI each July under the settlement. A services firm in that position is short wages and long indexation — it wins when it can automate the service desk faster than salaries climb, and loses quietly, contract by contract, when it cannot. The tax layer, at least, is neutral: GST flows through both sides of the ledger and distorts nothing except the headline retail comparisons, which is why the analysis above strips it before measuring anything.
On the customer side, the switching costs compound quietly. A client whose WAN, voice seats, cloud hosting, backup and security monitoring all terminate with one provider faces a migration project, not a purchasing decision, to leave — and the regulated layer adds its own friction: the monopoly's enterprise product carries early-termination charges of 85 per cent of remaining contract value where NBN Co funded a fibre build into the premises, a lock the reseller inherits and, rationally, mirrors in its own terms. Three-year terms like the church contract are the norm this machinery produces. The risk that shadows all of it is concentration: a boutique's book is lumpy, its largest clients are institutions with procurement cycles, and nothing public reveals what share of revenue the top five relationships carry. That opacity is the price of reading a proprietary company from the outside, and it is flagged here as the largest single gap in this analysis.
Consolidation pulls one way; the service layer pulls the other
The competitive map divides cleanly at the line this company drew for itself decades ago. Below the line, in the spread game, consolidation is relentless and the December 2025 market data shows why: six groups hold better than nine of every ten lines, the tail's seven per cent has been static since 2019, and every CPI reset squeezes the discounters whose headline prices already sit below the wholesale input ex-tax. The scale players' response to commoditisation has been to buy volume — the entire post-2020 history of Australian broadband retail is a consolidation ledger — because when the product is identical and the input is priced by the state, the only private variable left is cost per subscriber.
Above the line, in the wrap game, the pressures invert. blueAPACHE's competitors are not the 150 brands but the enterprise arms of Telstra, Optus and TPG, national managed-service specialists, and a peer group of mid-market MSPs bolted onto networks. Against the giants, the boutique's pitch is attention: the mid-market client who is a rounding error to a tier-one account team is a marquee client to a 27-year-old Melbourne firm. Against the MSP crowd, its pitch is the owned network core — most managed-service firms broker connectivity; this one routes it under its own autonomous system, with its own address space and its own peering, which is a genuine if narrow moat: harder to replicate than a reseller agreement, cheaper to run than a carrier licence. The substitutes worth watching are technological rather than corporate. Flat-rate national broadband at $57.60 is, on the numbers, the cheapest WAN underlay ever sold in this country, and modern software-defined networking lets a bold IT manager lash consumer-grade tails into a passable private network; enterprise 5G and satellite do the same for thin sites. The boutique's answer — that someone still has to run it all, at 3 a.m., accountably — is true, but it converts a connectivity margin into a labour margin, which is precisely the historical direction of this company's drift: the nine appearances in the national reseller growth rankings, 33 per cent year-on-year growth at the last of them, were earned as an IT-services firm with a network, not a telco with technicians.
One structural threat deserves separate billing because it comes from the monopoly itself. The wholesale tariff now lists consumer-grade tiers at 500/50 for $60.85 and 1000/100 for $75.88 — gigabit-class raw speed at one-tenth the price of the enterprise product's equivalent circuit. Every July that gap persists, the question a boutique's cleverest clients ask gets sharper: how much determinism do we actually need? The honest answer for many branch sites is "less than we're paying for", which is why the twelve small sites in the worked example above already ride the flat rate. The boutique's book migrates, tier by tier, toward cheaper inputs wrapped in the same management fee — good for margin percentage, corrosive for the enterprise-circuit revenue base that justifies the network investment. The determinism premium is the moat, and the state is slowly draining it from below.
What the unofficial record whispers
The informal evidence around this company is thin in exactly the way its strategy predicts, and the thinness is itself the finding. Australia's great consumer broadband forum, where every retail provider of consequence has a thread of grievances a decade long, has no meaningful blueAPACHE presence at all — the silence of a firm that has never billed a household. The review trail that does exist is occupational. Across Seek's employer pages and roughly 130 Glassdoor reviews, staff sentiment splits sharply: consistent praise for technical breadth and client exposure, consistent complaint about workload, management style and turnover, with only about a third of reviewers saying they would recommend the employer. Read economically rather than morally, that pattern is what a fixed-price, always-on service wrap looks like from the inside — the margin defended in the unit-economics passage above is extracted, in part, from the service desk's evenings. What the signal cannot establish is trend: platform reviews skew disgruntled everywhere, sample sizes are small, and the figures are the platforms' own. Sustained engineering-vacancy growth or an attrition disclosure would settle it; neither exists publicly.
