NTUC FairPrice, 13 June 2026. Watch it add up.
Source: NTUC FairPrice — 14 product pages, 13 Jun 2026 · [High]
SETHLUI — Hao Ge, Sep 2025 · [High]
…of your life, to cook one dish.
The hawker isn't cheap.
They price to the cent of what the ingredients cost, then absorb the rent, the gas, the ten-hour day. The $4 plate barely holds.
So who pays for it?
The plate barely holds. So who absorbs the difference?
The most common price for a plate of chicken rice in Singapore.
Take it apart.
Series high — CEIC / SingStat · Dec 2024 · [High]
56¢ of every operating dollar is gone before the hawker pays themselves a cent.
The owner's pay is whatever sits on top. If anything does.
NEA 2022 cost survey — MSE · 2 Jul 2024 · [High]
11 hours a day.
6 days a week.
Jonas Koh, 31, nasi lemak hawker. 286 hours a month.
"When I say break even, it covers my ingredients, my rent, my water, my gas. But not my salary."
Break-even 40 plates/day — Mothership · May 2026 · [High]
The hawker makes your $4 plate. The service worker rings up your groceries. Both on their feet, all day.
Hawker floor — Mothership 2026 · service median — MOM / SmartWealth · [High]
Same month worked. One gets employer CPF. One gets nothing.
Service median $3,315/mo + ~15% CPF; hawker no CPF — MOM / SmartWealth · [Med]
To earn what a clerical worker earns, a hawker would have to charge $6.66 a plate.
(≈100 plates/day, NEA rent)
The "$4–$5 hawker food" we expect is a subsidy. The hawker pays it, in their own labour.
Modelled — MSE cost split · NEA rent · MOM income · [Med]
Median age of a
Singapore hawker: 59–60.
The national labour-force median is 43. Hawkers skew 16–17 years older than the rest of us.
Preservation paradox — LKY School · NYC Food Policy · SilverStreak · [Med]
NTU 2021 youth survey — LKY School · [High]
The government even ran a course to train new hawkers. Follow what happened to the 566 who signed up.
HDP funnel 566→29→16 — MSE · 2 Jul 2024 · [High]
It isn't cheap because of efficiency.
Not because of subsidy. The $4 plate is cheap because the person cooking it earns $7.87 an hour.
And 87% of the next generation already know that. They said no.
$7.87/hr — Mothership · 87% — LKY School · [High]
The cook absorbs it. But the squeeze starts higher up.
SHEET 3 / 7
The 402
Two hawkers fry the same char kway teow in the same city. One pays $1,250 a month for the stall. The other pays $10,000. Here is the map of why, and who the machine runs for.
NEA median — MSE · coffeeshop band — Home & Decor · [High]
Same dish.
Same street.
From the street they look identical. Same wok, same queue, same $4 plate. The difference is invisible, and it decides everything: where you cook sets whether you can cook cheap and survive.
Median $1,250, Holland Drive overbid $7,012/mo — Mothership · DollarsAndSense · [High]
$1,250 a month. Every month. Hasn't moved since 2018.
In a city where a single coffee runs $7, a stall rent frozen at $1,250 is a real shield, confirmed in Parliament. The protected world is protected.
Median flat since 2018 — Mothership / Amy Khor · +0.3%/yr — MTI · [High]
The newcomer pays more than the incumbent. Every year.
2017–19 new-tender median $1,598 → ~$2,000 (2022) → $1,800 (2023). The shaded wedge is what a newcomer pays over the frozen median. Bid, then wait for the rent to settle.
New-tender median $1,598→$1,800 — NEA · MSE · [High · 2022 Med]
Win at any price.
Then wait.
For three years you pay exactly what you bid. Only then does the rent glide down to the assessed market rate. The system has a governor. First you have to survive your own bid.
Only 4% bid above AMR; Holland Drive $7,012 vs ~$720 (≈10×) — NEA, Nov 2024 · [High]
per month, for a single hawker stall.
Yang Ailan, 51, bid it for her son's cooked-food business. The highest on record in at least six years, several times the stall's assessed market rent.
Marine Parade record bid — Mothership · Aug 2024 · [High]
402 privately-owned coffeeshop units. HDB stopped selling them in 1998.
