What Happened
Telecommunications · Digital Divide · Satellite Internet · Malaysia · Column Briefcolumn of 6 Aug 2026 · record to 9 Sep 2026
Satellite direct-to-device · the case, and the table that makes it
Malaysia's Last Coverage Miles, and What a Satellite Can Actually Carry
On 30 July 2026 Communications Minister Fahmi Fadzil said Starlink representatives had told him they intend to trial direct-to-device coverage in Malaysia, and asked MCMC to bring the telcos to the table. His reason was arithmetic: a tower takes 18 to 24 months and, where the population is thin, is “neither cost-effective nor economically viable”. A week later a newspaper column made the wider case — that satellite is not a threat to Malaysia's operators but the piece of the map they were never going to finish alone. That case holds. Fibre beats satellite wherever fibre reaches, so the two serve density and distance rather than the same customer, and the mechanism keeps the operator in charge: the satellite acts as an LTE base station in space, on the operator's own spectrum, and the subscriber keeps their SIM and number. Two things in the arithmetic are softer than they look. The prices usually quoted for satellite — RM135 and RM235 a month — are for dish broadband, which is not the service being trialled and needs a kit nobody counts. And the direct-to-cell layer is not the constellation: of 11,124 working Starlink satellites, 639 carry it.
The numbers on both sides what a tower costs in time, and what the sky currently holds
- 18–24months to build a tower — the minister's own figure for why satellite is on the table
- 11,124working Starlink satellites, counted independently on 7 Sep 2026
- 639of them carry direct-to-cell — two shells, about 5.7% of the constellation
- 0next-generation V3 satellites in orbit; 20 launched, 20 recorded as failed to orbit
- RM135Starlink Malaysia's entry plan — up from RM129, and slower than the plan it replaced
- $10a month for satellite texting in the United States — a limited-time rate against $15
How much of this is established every claim on this page, graded — sorted by standing, never mixed
| Standing | What is claimed | Where it comes from |
|---|---|---|
| Confirmed | Starlink representatives told the communications minister they intend to trial direct-to-device coverage; MCMC was asked to invite the telcos to meet them | The minister, quoted at an MOU exchange in Cyberjaya, 30 Jul 2026. An intention to trial — no trial, licence condition or partnership is announced |
| Confirmed | A tower takes 18 to 24 months, and where the population is too sparse it is neither cost-effective nor economically viable | The minister's own words, not a commentator's inference |
| Confirmed | Direct-to-cell runs on an ordinary 4G handset over the operator's own licensed LTE spectrum, back to the operator's core network; the subscriber keeps their SIM, number and phone | Malaysian trade coverage of the same announcement, 1 Aug 2026. This is why the arrangement is wholesale rather than competitive |
| Confirmed | 11,124 working Starlink satellites, of which 639 are direct-to-cell across two shells — 674 launched | An independent orbital census, updated 7 Sep 2026 and read at its own summary table rather than through any report of it |
| Confirmed | 20 next-generation V3 satellites launched, 20 recorded in the failed-to-orbit column, none in orbit | The same table. It records the counts and gives no reason for them, and this page states no more than it does |
| Confirmed | The United States satellite-texting service costs $10 a month as a limited-time rate against $15, is free on the top plans, and now covers Canada, New Zealand and Japan as well | The operator's own product page, read in full. The $15 and the three extra countries are on it; neither travels with the figure when it is quoted |
| Confirmed | Starlink Malaysia's entry plan became RM135 for up to 100Mbps, replacing a RM129 plan rated at up to about 130Mbps; the standard plan rose from RM220 to RM235 | Malaysian trade coverage, 23 Jun 2026 — six weeks before the column that quotes the new prices without the direction they moved in |
| Confirmed | The Universal Service Provision fund takes 6% of weighted net revenue from licensees above an RM2 million annual threshold, under the Communications and Multimedia Act 1998 | The regulator's own published account of the fund and the 2002 regulations behind it. What the fund may be spent on was not established here |
| Confirmed, not verifiable here | Julau district in Sarawak had 4G across 58% of its populated areas in June 2025, and 1,391 of its 3,784 premises could get fibre | Attributed to a written parliamentary reply. The outlet carrying it refused every request made here and its archived copy could not be fetched, so it reached this page as a search extract only |
