May 4, 2026

Grab, Shopee, Rakuten: When The Super App Becomes A Bank

Fintech · Super App · Lending · Float · Risk 5/4/2026
Asia's Pattern
Why Grab, Sea (Shopee), and Rakuten all rushed into finance: the leverage of float, data, and the race to turn every user into multiple revenue streams at once

A ride-hailing company (Grab), an e-commerce company (Shopee), a Japanese online retailer (Rakuten) - three completely different starting points, yet all share one thing: a financial services arm ballooning at breakneck speed. This isn't a coincidence - it's the inevitable economic logic of the super app model once the core business hits a ceiling. And risk is stacking up on a $9 billion loan book that has never once passed through a recession.

+120%
Grab loan book YoY (Q4/25)
+80%
SeaMoney loan book YoY
$9.2B
Monee (Shopee) loan book
¥13.3T
Rakuten Bank deposits

Scope: This piece digs into the tech platform → financial services pattern through three examples: Grab (ride-hailing → GXS Bank + GrabFin), Sea Limited (Shopee + Garena → SeaMoney/Monee), and Rakuten (Ichiba e-commerce → Card + Bank + Securities). Figures are drawn from each company's Q4/2025 filings (SEC 20-F, disclosures to NYSE/TSE), Bloomberg, Business Times Singapore, Fintech News, and Rakuten IR.

Focus: explaining why (the economic logic), how (the float, data, and cross-sell mechanisms), and what risks (young NPLs, duration mismatch, regulatory exposure). This is not investment advice - 80-120% annual growth rates in a high-rate environment, paired with a loan book that has never been through a recession, is a warning, not a buy recommendation.

1 · Three Players - The Same Behavior

Before digging into the logic, a quick look at scale shows this isn't "adding a feature" - the financial services arm has become a core segment, at times bigger than the original business. Three companies starting from three completely different industries, ending up in the same place:

Grab Holdings
Singapore · NASDAQ: GRAB
Started asRide-hailing (2012)
Finance todayGrabFin + GXS + GXBank
Fin revenue Q4/25$99M (+34% YoY)
Loan book$1.18B (+120%)
GXS+GX deposits$1.6B
StatusLoss-making, 2026E breakeven
Sea Limited
Singapore · NYSE: SE
Started asGarena games (2009)
Finance todayMonee (formerly SeaMoney)
DFS revenue 2025$3.8B (+60%)
Loan book$9.2B (+80%)
NPL >90 days1.1%
StatusProfitable, major EBITDA contributor
Rakuten Group
Tokyo · TSE: 4755
Started asIchiba e-commerce (1997)
Finance todayCard + Bank + Securities + Insurance
FinTech revenue 2025~¥920B (+15%)
Rakuten Card GTV¥26.5T (#1 in Japan)
Rakuten Bank deposits¥13.3T (~$88B)
StatusSaved the whole group

Three companies, three completely different starting points (Southeast Asian ride-hailing, online gaming, a Japanese e-marketplace) - yet ending up in the same place: multi-billion-dollar loan books, digital banks with real licenses, cross-sell investment portfolios. That's no coincidence. It's the same economic logic pushing every platform with enough customer traffic in the same direction.

The core pattern to notice. All three started from a primary service that generates 3 data streams: (1) who the customer is - ID, KYC, device, location, (2) how the customer uses money - frequency, value, recipients, (3) whether the customer can repay - payment history, inferred cash flow. When you hold all three while traditional banks don't, you're sitting on a credit goldmine - but you only mine it by lending yourself. Selling data to another bank captures a sliver; lending it yourself captures the entire NIM (net interest margin) - and that is the highest margin business in the entire economy.

2 · The Real Economic Logic: Four Layers Pushing One Direction

Why does every platform large enough eventually build a bank? Four economic pressures push in the same direction - no conspiracy required; each pressure alone is enough to get a board to approve a financial services arm:

Layer 1 · Take-rate ceiling - the core business can't raise prices forever

Grab takes ~20% per ride - push it to 25% and drivers switch to a competitor, riders feel the price sting. Shopee takes ~5-8% commission from sellers - push it to 12% and sellers flee to TikTok Shop. Rakuten Ichiba takes ~10-15% from merchants. All of them have hit the ceiling. As user growth also stalls (Southeast Asia has no sudden population boom; Japan is shrinking), a core business that can't raise prices has to find secondary revenue from the same user. Finance is the only way - it's the only industry where you can multiply per-user revenue by 2-5x without needing more goods or more drivers.