The formal growth record, such as it is, points the same direction as the strategy. Nine appearances in the national reseller growth rankings — a record for those awards — and five "All Star" citations for sustained multi-year growth, the last at 33 per cent year-on-year, are the visible residue of two decades of compounding a services book; the Berry Street win, supporting an organisation that serves 40,000 people, shows the anchor-client machine still ran through 2020. What the record does not show is anything after: growth-award programmes lapsed or went unentered, and no contract announcement of comparable weight has been published since. That could mean maturity, discretion, or drift. The live vacancy list — Melbourne engineering roles posted continuously — argues against drift, but only weakly; replacement hiring and growth hiring look identical from outside.
The most interesting recent signal sits at the top of the company. In July 2025, the founder who had run the firm for 27 years handed the chief executive role to his long-serving technology general manager and repositioned himself around "global growth", sales and marketing, months after relocating both major offices into premium CBD towers. There are two standard readings. One is orderly succession in a founder business entering its second generation of management. The other is grooming: founder-led services firms of this vintage and profile are precisely what mid-market private equity and consolidating MSP roll-ups have spent the decade buying, and separating the founder from operations while polishing the growth story is what preparation looks like. The public record cannot distinguish the readings today. A change in the shareholding register, a nominated adviser, or the abrupt appearance of an earn-out vocabulary in company communications would.
The facts that would move this judgement
The judgement as it stands: emPOWER Data Services is the network identity of a genuine, continuously operating, founder-built B2B provider whose economics rest not on the regulated spread but on a service wrap over state-priced inputs — a defensible boutique position in a market whose consumer floor is consolidating toward half a dozen firms. Several discoverable facts would revise it. Audited financials, in any form — a credit filing, an acquirer's disclosure — could confirm or demolish the inferred wrap margins; they remain the largest absence. Resolution of the carrier-licence question in the affirmative would signal an infrastructure ambition this analysis assumes away. A sale to private equity or a listed consolidator would end the boutique thesis and start a different one about synergy extraction from exactly the sticky mid-market book described here. On the input side, the settlement that created today's arithmetic is not eternal: the regulator's pricing path binds NBN Co only within the current undertaking period, and a future variation that cut enterprise ethernet tariffs — or a genuine second enterprise fibre network reaching the mid-market — would compress the six-to-eight-fold determinism premium that funds every wrap in the country. Conversely, a sharp CPI cycle would widen the wedge between indexed inputs and contracted B2B revenues and squeeze the boutiques first. And the quiet signals bear watching in both directions: the routing table growing new transit, prefixes or downstream networks would suggest the network side is being invested in; the Glassdoor pattern hardening while vacancies stall would suggest the labour engine that actually produces the margin is wearing out. On today's evidence, the company is what the paperwork in three registries says it is — and the paperwork, unusually, all agrees.
Evidence register
The load-bearing sources for this piece, and what each carries. The APNIC registration of AS17473 ties the emPOWER Data Services name to Blue Apache Pty Ltd, its Abbotsford address and its founder. The ABR-derived company record carries the 1998 registration, the ABN and ACN, and the former trading names. PeeringDB carries the self-reported traffic scale, exchanges and facilities; RIPEstat carries the live prefix count, the 2001 first-seen date and the upstream set. NBN Co's July 2026 and July 2025 pricing statements carry the residential wholesale tariffs, and its Enterprise Ethernet price list carries every enterprise input figure and the early-termination formula. The ACCC's SAU final decision carries the cost-recovery cap; its December 2025 market indicators and 2019 competition release carry market shares, the tail's stagnation and the 121 interconnection points. NBN Co's FY25 results carry the monopoly's own margins. The company's Uniting Church announcement carries the one observed wrap transaction; its leadership statement, network page and growth-award record carry the self-reported claims flagged as such in the text. WhistleOut's July 2026 tracking carries the retail side of the price pair. Seek and Glassdoor carry the workforce signals, weighted accordingly.