Opening a café in Singapore, decoded.
The most romanticised small business in Singapore — and one of the deadliest. In 2024 a record ~3,000 F&B outlets closed; of the ones that died young, 82% never recorded a single dollar of profit. Here is the picture the dream skips: the real startup cost, the licences in the order that actually matters, the rent-and-wages maths that kills cafés, and the few places a new one genuinely wins.
3,047
F&B closures in 2024 (20-yr high)
82%
of young closures never profited
~S$100–250k
realistic all-in to open
S$195/yr
the SFA licence (the cheap part)
The MOAT Score: is a café worth building?
Before the details, our one-number read. The MOAT Score grades a sector's economic quality on the value-investing lens of Graham, Buffett and Munger — four pillars (Margin, Operating moat, Appetite, Treadmill), each out of 25. The café is one of the lowest-scoring sectors we cover.
Brutal
Brutal — most lose money.
How the score is built
The MOAT Score sums four pillars — each scored 0–25 — from the value-investing lens of Graham, Buffett and Munger. No black box: here is the working.
Margin
6/25Does the average operator actually keep money — real net margin and return on the capital tied up?
This sector: net margins are thin to often negative; 82%SourceMTI parliamentary reply, Nov 2025 of young F&B closures never recorded a profit.
Buffett 1979 — “a high earnings rate on equity capital… without undue leverage”; 1986 owner earnings.
Operating moat
6/25Pricing power and a durable competitive advantage — can a typical operator raise prices and have customers shrug?
This sector: a price-taker against the state-held hawker-kopi floor; concepts, recipes and fit-out copy overnight.
Buffett, FCIC 2010 — pricing power is “the single most important decision”; 1991 franchise; 2007 moat.
Appetite
14/25Demand durability — steady, recession-resistant repeat demand vs fragile, discretionary or faddish.
This sector: coffee demand is genuinely steady and habitual — but the individual café unit is fragile and footfall-dependent.
Graham, Security Analysis Ch.2 — inherent stability “derives from the character of the business”.
Treadmillinverted · less is better
8/25Capital intensity and structural drag — rent, churn, fashion, discounting. Scored inverted: less treadmill, more points.
This sector: rent + wages eat roughly two-thirds of revenue, with a 15–30% delivery-commission leak on every online order on top.
Buffett 2007 — “the worst sort of business… requires significant capital… Think airlines.”
M + O + A + T, out of 100
The MOAT Score is a transparent SGAI judgement on a sector’s economic quality through a value-investing lens — not a verdict on any individual business, and not a comment on an owner-operated livelihood (a sector can score low on capital returns yet work as a job).
The read: it's not a coffee business — it's a rent-and-wages business wearing an apron
The numbers are brutal and they're official. In a November 2025 parliamentary reply, MTI disclosed that of the F&B businesses that closed young (within five years), 82% had never recorded a profit in their tax filings. 2024 was the worst closure year in about two decades. The café isn't failing on taste — it's failing on arithmetic.
Here's why the maths is so unforgiving. A S$6 flat white has a fat gross margin — the milk and beans cost well under a dollar. But the state deliberately keeps hawker coffee cheap (kopi is S$1.20–2.00), so a café can never win on the cheap-fuel axis. It has to sell a different category — craft, setting, experience — while paying Singapore rents and Singapore wages. Get the rent-to-revenue ratio wrong and no amount of latte art saves you.
And it bleeds even the giants. Luckin Coffee lost ~S$8.8M in Singapore in 2024 despite 81 stores and S$4 coffee. Flash Coffee collapsed — all 11 outlets shut, ~S$14.9M owed, baristas unpaid — just months after raising US$50M. Capital and technology don't fix a bad lease and thin unit economics. Nothing does.
Why a café can't win on price
A café is fenced in: the state holds hawker kopi cheap below, so a café must charge a craft premium just to survive the rent — but customers anchor on the cheap floor.