| Confirmed, not verifiable here | 5G reaches 82.4% of populated areas nationally, 62% in Sarawak and 68.6% in Sabah | Government data reported by trade coverage; the national figure dates from a February 2025 statement and the state figures from mid-2025. Neither outlet's page could be fetched here |
| Not public | How many customers Starlink serves, and in how many countries | Widely quoted and not established by any source read here, so this page carries no figure for it |
Timeline
How it arrived the coverage gap is measured in 2025; the offer arrives in 2026
- Jun 2025Julau, Sarawak: 4G across 58% of populated areas, and 1,391 of 3,784 premises able to get fibre — the shape of a gap a tower has to pay for
- mid-2025Nationally 5G reaches 82.4% of populated areas; 62% in Sarawak and 68.6% in Sabah. The national figure had already been stated in February and does not move much afterwards
- 23 Jun 2026Starlink Malaysia withdraws its cheapest plan. RM129 for up to 130Mbps becomes RM135 for up to 100Mbps; the standard plan goes from RM220 to RM235
- 30 Jul 2026At an MOU exchange in Cyberjaya the communications minister says Starlink has told him it intends to trial D2D here, and that MCMC has been asked to invite the telcos to meet them
- 1 Aug 2026Trade coverage sets out the mechanism: the satellite is an LTE base station in space, on the operator's own spectrum, and the subscriber notices nothing but coverage
- 6 Aug 2026A newspaper column argues the wider case, and proposes three things: publish the trial's results with deadlines, redirect universal-service money to terminals and D2D access, and have the operators bundle satellite messaging before they are made to
- 7 Sep 2026An independent orbital census puts the working constellation at 11,124, of which 639 carry direct-to-cell — and records no next-generation V3 satellite in orbit at all
The Argument
Density against distance, which is the part that holds the case for satellite, and the table it is usually made with
The central claim survives every check made here, and it is worth stating before anything is taken off it. Fibre beats satellite wherever fibre reaches — on price, on speed and on reliability — and no constellation changes that. A fibre operator cannot economically reach a longhouse upriver, and a satellite cannot beat a gigabit line in a city. They serve density and distance, and the national map is only finished when both are working. The mechanism supports the argument rather than merely permitting it: because the satellite carries the operator's own licensed LTE spectrum back to the operator's own core network, the subscriber keeps their SIM, their number and their handset. Nobody is switching to anybody. That is the reason the arrangement is wholesale and not competitive, and it is a stronger foundation than any retention statistic. Where the case gets shaky is the table usually put underneath it. RM135 and RM235 are dish broadband, sold with a kit and pointed at a roof. Setting them against fibre at RM89 or a gigabit at RM199 demonstrates that satellite loses a race nobody is proposing to run in Julau — and it quietly compares a monthly rental on one side with a monthly rental plus hardware on the other. Both sides of it, in fairness, are promotional prices.
- RM135a month buys dish broadband, not the service being trialled
- same SIMsame number, same handset — which is why nobody switches operator
The constellation is not the capacity what the direct-to-cell layer actually consists of, counted independently
“More than ten thousand satellites” is true, and it is not the number that matters here. Counted independently on 7 September 2026, Starlink has 11,124 working satellites — and the direct-to-cell service runs on two named shells totalling 639 working satellites, 674 ever launched. That is about 5.7% of the constellation, and it is the whole of what a phone in Julau would be talking to. The figure most often quoted for that layer, more than 650, is close enough to today's launched total to suggest the shells have been roughly complete for some months rather than growing with every launch. The same census records something the optimistic version of this story does not survive contact with: of the next-generation V3 satellites — 20 launched — the table lists 20 in the failed-to-orbit column and none in orbit. It says the count and not the reason, and this page will not guess at one. But it is the difference between two sentences that sound alike. The satellite is already overhead is true. It improves with every launch is a forecast, and the capacity that would carry more than text is not up there yet. What a phone gets today is messaging and light data — which, for a longhouse with no coverage at all, may be precisely the thing worth having.