402 private units, sales frozen 1998 — Desmond Lee / Yahoo · 776 total — Dr Wealth · [High]
A coffeeshop that traded around $23.8m in 2013 fetched $41.68m by 2022.
3–5× in roughly a decade, on a gross yield near 4%. At a $40m price the owner has to pull ~$133,000/month off the stalls just to hit that yield, before a cent of profit. The unit is held like a collectible, not run like a kitchen.
$23.8m (2013) — Yahoo · $41.68m (2022) — ANN · ~4% yield — Dr Wealth · [High · yield Med]
The whole-unit head-rent has more than doubled in a decade. Someone has to cover it. The stalls below.
At $89,000/month head-rent, 8–10 stalls each owe $8,900–$11,000 just to cover the landlord. Before profit. Before ingredients.
Head-rent $89,000/mo — TOC · Sep 2025 · [High]
"That is not a negotiation. That is an eviction delivered in market language."
S$40m buy, $10k/stall, 10 of 14 quit — TOC · Sep 2025 · [High · secondary]
Not a villain.
A pass-through.
At an $89,000 head-rent, the operator's drinks monopoly isn't greed. It is load-bearing. Model a 10-stall unit at the Bidadari record and the operator nets roughly zero if stall rents hold at the band floor. The squeeze runs downhill.
Drinks 60–80% margin — KitchenUnion · Bidadari $73,888 — TOC · [Med · modelled]
Two interventions, years apart. Neither touches the thing at the centre: the unregulated spread on the private 402.
Unregulated spread, uncollected — NEA · HDB PQM · TOC · Sep 2025 · [High]
In a hawker centre, rent is a policy instrument.
In the private 402, rent is a yield.
One protects the plate. One prices it.
The machine runs. Who it runs for is the next question, and the answer is upstream. Follow the money.
NEA median stable — MSE · 402 owner-set, no AMR — Desmond Lee · [High]
Rent has an address. Look up.
Follow the
money up.
A plate of chicken rice. The wok is on the ground floor of a tall building. Climb it floor by floor, and watch who actually holds the bill.
Four dollars.
One plate.
A hawker. A wok. The lowest floor of a very tall building. The official average plate of chicken rice hit S$4.24 in Dec 2024, the highest in a series that runs back to 2014. Before we climb, stand here a moment, with the person doing the cooking.
S$4.24/plate — SingStat / CEIC · Dec 2024 · High
Eighteen stalls. One coffee shop. A whole year of cooking.
That works out to about S$3,000 a month each. The empty space above the bar belongs to people who never touch a wok. (Net figure modelled; no verified stall P&L exists.)
Median rent ~S$1,250 — MSE · deal — The Independent · Low · est.
He never
cooks a plate.
The operator holds the lease. He sublets each cooked-food stall, keeps the high-margin drinks to himself, and adds S$1,000–3,000 per stall in O&M fees. Together that roughly doubles the head-rent. Not one wok has fired up yet.
O&M S$1k–3k/stall — Home & Decor · Bidadari tender S$73,888/mo — TOC · High · structure
The money leaves
the room first.
The S$40m wasn't paid in cash. At ~70% LTV, ~S$28m was borrowed. At 3% over 25 years, the annual debt service comes to ~S$1.59m. That's about 74% of the gross stall rent, gone before the owner touches it.
Commercial LTV 60–75% — SMELoan · AceMortgage · High · math Med · rate
Across 25 years, who keeps more: the bank, or the landlord?
The bank collects about S$11.8m in interest. The owner's net stall-rent over the same 25 years runs roughly S$9–14m, before tax, O&M and vacancy. The bank out-earns the landlord on the landlord's own property. It just never gets blamed.
S$28m @ 3%/25yr → S$11.83m interest — SMELoan · AceMortgage · High · math Med · rate
Every resale,
a stamp.
Buyer's Stamp Duty on a non-residential sale runs 1%→5% across tiers. On a S$40m purchase that's ~S$1,969,600, the same published duty any commercial buyer pays, collected at completion. The Yishun shop changed hands in 2007 and 2022. Two transfers, two stamp cheques.
BSD top rate 5% — Yuen Law · PropertyGuru · Mothership · High
A flat, published 10% of Annual Value, every year, on every commercial property.