- Hawker kopi (kopitiam)S$1.6
deliberately kept cheap by the state
- Kopitiam-style latteS$3.5
- Specialty flat whiteanchorS$6
the café median — fat GM, but it must clear the rent
- Hotel / destination caféS$9
Source: SGAI price survey of SG coffee formats, 2026 (indicative)
The map: a ~S$12B F&B market — and the “coffee market” number everyone quotes is junk
Singapore's food & beverage services sell roughly S$0.95–1.1 billion a month (SingStatSourceSingStat monthly F&B receipts, 2025) — order of ~S$12B a year, with online already ~24–26% of sales. Ignore the aggregator reports claiming the “Singapore coffee market” is US$30MShaky figure — treat with cautionmarket-size aggregatorEstimates contradict each other 30–50×; one chain alone (Luckin) runs 81 stores. Anchor on SingStat.; they contradict each other 30–50× and one chain alone (Luckin) runs 81 stores. Anchor on SingStat, not market-size PDFs. And mind the distinction: a kopitiam is a landlord business renting stalls; a café is a third-wave, experience-led shop — different game entirely.
~S$12B
F&B services / year
SingStat 2025 monthly receipts
~24–26%
of F&B sales are online
structural, not a fad
3,047
F&B closures in 2024
~20-year high (ACRA/DOS)
2,431 vs 3,357
closed vs opened (Jan–Oct 2025)
the market runs hot
F&B closures hit a 20-year high
The closure count everyone half-remembers is years out of date. The line keeps climbing — 2024 set the record.
2024—a ~20-year high for F&B closures (ACRA/DOS)
Stale: the pre-pandemic figure older guides still cite
Source: ACRA / DOS closure counts, 2024. Pre-2024 points indicative of the rising trend; 2024 = 3,047 (reported).
The players: who you're really studying
The independents that survive almost all share one trait — they don't live on coffee-by-the-cup. The single most predictive question is whether the P&L is diversified beyond the counter.
The survivors diversify away from the cup
| Player | Model | Diversified P&L? | Pricing power? | SG profit signal |
|---|---|---|---|---|
| Common Man Coffee RoastersFlagship roaster-café (2013) | Cafés + roastery + wholesale + academy | The textbook survivor | ||
| PPP CoffeePioneer roaster (ex-CSHH) | Roaster + café + wholesale | Specialty originator | ||
| Atlas / Assembly CoffeeBrunch-led independent group | Own-brand multi-outlet | ~S$200k founder capital to start | ||
| Luckin CoffeeApp-first value chain, 81 stores | Tech / scale, S$4 coffee | −S$8.8M (2024) | ||
| Toast Box / Ya KunLocal kopi chains | Franchise / family-run scale | The cheap-fuel incumbents |
Common Man Coffee Roasters
Flagship roaster-café (2013)
- Model
- Cafés + roastery + wholesale + academy
- Diversified P&L?
- Pricing power?
- SG profit signal
- The textbook survivor
PPP Coffee
Pioneer roaster (ex-CSHH)
- Model
- Roaster + café + wholesale
- Diversified P&L?
- Pricing power?
- SG profit signal
- Specialty originator
Atlas / Assembly Coffee
Brunch-led independent group
- Model
- Own-brand multi-outlet
- Diversified P&L?
- Pricing power?
- SG profit signal
- ~S$200k founder capital to start
Luckin Coffee
App-first value chain, 81 stores
- Model
- Tech / scale, S$4 coffee
- Diversified P&L?
- Pricing power?
- SG profit signal
- −S$8.8M (2024)
Toast Box / Ya Kun
Local kopi chains
- Model
- Franchise / family-run scale
- Diversified P&L?
- Pricing power?
- SG profit signal
- The cheap-fuel incumbents
The cautionary tale: Flash Coffee
US$50M raised, 11 Singapore outlets, then dead within months — ~S$14.9M owed to creditors, including unpaid barista wages and CPF. Store count is vanity; cash runway and unit economics are survival. If a VC-backed chain with that war chest couldn't make the maths work, respect the maths.
The customer: it's not who orders — it's which channel they order through
The flat white is the same drink whether someone walks in or taps an app — but the channel decides what you keep. Dine-in carries the full price; every delivery order hands 15–30%Sourceplatform commission norms, 2025 to the platform. The revenue channel mix is the split that quietly rewrites the P&L.
