Summary
- NBN Co's October 2026 Network Premises Report defines both of its premises nouns in adjacent footnotes on the same page as the national totals, and prints column formulas that reproduce those totals exactly.
- The count that once made the larger total legible as a build measure — premises inside a declared ready-for-service area that cannot yet order a service — was printed at 60,000 in November 2020 and 1,138 in September 2023, and is absent from the October 2026 report.
- The regulator counts a third noun, Access Virtual Circuit services in operation, whose only printed relation to premises is a lower bound, so the ratio a reader will compute is a construction, not a published measure.
I reserved this piece to accuse a network of printing a footprint noun it never defines. NBN Co does not do that. Its monthly Network Premises Report carries two footnotes, one above the other, on the same page as the numbers, and both are definitions rather than labels:
Ready to connect: A premises is ready to connect when an nbn service order can be placed, and the service can be connected within an area that has been declared ready for service.
Premises activated: Refers to premises which have an active service installed.
That is a better document than the one I set out to criticise, and the correction belongs up here rather than in a closing caveat. The report is dated 11 November 2026, covers the month ending October 2026, and states its own basis in its opening sentence: "The data contained in this document reflects NBN Co's position for the month ending October 2026" (October 2026 report).
The national row prints 11,606,064 fixed line ready and 8,216,734 fixed line activated, 305,940 satellite ready and 57,406 satellite activated, 802,573 fixed wireless ready and 375,003 fixed wireless activated, and then the two grand totals: 12,714,577 Ready to Connect and 8,649,143 Premises Activated. The formulas sit beside them, (I)=(A+E+G) and (K)=(B+F+H). My arithmetic, and it is mine rather than the publisher's: 11,606,064 + 305,940 + 802,573 = 12,714,577, and 8,216,734 + 57,406 + 375,003 = 8,649,143. Both totals reconstruct from the printed components, so a stranger can audit this table with a calculator.
Then the arithmetic nobody printed. 8,649,143 divided by 12,714,577 is approximately 0.6803, or 68.0 per cent. 12,714,577 minus 8,649,143 is 4,065,434 premises that can order a service and have not activated one. Those are my sums. Across the fifteen documents read for this article on 21 November 2026 — the report and its landing page, the September 2026 tracking dashboard, the July 2025 and October 2024 editions, the 2018 and 2023 progress reports, the operator's glossary, the wholesale product description, the departmental methodology report and four ACCC publications — no page prints that percentage, that ratio or that difference.
The residue that stopped
What changed here is not a definition. In July 2018 the weekly rollout report carried four glossary entries rather than two, and the third named the thing the current report will not:
Not yet ready to connect: When a premises within an area declared RFS cannot currently order a service until further work is completed ... This includes SC0 (FTTP), SC10 (FTTN/B), SC20 (HFC) and SC30 (FTTC).
Four service classes, named, carrying the negation (rollout information, week ending 12 July 2018). That entry is what lets a reader read the larger total as a build measure: a premises counted inside a declared ready-for-service area, and explicitly not yet orderable.
The November 2020 assessment methodology report put a number on it. It records that NBN Co had delivered 11.854 million premises ready to connect, "representing 99.5 per cent of its footprint", and that premises yet to ready to connect fell from 109,000 at 30 June 2020 to 60,000 as at 20 November 2020. The same document says the quiet part out loud: "In practical terms, the RTC footprint can never reach 100 per cent, with a small proportion of premises being unserviceable at a given point in time" (methodology report, November 2020). That 99.5 per cent belongs to a 20 November 2020 datapoint and to a reporting surface the operator has since replaced. It is not a current figure and I am not using it as one.
Three years later the residue was still printed and the number was tiny. The progress report for the week ending 2 November 2023 carried 12,353,303 Ready to Connect and 8,582,189 Premises Activated, plus a separate "Yet to ready to connect" count of 1,138 premises as at 30 September 2023 (progress report, 2 November 2023).
The October 2026 report carries neither the count nor the term. So the residue has three dated printings in the record I read: 60,000 as at 20 November 2020, 1,138 as at 30 September 2023, and nothing at all in October 2026. That is a count of printings, not a trend; the three figures come from three documents in three vintages.
I will not tell you why it stopped, because no reviewed document does. What can be said is what a reader loses: the number that made "ready to connect" legible as construction rather than as a catalogue of orderable addresses is gone, and the two remaining totals now sit on the page without it.
The break the report dates itself
The October 2026 report also carries its own series break: "The methodology in this report was updated to reflect operational reporting changes from the month ending July 2025." The July 2025 edition, the first month under the new method, prints the same two definitions and 12,578,864 Ready to Connect against 8,631,386 Premises Activated (July 2025 report). The October 2024 edition prints the same two definitions and no methodology sentence (October 2024 report). So the definitions did not change at the break. Something in the operation of the count did, and the report says so in one sentence without saying what.
Between the month ending July 2025 and the month ending October 2026 — fifteen months, both endpoints inside the new method — Ready to Connect rose by 135,713, about 9,048 a month, and Premises Activated by 17,757, about 1,184 a month. Those are my subtractions and divisions and I am not projecting them forward.