Revenue per user · Grab Ride/Food
~$15-25/year
20% take rate on ~$100/year average GMV per user. Growth has slowed to single digits.
Revenue per borrowing user · Grab Fin
~$80-150/year
A $500 loan at 18-24% annual interest plus fees generates 5-10x the revenue of the core take rate.
Fin users / total users
5-15%
Grab has ~40M MTU, only ~3-5M using GrabFin. Huge room left to raise internal penetration.

Layer 2 · Customer acquisition cost (CAC) is already paid - now it's just multiplying

Traditional banks spend $100-300 to acquire one customer (advertising, account-opening perks, agent commissions). GXS Bank spends ~$8 - because the customer is already inside the Grab app; all it takes is a banner saying "open an account, move your ride credit into your wallet." This is called "sunk CAC arbitrage" - the marketing cost to acquire the user was already paid by the core business (ride-hailing), so any secondary revenue pulled from that same user carries an enormous margin because the marketing-cost denominator is already zero.

Customer acquisition cost (CAC) - Traditional banks vs. super apps
Citibank SG
~$280
Traditional bank
DBS digibank
~$165
Digital bank (own brand)
Trust Bank (Standard Chartered + NTUC)
~$70
Has an ecosystem partner
GXS Bank (Grab+Singtel)
~$8-15
Sunk CAC arbitrage
Monee (Shopee user → loan)
~$3-6
Natural conversion

Estimates based on MAS reporting, Trust Bank IR, and Bain Asia analysis. With an $8 CAC and a per-borrower LTV (lifetime value) measured in thousands of dollars, GXS's theoretical marketing ROI is 50-100x that of a traditional bank - if it can keep NPL low.

Layer 3 · Cross-sell flywheel - each product multiplies the value of the others

This is the part Rakuten has executed best over the past 20 years, and Grab/Sea are now copying it. The mechanism is called the "Rakuten Ecosystem" or the "super app flywheel":

STEP 1
Pay with a Rakuten Card

On Ichiba e-commerce, paying with a Rakuten Card earns +1% Rakuten Points. Card-opening rate within the Ichiba base: ~70%.

STEP 2
Points become currency at Securities

~70% of fund buyers at Rakuten Securities pay with Points. Brokerage accounts open "free" for cardholders.

STEP 3
Open a Rakuten Bank account for payroll

Deposit rate 0.1% higher than Mizuho/MUFG. ~16.83 million accounts, +10.4% YoY - the largest in Japan.

STEP 4
Take a mortgage / buy insurance

Pay with Points. Rakuten Life insurance. A closed loop - the customer never leaves.

Hiroshi Mikitani (Rakuten's CEO) calls this the "vertical and horizontal expansion of the ecosystem." The real numbers: the share of users using multiple Rakuten services rose from 64.9% (2017) to 72.3% (2020) and keeps climbing. The average user uses 4.2 services - compared to a traditional bank, where the average user uses just 1.3 products. That's why Rakuten Bank outpaces every other Japanese internet bank in acquisition speed - not because of better rates, but because it already had 100M Ichiba users on tap.

Why did Sea rename "Monee" - and why does Grab call it "GrabFin" instead of "Grab Lending"?

Sea Limited renamed SeaMoney to Monee in September 2025. The surface reason: shorter, more memorable branding. The real reason: separating the brand from Sea/Shopee to expand outside the ecosystem. Once Monee stands alone, it can lend to users who aren't on Shopee (other marketplaces, other apps), and more importantly, it can successfully spin off as an independent company with its own IPO - similar to how Ant Group separated from Alibaba. This is a "value-optionality" strategy - preserving the right to IPO the fintech arm at a separate valuation when the market is hot.

Grab uses the name GrabFin/GXS instead of "Grab Bank" or "Grab Loans." The reason is regulatory: under MAS rules, "Bank" in a name requires a full banking license - GXS is only a digital bank with a $5,000-per-user deposit cap in its early phase. "Grab Loans" would be too narrowly defined - limiting its ability to do wealth management, BNPL, or insurance. The most neutral choice is "Fin"/"GXS" - easy to reshape in scope.