At ~S$216k a year, that's roughly S$5.4m over 25 years. The tax tracks Annual Value, so it moves with the market the assessor observes. A published, automatic rule, not a discretionary call. (AV is set by IRAS independently; gross-rent proxy makes this an upper estimate.)
10% of AV — IRAS · SGMoneyMatters · High · rate Med · AV proxy
Fixed. Finite.
By design.
Since 1998, no new private coffeeshop units have been released. The class is capped at about 402, a deliberate planning choice. Supply is fixed, a city keeps rising on top, and scarcity does the pricing. The land underneath reverts to the State at lease-end, as every 99-year lease does.
Sales frozen 1998, ~402 units — EdgeProp · EdgeProp · Yishun S$6m→S$40m — Mothership · High Med · 402
Who captures most
of your S$4 plate?
Like-for-like, in a single year of the Tampines model.
S$1.59m/yr debt service, ~S$11.8m interest.
18 × ~S$3k/mo ≈ S$0.65m/yr (modelled; no verified P&L).
Residual after bank + tax: ~S$0.35–0.57m/yr.
Tax S$216k/yr + BSD ~S$1.97m/resale; owns ~90% of SG land.
Per-year flows, Tampines S$40m / 18-stall model. 25-yr cumulative: Bank ~S$11.8m · State ~S$7.4m recurring. High/Med/Low per row.
You came looking for a villain.
There isn't one. The coffee-shop towkay everyone blames is a leveraged middleman, sandwiched between a bank that takes most of his rental cash flow and a State that planned the asset class, taxes it at published rates, and ultimately owns the land. The squeeze is transmitted at his floor. It isn't born there.
~84% of rent pre-committed; State owns ~90% of SG land — Land Portal · High · mechanism Med · proportions
Same S$10,000 stall rent. Watch what the cost of money does to it: is the landlord covering a cost, or pocketing a gap?
2022 reset earned ~0.2% cash-on-cash at 4.5% — SMELoan · Med · modelled
The land money sits at the top. It doesn't stay there. It cycles.
The hawker's capitalised labour partly funds the cheap flat the hawker lives in. The State is the apex owner, but the money is converted, not consumed.
GLS → Reserves, ≤50% via NIR — MinLaw · MOF · HDB S$6.78b deficit — TOC · High · mechanism
The hawker cooks on the ground floor of a tower whose every storey he helps pay for. Including, somewhere up there, his own home.
Is that a system that works, or one that works against him?
There is no villain.
There is an address.
Not the rent but the price. Not the towkay but the capped, planned scarcity that sets it, amplified by a bank that pre-claims most of the rent, softened by land money that converts back into the homes above. The squeeze is real. The villain is structural.
A note on framing. This chapter traces published duty rates, a published 10% property-tax rate, and transparent planning settings (the 1998 freeze, the 99-year leasehold). It maps where the dollars in one coffee-shop deal go; it imputes no wrongdoing to any party. Figures are confidence-graded and modelled on the Tampines S$40m / 18-stall case with stated assumptions; the bank rate, the IRAS Annual Value, and the S$3k hawker net are estimates, flagged in the receipts. Personal investigation · not for publication · every line sourced.
The money flows back to the land. The cooks are ageing out.
Two Futures2036
The warm future shrinks. The cold one fills the frame. That is the default.
One road needs nothing to happen. The other needs everything.
The shutters are down. Not a slow morning. There is no one to open them.
Sixteen years
older than the work.
The wonton-noodle uncle is 67. His body made the retirement decision. The median hawker is 59–60. The national workforce median is 43. That 16-year gap has no parallel in any other trade. By 2034 the cohort that was 60 in 2026 turns 68, and a ten-hour day in an un-air-conditioned stall is a physical fact.
Sources: LKY School of Public Policy (2024); SingStat Labour Force Survey.
Hawker age 59–60 vs workforce 43–44, ~16-yr gap — LKYSPP · 2024; SingStat · LFS
59–60 cites 2018–21 data; direction firmer than exact number.
They were reading
the arithmetic.
A hawker works 286 hours a month at the documented floor and earns S$7.87/hr. A clerical worker, same education profile, earns S$15.95/hr for 220 hours. In a 2021 NTU survey, 87.3% of respondents did not want to be hawkers.
Hawker floor: Mothership nasi-lemak case. Clerical: MOM 2025 via SmartWealth. NTU survey 2021 via LKY/NYC Food Policy.
