What rural internet actually costs in New Zealand

1. Fibre (UFB). The urban rollout is complete. If your address has it, stop reading — the Commission measured Fibre 500 at 512 Mbps down, 109 up and 6 ms, holding almost perfectly through peak hours. Nothing else here comes close on latency.

2. 5G fixed wireless, if you can get it. This is the step most rural guides skip, and on the Commission’s own numbers it beats satellite comfortably. Report 29 measured Spark’s Max Wireless 5G at 417 Mbps across all hours and 410 Mbps in peak, with 58 Mbps upload and lower latency than Spark’s own 4G plans. Against the 4G fixed-wireless average of 58 Mbps all-hours and 43 at peak, that is a different class of service. Two caveats that matter: the 5G results are Spark-specific, because each provider’s fixed-wireless infrastructure differs and the Commission says not to compare RSPs directly; and 5G coverage is limited, so for much of rural New Zealand it simply is not an option. Check your address before assuming.

3. A local WISP. Genuinely worth a call, and the option most people skip. Lightwire — the only WISP in the Commission’s programme — averaged 73 Mbps down (about 61 at peak) and 17 up. That is slightly ahead of the 4G fixed-wireless average and more consistent — 17% of its tests fell below 25 Mbps at peak against 35% for 4G in non-fibre areas — and the Commission’s own market report found 46% of rural households leaving copper chose a local provider over Starlink, and they are not making a mistake: they are buying somebody nearby to call.

4. 4G fixed wireless. In non-fibre areas it averaged 52 Mbps, falling to 37 Mbps in the 7–11pm peak. The average understates the problem: 35% of its speed tests in non-fibre areas came in under 25 Mbps during peak hours, against 24% off-peak. By comparison the WISP measured only 17% below 25 Mbps at peak. It is the most variable option on this list, and it has the highest latency of any technology the Commission measures.

5. Starlink. The most expensive monthly, and usually the fastest where the first three fail.

This is the part no comparison site does properly. Starlink moved New Zealand to a tiered structure in early 2026, and the tiers do not perform alike.

Commerce Commission, Measuring Broadband New Zealand Report 29, published October 2026, testing 1–31 July 2026, Table 3; the Lightwire row is from the report’s WISP section. Sample sizes are small on some rows — 51 Whiteboxes on Residential Max, 60 on Residential 100, 22 on 4G fixed wireless in non-fibre areas, 14 on VDSL — and the Commission says to factor in the error bars. Report 29 had too few ADSL testers to report, so that row is the last published figure, from Report 28 (April 2026). The Commission says the Starlink changes since Report 28 “reflect an improved validation of the sample, rather than a direct change in Starlink network performance”. There was not enough sample to report Residential 200 Mbps at all.

Starlink’s cheaper residential tier is capped at 100 Mbps and measured at 97 all-hours, 92 at peak. Max measured 284 and 253. So Max is nearly three times the speed — but the honest question is whether you need it.

The Commission tested that directly. 95% of Residential Max households could stream two simultaneous UHD Netflix streams at peak. On the 100 Mbps plan it was 87%. An eight-point difference. On game downloads the gap is wider: Hogwarts Legacy took under 50 minutes on Max and up to two hours on the 100 Mbps tier.

The practical read: if your household streams and browses, the cheaper tier is very likely enough — 87% versus 95% is not what most people are imagining when they pay for the top plan. If you move large files, have five or more people online at once, or are in a congested beam, Max earns its premium. Only 1–3% of tests on the 100 Mbps tier fell below 50 Mbps, and 2% on Max fell below 100, so the floor is decent on both.

One thing the Commission says plainly that nobody repeats. Starlink retired its old “Residential” and “Residential Lite” plans for the tiered structure in early 2026. Residential Max performs the same as the prioritised plan it replaced — but of the new 100 Mbps tier the Commission writes that it “shows a large decrease in performance compared to results from the previous ‘Residential Lite’ plan”. If you were on Residential Lite and moved across, you very likely got slower.

The distribution shows where the gap bites. 51% of Residential Max tests hit 300 Mbps or more off-peak, and 32% at peak. On the 100 Mbps tier, 64% of tests reached 100 Mbps or more off-peak, falling to 43% at peak. But the floor is good on both: only 1% of tests on the 100 Mbps tier fell below 50 Mbps off-peak, rising to 3% at peak. So the cheap tier is reliable and slower, not unreliable.

Starlink notified customers in mid-May 2026 and the new prices took effect from the mid-June billing date. These figures are corroborated by the NZ Herald, Farmers Weekly and customer emails posted publicly.

Starlink’s customer email was titled “Upcoming Price Adjustment” and attributed the rise to network capacity, coverage and reliability investment. The company did not answer the NZ Herald’s questions about it. Prices are per month in NZD and vary by address — confirm at the checkout for your own location.

The hardware question, which most guides get wrong

You will see the kit quoted anywhere between free and $799, and all of those figures have been true at some point. Starlink changed its New Zealand sales model in January 2026: instead of buying the dish outright, you can now rent it free, paying only a small shipping and handling fee. Outright purchase was around $799 before the change, and there have been rural promotions well below that; Starlink’s New Zealand site listed the standard kit at $599 to buy when we checked in October 2026.

The catch is the reason to read the fine print. Plans are generally no-contract — but a 12-month minimum term can apply if you are on a free-hardware deal. If keeping the freedom to leave matters to you, buying the kit is what buys it.

Sales-model change reported by rural connectivity providers from January 2026; the minimum-term caveat is from NZ Herald reporting of the price rise. We have not tested a checkout ourselves and say so rather than presenting these as measured.

Disconnections. The Commission counts dropouts lasting over 30 seconds. In Report 29 the median for Starlink was no more than 0.5 a day to New Zealand servers — so half of satellite households see no more than that. Fibre is lower again. It is a small number either way, and it is the one most people assume satellite fails on.

Bufferbloat. This is the metric that decides whether a webpage still loads while somebody else is streaming, and it is measured under load rather than idle. The finding runs against expectation: satellite’s idle bufferbloat is higher than fibre’s, but under download and upload load the results are “more comparable” — and both handle it far better than ADSL, VDSL or 4G fixed wireless, which degrade noticeably.

And a third that inverts the usual assumption: latency. Starlink measured 26 ms. 4G fixed wireless measured 42 ms across all areas and 46 ms at peak in non-fibre areas. Satellite is markedly more responsive than the terrestrial wireless option most people would assume is quicker, because the constellation is in low orbit and the cellular network is congested.

Disconnections and latency are from Measuring Broadband New Zealand Report 29. Bufferbloat is from Report 28: technical problems during Report 29’s test month meant the Commission left bufferbloat out of that report. These metrics exist in every MBNZ report and are almost never cited in consumer comparisons, which tend to stop at download speed.

Fibre available: take it. Cheaper, faster, and 6 ms against 26.

No fibre, but 5G reaches you: take it. On the Commission’s figures it is several times faster than anything else short of fibre. Coverage is the constraint, not performance.

No fibre, no 5G, WISP in range: get a quote. Similar speed to 4G fixed wireless, local support, and often cheaper than satellite. Nearly half of rural switchers choose this route.

None of the above, ordinary household: Starlink’s cheaper residential tier. 87% of those households stream two UHD feeds at peak, which is most of what most people need.

Heavy use, several simultaneous users, or a congested area: Residential Max.

Still on ADSL: at 11 Mbps (the last figure the Commission published, in Report 28), almost anything is an upgrade. This is where the Commission’s “twenty times faster” line genuinely applies. On VDSL at 69 Mbps it does not — that gap is closer to four times, and worth thinking about before you spend.