Both naming choices reveal the same thing: these companies know the fintech arm will eventually outgrow the core business, so they're already preparing the brand architecture for that day.

Layer 4 · NIM - the highest margin in the real economy

This is the final piece - and the most beautiful part financially. The Net Interest Margin (NIM) of Southeast Asia's digital consumer lending industry typically looks like this:

Funding cost → Lending rate → Gross NIM (% APR, scale 0-50%)
The striped zone = funding cost (what Monee/Grab must pay). The solid zone = the net margin (NIM) captured by the platform. Highest in the industry: Indonesian BNPL at 30%+ NIM.
SPayLater BNPL IndonesiaSea / Monee
NIM ~30%
↳ 40% APR
SLoan cash loan PhilippinesSea / Monee
NIM 30-50%
GXS personal loan SGGrab · capped by MAS
NIM ~11%
↳ 18% APR
Rakuten Card revolvingJapan · BOJ "free money"
NIM ~15%
↳ 18% APR
Traditional VN bankcorporate lending · reference
NIM ~4%
↳ 12% APR
JPMorgan USblended portfolio · reference
NIM ~3%
↳ 6% APR mortgage
APR scale
0%10%20%30%40%50%
Funding cost (deposits + debt) High NIM (BNPL/cash loan) Medium NIM (regulated) Traditional bank NIM

A 20-30% NIM on BNPL/cash loans in Southeast Asia is a staggering figure. Compared to Grab's core business (ride-hailing gross margin ~25%, net after driver costs ~3%) or Shopee's (gross take 8%, net -2 to +5%), one dollar invested in lending returns 5-10x the margin. That's why no CFO sitting on top of a super app would not push hard on the financial services arm - that would be financial malpractice.

Why a 25% margin isn't free money. BNPL lending rates in Indonesia are that high not because Monee is greedy, but because it has to provision in advance for the fact that some borrowers won't repay. The simple way to see it: for every 100 units lent out, Monee collects 25 units of interest - but must set aside a portion as a "buffer fund" for uncollectible loans. If in reality only 1-2 units are written off, Monee nets ~23 units (very healthy). If 5 units are lost (the normal scenario once the loan book matures), it nets ~20 units (still fine). But if a shock hits and 10-15 units are lost (a recession scenario), Monee starts posting net losses. That is exactly what happened to Klarna in the US in 2022, LendingClub in 2017, and the entire US subprime mortgage market in 2007.

3 · Grab - From Ride-Hailing To Digital Bank: The $1.6 Billion Deposit Journey

Grab is the clearest case for watching the transformation mechanism unfold. It started in 2012 as MyTeksi in Kuala Lumpur; by 2026 it has:

Q4/2025 revenue
$906M
+19% YoY group-wide. First full-year profit in 2025: $268M (Business Times).
Financial Services segment
$99M (+34%)
Grew 34% YoY last quarter while On-Demand grew only 17%. Twice the pace.
GXS+GXBank loan book
$1.18B
+120% YoY (versus $536M in Q4/2024). $979M disbursed in a single quarter.
SG+MY deposits
$1.6B
Up from $1.2B the prior quarter. Mostly Grab users (drivers, merchants, consumer ecosystem).

The GXS Bank story: why did MAS grant Grab a license?

In 2020, Singapore's MAS decided to award 4 digital bank licenses - the first time it opened the door to non-bank operators. GXS (a Grab + Singtel joint venture) was one of two digital full banks licensed. MAS's reasons for choosing Grab:

  1. Serving the underbanked - Grab holds data on millions of drivers, gig workers, and small merchants that traditional banks never bothered courting.
  2. Competing in SME and micro-credit - a segment DBS/UOB/OCBC care little about because margins are thin and due diligence is expensive.
  3. Pressure to lower costs at traditional banks - one of MAS's core policy goals to keep Singapore's fintech-hub status.

In FY2025, the GXS group (Singapore + Malaysia combined) posted a loss of S$208.1 million, narrowing from S$214.3M the prior year (Business Times). Of that, GXS Singapore alone lost S$132.1M. But net interest income rose 84% to S$55.6M, and the loan book grew 323% to S$1B. The target: profitability by March 2027 - right on the 5-year timeline MAS set for digital banks.