S$7.87/hr, 286 h/mo hawker floor (Koh case) — Mothership · May 2026
Clerical S$3,510/mo = S$15.95/hr @220h — SmartWealth · MOM 2025
87.3% wouldn't be hawkers — LKY/NYC · NTU 2021
87.3% via secondary citation chain, not primary instrument.
566 trained.
16 stayed.
Singapore's Hawkers' Development Programme trained 566 people. 16 remain operating stalls today. 2.8%.
Hawkers' Succession Scheme: 6 transfers / 3.5 years / 13,400 stalls.
Source: MSE Parliamentary Reply, 2 July 2024.
HDP 566→120→29→16 = 2.8% — MSE · 2 Jul 2024
HSS 6 transfers/3.5yr, 13,400 stalls; ISP 37/99 = 37.4% — MSE · Feb 2025
The machine
that feeds itself.
When a stall closes with no successor, it goes to tender. New tenders run at a 44% premium over the frozen incumbent median. Every retirement tightens the economic case for the next entrant.
New-tender median S$1,800 vs sitting median S$1,250 (+44%). Source: MSE, Feb 2025.
S$1,800 vs S$1,250, +44% — MSE · Feb 2025
Staggered-AMR reform context — NEA · Nov 2024
Private stall rents S$4,000–10,000 — Home & Decor · 2024
Sub-lease S$4k–10k is commentary; HDB data not yet published.
The uncle who sold wonton noodles for 31 years retires in 2033. No one takes the stall. NEA re-tenders twice. One bid arrives, a bubble-tea chain. Rejected. The centre runs at 55% occupancy.
The heritage is preserved in aspic. The trade hollows out, stall by stall. This is the road where nothing changes.
"55% occupancy" is a scenario projection, not a 2026 datum.
Two roads
out of 2036.
The default path requires nothing new. It is the extrapolation of what is already measured. The path that pays requires three things to happen, in sequence, within six years.
These probabilities are a judgment, not a forecast.
Choose a path, or just keep scrolling for the full walk.
Keep scrolling: the full walk (Lock 1 → Lock 2 → Lock 3 → the judgment) follows below.
Split ~65% Ditch / ~25% muddle / ~10% Golden (dossier §10.5, judgment NOT forecast). Data — MSE · food prices; NEA · measures
There are no locks to walk. The default path requires nothing new, only that nothing changes. The uncle retires in 2033. No one takes the stall. The centre runs at 55% occupancy.
The heritage is preserved in aspic. The trade hollows out, stall by stall. This was already the default.
"55% occupancy" is a scenario projection, not a 2026 datum. Tap "back to the fork" to walk the other road.
The road
that works.
From here the page goes from grey to colour. Three locks, in sequence.
A coffeeshop bought at S$40M in 2022 required ~S$10,000/stall/month to cover debt service. The loans have halved in cost since. The rents have not.
SORA fell. The 2022-era debt burden eased. The gap between what the loan needs and what the stall is charged widened into extractable surplus. The rent machine has to stop before anything else can work.
Coffeeshops sold S$40M+ (Tampines S$41.7M, Yishun S$40M) — Mothership · Jun 2022
Stall rent reset to S$10,000/mo — Mothership · Sep 2022
SORA ~4.5%→1.0%; HDB unit record S$73,888/mo — TOC · Sep 2025
Rent control rejected — Malay Mail · Jan 2026
S$10,000 is the verified Yishun/Tampines reset, not a general rate.
Cheaper than
cooking at home.
Home-cooked chicken rice (4 servings, NTUC FairPrice, June 2026), with your time costed at the median wage: S$8.73 per plate. Hawker plate (SingStat official average, December 2024): S$4.24. The hawker is 2× cheaper than cooking at home, and still earning S$7.87/hour. The $5 ceiling is cultural, not economic.
The Parity Slider
What would a plate cost if the hawker earned a white-collar parity wage? Drag the volume.
| Plates/day | NEA | Coffeeshop |
|---|---|---|
| 50 | S$13.55 | S$19.05 |
| 80 | S$8.47 | S$11.90 |
| 100 | S$6.77 | S$9.52 |
| 120 | S$5.64 | S$7.93 |
| 150 | S$4.51 | S$6.34 |
| 200 | S$3.38 | S$4.76 |
Assumptions: Owner-draw parity = clerical-worker comparable (S$4,290/mo). Ingredient COGS 38% of revenue (gross margin 62%). 22 operating days/month. Model: dossier §02-D.5.