Where café revenue actually comes from
Channel mix for a typical SG shopfront café (indicative). Dine-in keeps the full margin; the delivery slice is the commission-bleed channel.
- Dine-in62%foodservice-wide figure (Mordor 2025) — not café-only; flag for verification
- Delivery (apps)25%online is ~24–26% of F&B sales (SingStat 2025) — the commission-bleed channel
- Takeaway / counter13%indicative, residual
Source: SingStat 2025 (online share) + Mordor 2025 (foodservice-wide dine-in share, NOT café-specific — treat as indicative). Takeaway is the residual.
The economics: it isn't the cup that kills you
The internet says you can open a café for “S$50k.” A real shopfront is ~S$100–250k+ all-in (the founders of Assembly put in ~S$200k). The coffee carries a healthy gross margin — what kills cafés is the combination of rent and wages. Healthy occupancy cost is ≤10% of revenue; many SG cafés run mid-teens to mid-20s%. On top: a services-sector foreign-worker quota (DRC 35%, the strictest S Pass sub-quota at 10%), the mandatory Food Services Progressive Wage Model (floor rising to S$2,500 by 2028), GST at S$1M turnover, and delivery commissions of 15–30% on every online order.
All-in cost to open a real shopfront café
Fit-out, deposit, equipment, and several months of working capital. The “~S$50k” figure circulating online is not credible for a real shopfront — under-capitalisation is a leading cause of failure.
Source: Operator reports; Assembly founders ~S$200k
Where the café dollar goes
A typical SG shopfront café, modelled per S$100 of sales. Rent and wages alone take the lion's share — what's left for the owner is thin to negative.
- COGS (beans, milk, food)−32%68% left
High GM on coffee, lower on food · SG F&B benchmarks / SGAI calc
- Labour + CPF−28%40% left
PWM floor rising to S$2,500 by 2028 · MOM PWM
- Rent + occupancy−18%22% left
Healthy is ≤10%; sector avg ~11%, many run mid-teens to mid-20s% · MTI PQ 2024 / operator reports
- Other fixed (utilities, licences, marketing, POS)−9%13% left
- Delivery commission haircut−4%9% left
15–30% on the ~24–26% of sales placed online · SingStat 2025
What the owner actually keeps
Verdict: A healthy 9% — there is real margin of safety here.
Illustrative model on SG benchmarks (2023–2026); café net margins are thin to often negative. Not financial advice.
Cafés & specialty coffee
Would a value investor own the average operator here?
A value investor would not want the average café — it's a price-taker on a rent-and-wages treadmill with no moat, where coffee demand is steady but the unit is fragile.
Hawker kopi is held cheap by the state; cafés discount to win footfall and sweat every price rise.
Buffett, FCIC 2010 — pricing power is “the single most important decision”
Concepts copy overnight; location + brand are the only edges, and both are rentable.
Buffett 2007 — an enduring moat protects returns on capital
ROIC barely clears the cost of capital after fit-out + maintenance.
Buffett 1979 — a high earnings rate on capital, unleveraged
Fit-out + equipment + working capital; to grow you sink more capital.
Buffett 2007 — the worst business needs much capital, earns little
Repeat coffee demand, but discretionary and footfall-dependent.
Graham, Security Analysis Ch.2 — inherent stability is qualitative
Diversified P&L (roastery + wholesale), a real brand, or low-cost scale — not coffee-by-the-cup.
Assessment uses the value-investing lens on SG café unit economics (2023–2026). A lens on economic quality, not a verdict on an owner-operated livelihood.
Model your own survival line — drag the sliders:
Break-even
172
customers a day, just to break even
Monthly net
−S$7,967
Rent / revenue
32%
danger
Verdict: At these numbers you'd join the 82% that never turned a profit.
Illustrative model on SG benchmarks (2023–2026). Café-specific margins are scarce, so this is a starting frame, not financial advice. Rent gauge: ≤10% healthy, 10–18% tight, >18% danger.