Why does GXS have an expected credit loss ratio of 4.6%? Versus 0.5-1% at traditional banks

4.6% is the real FY2025 figure, down from 6.8% the year before - but still roughly 6x a typical Singapore commercial bank. The reason isn't that GXS lends poorly, but that it deliberately targets a higher-risk segment:

  • Underbanked customers - lacking the thick credit history traditional banks rely on. DBS/UOB decline them; GXS accepts and compensates by charging higher rates.
  • Small loans ($500-5,000) - processing cost per loan is roughly the same whether it's $500 or $500,000, so small loans naturally carry a higher risk margin.
  • A young loan book - up 323% in a single year, with most loans not yet old enough to "mature" into bad debt (see the detailed explanation in Section 7). As disbursement slows, this ratio will rise further before stabilizing.

GXS's survival math is simple: the net interest margin must be large enough to cover the portion of borrowers who don't repay. Currently, the gross margin is ~10-12%, minus an expected 4.6% credit loss rate, leaving a ~6% net margin - just enough to cover operating costs on the path to profitability. But if a Singapore recession (or just a shock to the gig economy - Grab drivers laid off as AVs/robotaxis arrive) pushes losses to 8-10%, the net margin turns negative and the 2027 breakeven timeline slips far out.

For comparison: DBS has an ECL of ~0.4% on a $400B portfolio - but that's the result of 50 years of screening and relationship lending. GXS has been building that risk infrastructure only since 2022, with a loan book growing 4x a year. This is a race between how fast risk models can be built and how fast the book grows.

Why Grab is the best pure-play to watch this pattern

Grab lets us watch the pattern in real time: the transition from ride-hailing to a fintech-driven super app is unfolding right now. In Q4/2024, financial services was just 9.7% of revenue. By Q4/2025 it was already 10.9%. At this pace, by 2028 financial services could account for 20-25% of revenue - and at that point Wall Street will be forced to price Grab at fintech multiples, no longer at gig-economy multiples. That is the ultimate prize of this journey: a 3-5x increase in market cap without needing any more users.

4 · Sea Limited - The Biggest Money Printer In Southeast Asia

Sea Limited is a very different case from Grab: Sea has already succeeded spectacularly at fintech, to the point that Monee now contributes a large share of the entire group's profit. FY2025 Sea Limited:

Full-year 2025 revenue
$22.9B
+36.4% YoY. Net income $1.6B - 3.6x 2024 ($447.8M).
Monee revenue 2025
$3.8B
+60.3% YoY. Q4 reached $1.1B (+50%). The fastest-growing, highest-margin segment.
Loan book as of 12/31/2025
$9.2B
+80.4% YoY. $8.2B on-book + $1.0B off-book. 20 million first-time borrowers in 2025 alone.
NPL >90 days
1.1%
Stable quarter-on-quarter, a slight improvement from 1.2% the prior year. A good-looking number - but one that needs context.

SPayLater - The killer product

SPayLater is a BNPL product built directly into Shopee checkout: buy something for $50, choose to pay over 3/6/12 months. It differs from Klarna/Affirm in the US:

  • Merchants receive 100% of the payment immediately from Monee Capital. Monee bears the entire credit risk.
  • Real-time underwriting - credit decisions made in milliseconds, using Shopee data (purchase history, payment behavior, app usage).
  • Starts with a low credit limit ($50-100), rising gradually with on-time payments. Behavioral underwriting is the core risk model.
  • 1-5% monthly interest depending on tier - effective APR of 12-60%. Some 0%-interest promotions exist for top-tier customers buying promoted items.

The pattern here is quite clever: SPayLater isn't just a lending product - it's also a conversion booster for Shopee itself. Sea's own published research shows: Shopee checkout completion rates rise 27% when a user activates SPayLater. Average order value (AOV) rises 40%. That means even if Monee only broke even on lending, it would still create value for the core business - a kind of strategic loss leader Shopee can well afford.

Why Indonesia is the main battlefield - and why that's Monee's biggest risk

Indonesia accounts for ~50% of Monee's loan book (an estimate, not officially disclosed). The natural reason: Indonesia is Southeast Asia's largest e-commerce market, with a population of 280M and extremely low traditional credit access (only ~30% of adults hold a credit card).

