The three Singapore carriers, from an operator who runs all of them

I pay three mobile bills a month that have nothing to do with my own phone. Three accounts, three billing dates, three portals that each hold a different opinion about what a password should look like.

That is the first thing that differs between Singtel, M1 and StarHub once you run lines on all three, and it never turns up in a comparison table. Before any of this is about radio, it is about being a customer of three separate companies.

I run roughly a hundred mobile lines out of Singapore on real carrier SIMs, split across all three networks, and the support inbox comes straight to me. So this is the view from the admin desk and the rack rather than the coverage map.

What the second and third carrier actually cost

Every line here is a registered line tied to a real identity, with real limits on how many one holder can pick up at a time. Stock grows at the carrier’s pace.

Multiply that by three and the fleet stops being one pool of a hundred lines. It becomes three pools sharing shelves and power, with three renewal calendars, three ways a payment can fail, and three separate chances to miss a date and let a line go quietly cold.

Then add the reserve. Moving a customer back onto the same carrier after a hardware failure means holding idle cards on every network. Each costs its monthly, around $10 a line here, plus its share of power and the hub it hangs off, and earns nothing until the night somebody needs it.

None of that reaches the person renting a port.

The lead time nobody prices in

Ordering M1 cards and getting them activated takes about three days here. Every time. I have never shortened it by asking nicely. The other two move at their own pace, and the gap shifts with how the shops are staffed that month and whether I am picking up two cards or a batch.

Three days is nothing until a customer needs six lines on that specific carrier this afternoon and the honest answer is Friday.

Which makes this the most useful question you can put to a proxy seller, and almost nobody asks it: what happens if I need four more lines on the same carrier tomorrow?

If the answer is instant, on any network, in any quantity, then either they are sitting on a lot of idle stock across three separate companies, which is expensive and rare, or they do not own the SIMs. I have watched enough sellers advertise all three carriers and then take four days to produce one line to be comfortable saying that out loud.

Why “I need M1” is an inventory question

For a long time I read a named carrier as fussiness. It is almost always a request for a constrained item, and the constraint sits on my side of the transaction.

The reason is usually dull and legitimate. The customer’s accounts have spent months being seen on one network’s addresses. Move them to a different network during a fault and you have handed those accounts a new neighbourhood to explain. Some will not care. Some will, and neither of you finds out which for a week.

So a named carrier is a hard requirement here now, and the spare cards are held per network instead of in one pile.

The coverage difference is real, and it is yours rather than mine

There is a genuine difference between the three at a given address at a given hour. I can measure it on my own floor: one network strong at the front of the rack room and thin at the back, another the reverse.

The ordering changed when we shifted a shelf about a metre.

The difference is real and it is local. It belongs to that building, that wall, that hour, and it does not survive the trip to your office. A customer once ran a week of his own tests from a unit in the east and came out with an ordering that disagreed with mine. Neither of us was wrong.

Treat any carrier speed ranking as a local reading, this one included. I can tell you how three networks behave inside one metal-heavy room in Singapore. I cannot tell you which is fastest where you are sitting.

The one pattern all three share is the evening. Around the time a few million people are on the train home watching video, throughput sags and latency climbs on every network. That congestion is the crowd of real subscribers you are paying to sit inside of, and since all three do it, it tells you nothing about which to pick.

What happens past the data allowance

This is the difference I actually plan around, and it gets almost no airtime.

The plans are not built the same way, and the behaviour past the ceiling varies. One arrangement stops the line outright. One slows it to something technically alive and practically useless. One keeps going and bills the next block automatically.

On a personal phone those are three grades of mild annoyance. On a line carrying somebody’s scraping job they are three different failure modes, and only one of them is honest.

A hard stop announces itself. The line is dead, you can see it is dead, you top up or move onto another line, and the incident costs an hour.

The throttled line is the one that hurts. It looks alive. It answers. It passes every health check I can point at it: exit address correct, carrier correct, response under a second. Meanwhile the customer’s job crawls or times out and the dashboard sits there insisting everything is fine. I have lost a full day to that shape more than once.

The line that quietly keeps billing is easy to operate and unpleasant on the first of the month.

So choosing a plan shape per carrier is choosing a failure mode. I want lines that die loudly.

Address behaviour is weather, not a spec

People want a ranking of which network hands out a fresh IP on reconnect and which holds one address for hours. I will not give that as a property of a brand, because I have watched it flip: a network that reliably released an address on every reconnect started clinging to them for hours after a round of upgrade work, and nobody tells retail subscribers when it happens.

Measure it per line, this week, and put the customer who needs a session to hold still onto a line you checked recently. Sold as a permanent carrier feature, it becomes a promise you cannot keep next quarter.

The seven months I spent wrong about a carrier

A couple of years in I concluded one of the three was measurably worse, and I had numbers for it. Throughput and latency logged over weeks, same targets, same method, one carrier consistently behind the other two by more than noise.

So I stopped buying it. Let the existing cards run down as customers moved off them, and put the money into the two that tested better. It felt like discipline.

Seven months later we rebuilt the racking and I re-ran the tests because it was cheap to. The gap was gone. Not narrower. Gone.

Almost every line on that carrier had been sitting on one shelf, at the back, against an outside wall. I had measured a corner of my own room and written it down as a fact about a national network.

The testing cost nothing. The seven months cost real money, because when a customer turned up wanting a serious number of lines on that carrier I had four of them and a three day lead time. I lost the order and I deserved to.

What changed afterwards was that I stopped comparing carriers against each other. Each line now gets compared to what that same line did last week. A line is allowed to be slower than another line. A line is not allowed to be slower than itself.

Three suppliers is worse margin, and I keep it anyway

Running three is worse for margin than running one. More accounts, more admin, idle cards on every network, and volume split three ways so no single relationship gets big enough to matter to the other side.

I would still not run one.

A carrier decision lands on every line of that carrier at the same moment. A plan gets restructured. Something tightens at the account level. A large platform starts treating one range differently. It arrives all at once because it is all one network, and if that network is the whole fleet, the business sits one decision by one company away from a week that cannot be fixed or appealed.

Across three, the same event is a shrug and an afternoon of moving people around. That is insurance, paid monthly in idle cards and admin I dislike doing.

Four reasons people ask for a carrier

Most jobs have no carrier variable in them at all. Account actions, posting, logins, warm-up, ordinary scraping: all three are genuine Singapore mobile networks, and an address on any of them looks like a normal local person on a normal phone.

A fair share of the carrier requests I get are superstition, built on a forum post from four years ago or one bad afternoon nobody retested. So before paying a scarcity price for a named network, work out which of these you are holding.

Your accounts have already lived on that network for months. Real, and worth holding cards for.

A specific target has started challenging one range while leaving the other two alone. Also real, though worth testing rather than assuming, because those things reverse on the platform’s own schedule and nobody sends a notice.

Coverage where the hardware sits. That only matters when the hardware is in your building instead of mine.

You read somewhere that one of them is better for proxy work. This is the one I push back on, because you are about to pay for a preference you have no way of measuring.

The boring move is the right one. Start on whatever the default is, run your own target through it for a few days at a normal pace, and move only if that target tells you to. One afternoon on the thing you actually care about beats every carrier comparison you will read, this one included. If you want lines from someone who holds separate accounts on all three networks and will say plainly when the carrier you asked for is not what you need, that is what I run.

Get new guides and videos first — join the Telegram channel.