The variations note names "validation of address data or statutory infrastructure provider service areas" among the causes of month-to-month movement, and says the brownfields volume is understated where brownfields are replaced by greenfields. Read beside a headline figure, that is a total which moves for reasons having nothing to do with anyone digging.
The regulator counts a different object
The ACCC's quarterly surface is not a premises count and does not pretend to be. Under the NBN Services in Operation Record Keeping Rules, issued in September 2014, the regulator collects "the number of Access Virtual Circuit (AVC) services in operation" (ACCC programme page). Its glossary is exact about the unit: an AVC is "an Ethernet-based, Layer 2 virtual connection on an NBN Co Network that carries traffic to a premise connected to that NBN Co Network", and a service in operation is "a service supplied to a premise connected to the NBN network" (statement of reasons).
The June quarter 2026 report prints 8,839,253 Total TC4 AVCs for March 2026 and 8,817,314 for June 2026 in its migration table (quarterly report). The page is served at an address ending "march-quarter-2026-report" and is headed June quarter 2026; the March row survives inside the June page's time series.
Set 8,817,314 TC4 AVCs for June 2026 beside 8,649,143 premises activated for October 2026. The service count is larger by 168,171, a ratio of about 1.019 — my arithmetic, offered only to show the ordering of two figures that are neither the same unit nor the same period. That ordering is what a reader reaches for, and nothing reviewed licenses the reach.
What the documents do say is a lower bound. NBN Co's wholesale product description for nbn Ethernet, version dated 1 September 2026, requires that a retail provider "must order an AVC for each Premises to which nbn Ethernet will be supplied", then permits it to "additionally elect to order one AVC TC-1, one AVC TC-2, or both" alongside the TC-4 (product description). The ACCC's October 2019 access pricing discussion paper points the same way: "RSPs need at least one AVC for each premises" (discussion paper). At least one. A lower bound is not an identity, and a permission is not a use: nothing reviewed shows a premises holding three AVCs, and nothing shows one holding exactly one.
Two nouns, two publishers, no bridge
The operator also runs a dashboard, in a third vocabulary. Its September 2026 edition states that "12.71 million Australian homes and businesses are able to connect to the nbn access network", 12.6 million in September 2025, and that "8.65 million homes and businesses connected to a plan over the nbn access network", defining the metric as "the number of homes and businesses that can order a plan via a phone and internet provider and connect" (tracking dashboard). Rounded millions, and the counted noun is homes and businesses. The October 2026 report gives exact integers and the counted noun is premises. Both are quoted here; neither is merged into the other.
What the reviewed documents do not print
Read as a set, across the fifteen documents consulted for this article on 21 November 2026, the absences are specific:
- No current page prints a premises-passed, homes-passed or ready-for-service-area total larger than Ready to Connect itself.
- No current page prints a percentage of footprint, a ratio of activated premises to ready-to-connect premises, or a ratio of services to premises. The only printed footprint percentage in this reviewed set is the November 2020 departmental figure.
- The October 2026 PDF was never rendered directly: its text came from extraction with spacing lost, and the headings for the third and fourth numeric columns are garbled. This article therefore quotes only the national row's positions A, B, E, F, G, H, I and K, whose arithmetic reproduces the printed totals, and quotes no state row and no column C or D value.
- No document in the reviewed set prints a "not yet ready to connect" count for any month after 30 September 2023.
- The operator's glossary page was retrieved only in part, so the absence of an entry for ready to connect, premises activated, serviceable or services in operation is limited to the inspected text.
- The ACCC page at the March-quarter-2026 address serves a June-quarter 2026 report, and no separate page for the March quarter itself was obtained.
- No reviewed document states that one Access Virtual Circuit equals one premises.
- No reviewed document prints a crosswalk, table or note reconciling the regulator's service counts with the operator's premises counts.
What this leaves a reader
A table that closes, two definitions that hold, and one missing subtraction. NBN Co's current report will tell you, to the premises, how many addresses in Australia can order a service and how many have activated one. It will not tell you how many sit inside a declared ready-for-service area waiting for work, because that line is gone. And it will not tell you what fraction of the orderable set has been taken up, because no page prints that either.
The division is yours. Do it, and label it: 68.0 per cent activated against ready to connect for the month ending October 2026, on the operator's own two integers. What you cannot do is point to a page that says so.
Three documents in that set do different work here. The report landing page carries the edition date and the scope sentence a reader needs before citing the PDF at all. The operator's glossary of terms defines neither premises noun in the portion I could read, so the definitions live beside the figures rather than in the glossary. And the ACCC's guide to interpreting the market-indicators data explains what the regulator's own columns mean — the missing step between its service counts and any premises denominator.
Member Briefing
Read the full trend briefing
Members read the full briefing with the source notes behind every finding.
Strategic Circle
A daily briefing in your language, built from changes we have verified.
Join for USD 20/monthCancel anytime.
Members also get the full research library, a daily briefing on the networks they follow and the private forum.