But Indonesia is also where systemic risk runs highest:

  • Total multifinance NPL in Indonesia rose from 3.37% (January 2025) to 3.68% (February 2025) in a single month (Business Times). The upward trend continued through all of 2025.
  • OJK (Indonesia's financial regulator) has tightened BNPL/p2p lending rules since Q3/2025: minimum age of 18, minimum income of IDR 3M/month, restrictions on predatory advertising.
  • Layoffs across Indonesian tech (GoTo, Bukalapak, and Traveloka cutting thousands of jobs in 2024-2025) have pushed a wave of BNPL users into repayment difficulty.
  • The IDR depreciated against the USD by ~8% in 2025 - if Monee funds itself in USD/SGD but lends in IDR, that adds FX risk.

Monee's 1.1% NPL looks good but is blended across the whole portfolio - it doesn't reveal risk concentration in Indonesia or the Philippines. This is exactly the question long-term investors should be asking - and Sea's own management acknowledged as much on the earnings call, noting "we are closely monitoring the Indonesian macro environment."

5 · Rakuten - 20 Years Ahead, And A Survival Lesson

Rakuten is the OG (original gangster) of this pattern - starting in 1997 as a small e-marketplace in Japan, and by 2026 it's Japan's largest fintech conglomerate, with:

Group revenue FY2025
¥2.5T
~$16.7B. +9.5% YoY. The 29th consecutive year of record revenue.
FinTech segment 2025
~¥920B
Q1-Q4 total ~¥912B (Q1+Q2 already ¥456B). +14-16% YoY. Operating profit ¥170-180B.
Rakuten Card GTV
¥26.5T
#1 in Japan for shopping transaction volume. 33.41 million cards issued (12/2025).
Rakuten Bank deposits
¥13.3T
~$88B. Japan's largest internet bank, 16.83M accounts. Consolidated total assets ¥16.9T.

Why FinTech "saved" Rakuten

This part often gets overlooked: Rakuten posted net losses for four consecutive years, 2021-2024, as Rakuten Mobile (its telecom arm) burned cash building a 5G network to compete with NTT/SoftBank/KDDI. FinTech was the only segment that stayed consistently profitable, growing double digits every year. The FinTech segment literally kept the whole group alive during the period Mobile was burning cash.

"FinTech segment achieved both revenue and profit growth throughout FY2025, with Rakuten Card and Rakuten Bank as major contributors. The Mobile segment is on track to break even in FY2026 - allowing the Group to fully realize the synergies of our ecosystem strategy." - Rakuten Group FY2025 Earnings Highlights, February 12, 2026

This is the model's insurance policy - when the core business struggles (Ichiba e-commerce squeezed by Amazon Japan, Mobile losing money), the stable financial arm keeps overall results heading upward. This is also why Grab and Sea insist they'll keep pushing fintech: it's diversification of the company's own cash flow, all within one company.

The quiet lesson from Rakuten Bank's 2023 IPO

In April 2023, Rakuten Group IPO'd Rakuten Bank on the TSE Prime market - successfully spun off as an independent listed company. The ¥80 billion raised helped Rakuten Group pay down debt. This is the model Sea and Grab are now preparing to copy:

  1. Build a fintech business inside the group → raise revenue, raise internal valuation.
  2. Once scale is sufficient, spin it off in an independent IPO at fintech multiples (P/B 2-4x for a bank, P/S 6-10x for a lending fintech).
  3. The parent retains ~60% ownership - collecting the IPO cash while still capturing further upside.

This is the hidden value optionality embedded in every super app's fintech arm. It isn't yet fully reflected in Grab's or Sea's share prices today - but it will be once Sea spins off Monee (rumored for 2027) or Grab IPOs GXS (2027-2028, after reaching profitability).

6 · A Direct Comparison Table - One Page

CriterionGrab HoldingsSea LimitedRakuten Group
Year entered fintech 2017 (GrabPay) 2014 (AirPay) 2002 (Card)
Original core business Ride-hailing Games (Garena) E-commerce (Ichiba)
Fin revenue as % of total ~11% ~17% ~37%
Fin growth rate/year +34% +60% +15%
Holds a bank license? Yes (GXS+GXBank) Indonesia/PH/SG digital licenses Yes (Rakuten Bank, separately listed)
Loan book (2025) $1.18B $9.2B ~¥3T (~$20B, including card revolving)
NPL/ECL 4.6% (GXS) 1.1% blended <1% (Card revolving)
Fin segment profitability Loss-making, 2026E breakeven Profitable, high margin 25%+ profit margin, consistently
Lifecycle stage Scaling Mature, near peak growth Steady state, IPO spinoff already done
Main risk NPL on a young loan book, regulatory Indonesia macro shock, maturing NPL Competition from neobanks, BOJ rates