How to actually open one (in the order that matters)
Almost every guide leads with the cheap, fast SFA food-shop licence (S$195/yr). That's the last easy step. The real bottleneck is the premises.
The approval stack, sequenced
- ACRA — register the business.
- URA change-of-use (~S$500) — the premises must be approved for F&B use. Gate 1.
- SCDF fire safety — plans via a Qualified Person → Fire Safety Certificate. Gate 2.
- WSQ Food Safety Course + a Food Hygiene Officer.
- SFA Food Shop Licence — S$195/yr, in-principle approval ~7 working days (the cheap, late step).
- By concept: liquor licence (SPF), MUIS halal (widens the market; mutually exclusive with alcohol/pork), signboard, public-entertainment.
The real bottleneck
It is the premises trifecta — URA change-of-use + SCDF fire certificate + the tenancy itself — sequential, QP-dependent, weeks to months. Get it wrong and you're paying rent on a café you can't legally open. Negotiate a fit-out / rent-free period around these approvals, not after. Note too: from 19 Jan 2026 the SFA SAFE framework replaces the old A/B/C/D hygiene grades — any older guide citing letter grades is already out of date.
Where a new café actually wins
Coffee-by-the-cup won't carry a Singapore lease. The survivors diversify the P&L away from the counter.
Vertical hybrid
Café + roastery + wholesale + workshops — the only evidence-backed route to net profit (the CMCR model). Beans and wholesale carry the P&L, not lattes.
Specialty in the heartland
Supply clusters in the CBD, Tiong Bahru, Katong. Under-served HDB towns mean real demand at lower rent.
Halal-certified specialty
Most specialty cafés serve pork/alcohol and skip halal — a deliberately under-served intersection that unlocks a far larger market.
Beans DTC + subscription
Diversify beyond footfall and rent dependency with direct-to-consumer beans and recurring revenue.
A distinct niche
Work-friendly, single-origin, local-fusion — anything to escape the "nothing uniquely yours" oversupply trap.
The honest AI edge
Demand forecasting, inventory and labour scheduling trim 2–15% off cost lines — real, but it won’t rescue a structurally bad lease. Use it to sharpen a good business, not save a doomed one.
Questions founders ask
How much does it cost to open a café in Singapore?
Realistically S$100,000–250,000+ all-in (fit-out, deposit, equipment, and several months of working capital). The widely-quoted "S$50k" figure is not credible for a real shopfront café — under-capitalisation is a leading reason cafés fail.
What licences do I need to open a café in Singapore?
In order: ACRA registration, URA change-of-use (or HDB approval), an SCDF Fire Safety Certificate via a Qualified Person, a WSQ Food Safety Course plus a Food Hygiene Officer, and the SFA Food Shop Licence (S$195/year). The bottleneck is the URA + SCDF + tenancy chain, not the cheap SFA licence.
Are cafés profitable in Singapore?
Coffee carries a high gross margin, but rent and wages eat roughly two-thirds of revenue. Of F&B businesses that closed within five years, 82% had never recorded a profit (MTI, Nov 2025). A café only works at high, consistent utilisation and usually with diversified revenue (wholesale, roastery, workshops).
Why do so many cafés fail in Singapore?
It is a rent-and-wages business, not a coffee business. Under-pricing, under-capitalisation and high occupancy cost kill most of them — even Luckin lost about S$8.8M in Singapore in 2024, and Flash Coffee collapsed despite raising US$50M.
Want a report like this for your company or sector?
Tell us what you're trying to figure out. We'll build you a Deep-Context report — the same depth, focused on your business.
About this report. Built with SGAI's Deep-Context Engine — human-directed, AI-accelerated. Figures draw on SingStat, an MTI parliamentary reply (Nov 2025), SFA, URA, SCDF, MOM, IRAS and market reporting (2023–2026). Café-specific net margins are scarce; where only global benchmarks exist we say so, and SG net margins are thin to often negative. The MOAT Score is a transparent SGAI judgement on economic quality, not a verdict on an owner-operated livelihood. Verify fees and regulatory steps with each agency before acting.
We're not a grant office — we build AI and intelligence for Singapore SMEs. Talk to us.