Demand elasticity (~20% drop per S$0.50) is a model assumption described in the dossier, not a directly-linked dataset.
FairPrice ingredient snapshot: 13 Jun 2026 (several promos expire late June 2026).
Home S$8.73/plate (ingredients S$4.15 + ~90min @$25/hr ÷4) — FairPrice · Jun 2026
Hawker avg S$4.24/plate — CEIC/SingStat · Dec 2024
Hawker-food CPI +6.1% — MSE · 2 Jul 2024
Parity model & rent anchor — MSE · Feb 2025; ingredient share 56% — MSE · food prices
FairPrice promos captured 13 Jun 2026; elasticity is a model assumption.
The documented hawker month: 286 hours (11h/day × 26 days). A standard full-time white-collar month: 220 hours.
The path that pays does not need full automation. It needs to bring 286 hours toward 200, the threshold where "demanding" becomes "comparable." Central kitchen prep. QR ordering. Shared cleaning. The technology exists. The cooperative infrastructure does not.
286 h/mo (11h×26d), Koh case — Mothership · May 2026
Full-time 220 h/mo (MOM); white-collar parity ~200 h (§10.2 scenario).
Chains use central prep; Kimly FY25 S$322.1M, 89 shops — Yahoo/The Edge · Nov 2025
We tested
every fix.
Rent ceiling alone: margin improves, still sub-parity. Price permission alone: revenue rises, then gets captured as rent at the private tier. Automation alone: hours fall, wages don't. Rent breaks first, so price increases flow to the operator, not the landlord. Then cultural permission to charge S$6.50–S$8. Then the hours reduction that makes the economics defensible to a 28-year-old.
Only one passes.
IN ORDER
In order.
Rent breaks first, so every extra dollar reaches the cook, not the landlord.
S$6.66/plate clears a white-collar wage, but only if rent breaks first. Source: §02-D.5 model, dossier.
Rent-capture: Tampines rent reset S$10,000/mo post-sale — Independent SG · 2022
Parity S$6.66–8.82 (§02-D.5); rent control rejected — Malay Mail · Jan 2026
Tender cross-ref — MSE · Feb 2025
A 31-year-old Nanyang culinary graduate opens a laksa stall in Buona Vista. She charges S$7.80 per bowl. 80 bowls before 11 a.m. Central-kitchen base. She finishes the braise herself. Her net draw: ~S$6,000–S$6,500/month. With CPF.
This is not a fantasy. It is the end state of three specific things going right.
The take-home/CPF figures are a modelled scenario grounded in verified inputs, not a measured fact.
S$7.80/bowl, 80/day, net ~S$6,000–6,500/mo +CPF (§10.2 scenario).
CPF gap ~S$940/mo; white-collar take-home ~S$4,800/mo — SmartWealth · MOM 2025
Ingredient share 56% — MSE · 2 Jul 2024
Modelled: take-home/CPF grounded in verified inputs, not a measured fact.
The default requires nothing new to happen. The renewal requires three interventions, in sequence, in the next six years. Against a political economy that has rejected rent control, against programmes that convert at 2.8%, against a median age of 60 already retiring.
The fork is real.
The clock is running.
The default is Ditch the Wok.
Which wok does Singapore choose?
Rent control rejected — Malay Mail · Jan 2026
HDP 2.8% conversion — MSE · 2 Jul 2024
Median age 60; ~65% Ditch (§10.5 judgment, NOT forecast).
A note on framing: this chapter traces published rates, transparent planning settings and verified succession data; it imputes no wrongdoing to any party. The 65/25/10 split is an explicitly-flagged judgment, not a forecast. Scenario figures (55% occupancy, the S$6,000–6,500 take-home) are modelled and labelled as such. Personal investigation · not for publication · every line sourced.
Two roads to 2036. One verdict left to give.
Hawker Thesis · Chapter 6 of 6 · Full Circle
A finale by Hoho. The verdict is a shape, not a name. Sources are cited inline; the modelled figures are tagged and caveated. The State is framed structurally, never as a villain.
The verdict is a shape, not a name. Tomorrow, it's still a $4 plate.