- 639 / 11,124working satellites carry direct-to-cell — about 5.7% of the constellation
- 20 / 0next-generation satellites launched, and in orbit, on the same census
What Others Add
Two products, and only one of them is on the table what the dish sells, and what the phone gets
| Satellite broadband, with a dish | Direct-to-device, on an ordinary phone | |
|---|---|---|
| What you buy | A subscription and a kit pointed at the sky | Coverage, added to the mobile plan you already have |
| What it carries | Up to 100Mbps on the entry plan, up to 400Mbps on the standard one | Texting and selected apps. Data speeds are limited and some apps will not work |
| Who bills you | The satellite operator, directly | Your own mobile operator. The satellite is wholesale capacity behind it |
| Where it is priced | RM135 or RM235 a month in Malaysia, plus hardware | $10 a month in the United States as a limited-time rate against $15, and free on the top plans |
| Is it what is being trialled | No | Yes |
Two things the argument leaves open one is a question of law, and one is a date that helps the case
The fund is easier to name than to spend. Malaysia's Universal Service Provision fund does take 6% of weighted net revenue from licensees above an RM2 million annual threshold, under the Communications and Multimedia Act 1998, and redirecting it from towers to terminals where a tower cannot pay for itself is a coherent proposal. What is established is what the fund collects. What it may be spent on is set by regulation, and whether a satellite terminal or a D2D subscription is an eligible project was not established here — which is the question anyone acting on the proposal would have to answer first, and it is not asked. And the coverage gap is older than it is usually dated. The 5G figures — 82.4% nationally against 62% in Sarawak and 68.6% in Sabah — are habitually given as a late-2025 snapshot. The national figure was stated in February 2025 and has barely moved since. That is not a correction so much as a strengthening: a number that has stalled for eighteen months makes the case for a different kind of infrastructure far better than a fresh one would, because it is evidence that the remaining miles are not merely unbuilt but structurally unbuildable on the current economics. The argument had that in hand and reached for the newer date instead.
Conclusion
So what what to take from a proposal that is still only an intention
The case is right, and it is not the case the prices make.
Density against distance survives every check here. A fibre operator cannot economically reach a longhouse upriver and a satellite cannot beat a gigabit line in a city, so the national map is finished by both or by neither. What does not survive is the table underneath it: RM135 and RM235 are dish broadband, and the thing being trialled is a phone getting a text. If you want to know whether direct-to-device is worth doing in Malaysia, the comparison to make is against no coverage at all — which is what Julau has across 42% of its populated area — and not against fibre in Bangsar.
Count the layer, not the constellation.
Eleven thousand satellites is a fact about Starlink. 639 is the fact about the service a phone would use, and on an independent census the next-generation satellites that would carry more than text are not in orbit at all. That does not make the trial a bad idea — messaging where there is nothing is a large improvement on nothing. It means the promise should be sized to messaging, and a promise sized to broadband will be broken by physics rather than by anyone's bad faith.
The operator keeps the customer, and that is the design.
The reason a telco has nothing to fear here is not commercial goodwill but architecture: the satellite carries the operator's own licensed spectrum back to the operator's own core, and the subscriber keeps their SIM, their number and their handset. Nobody switches, because there is nothing to switch to. That is worth stating as a structural fact rather than as a retention statistic, because no source read for this page reports a customer count either way — and the structural version is the stronger claim in any case.
Nothing has been agreed. An intention was reported.
What exists is a minister saying Starlink told him it intends to trial something, and a regulator asked to bring the operators to a meeting. No trial, licence condition, spectrum arrangement or partnership has been announced, and no source read here reports that any Malaysian operator has met Starlink, agreed terms or commented. The proposals that follow — publish the results with deadlines, redirect universal-service money, bundle the messaging early — are reasonable, and every one of them is downstream of a conversation that has not yet been shown to have happened.