7 · Risk - The Part The Pretty Story Usually Skips

By this point, the piece has explained why the model makes sense and why growth looks great. But the most important closing part is what's never shown on the sales slide: the risk that's actually stacking up. There are 5 specific risks, in rough order of likelihood of materializing:

Risk 1: 1.1% NPL looks good - but it may be deceiving

Monee reports NPL of only 1.1% on a $9.2B loan book - which sounds safe, even better than many traditional banks. But there's an arithmetic trap in the calculation: bad debt takes 12-18 months to "mature" - a few days late in month one, a week late in month two, only classified as over 90 days past due by month 7-9. A loan disbursed today hasn't had time to go bad in next quarter's report.

That means as the loan book grows fast, the denominator (the loan book) balloons faster than the numerator (matured bad debt) - the reported NPL ratio is artificially low. A simple simulation: the loan book doubles every year, and the "true" NPL of each cohort once it matures is 8%:

The dilution effect - why reported NPL looks low during fast growth
Simulation: loan book doubling every year, "true" NPL of 8%. In Year 4, disbursement growth stalls → the ratio automatically jumps from 4% to 8% with no shock needed.
0% 2% 4% 6% 8% NPL ratio Year 1 $1B book 0% / 0% Year 2 $2B book 4% 8% Year 3 $4B book 4% 8% Year 4 $4B (flat) 8% 8% BOOK FLATTENS → NPL REVEALS ITSELF
Reported NPL (standard calculation) "True" cohort NPL Loan book size

In the first three years, reported NPL looks like just 0-4% - very pretty. But that's not because the loan book is good - it's because the rapidly growing denominator is diluting the ratio. When disbursement stalls (saturation, regulatory tightening, recession), the numerator keeps maturing while the denominator stands still - NPL jumps from 4% to 8% in just 2-3 quarters with no macro shock required.

Applying this to Monee. A $9.2B book growing 80%/year → average loan age of ~6-9 months, not yet old enough to mature. If Sea's growth stalls and the "true" cohort NPL once matured is 4-5% (a normal level for BNPL in the region), the reported ratio could jump from 1.1% to 4-6% within 12-18 months - simply by growth stopping, the truth reveals itself.

On top of this arithmetic trap sits a bigger one: Southeast Asia's entire digital consumer lending industry (2018-2025) has never been through a real recession. COVID-2020 was the only shock, and it was met with massive government stimulus - Monee's and GXS's risk models are calibrated on data from an uninterrupted growth period. History shows how fast NPL can blow up when a real shock arrives:

  • LendingClub US 2017: NPL spiked from 4% to 9% in 2 quarters as the Fed raised rates. The stock fell 80%.
  • Klarna 2022-2023: losses grew from $200M to $1B a year. Valuation fell from $46B to $6.7B (an 85% drop).
  • Andhra Pradesh microfinance 2010: NPL jumped from 1% to 30% in 6 months when the state passed a law restricting debt collection.

Risk 2: Regulatory - governments tighten once they've "seen enough"

The historical pattern is clear: governments always loosen the reins to let fintech grow, then tighten when (a) a major scandal hits, (b) traditional banks lodge political complaints, or (c) systemic NPL rises. Vietnam, Indonesia, the Philippines, and Thailand are all currently in phase (a) or (b):

  • Indonesia: OJK tightened BNPL regulations in Q3/2025 - minimum age, minimum income, advertising limits.
  • Philippines: BSP capped credit card rates at 36% APR (2022). Now proposing similar caps for cash loans.
  • Vietnam: Decree 13/2023 requires consumer credit funds to hold charter capital of ≥VND 500 billion. This effectively wiped out most p2p operators.
  • China: the Ant Group IPO block of 11/2020 - the biggest warning shot for every fintech super app in Asia.

Regulatory risk isn't a question of "if" but "when" and "how hard." Once natural growth stalls, tightening policy will turn growth negative.

Risk 3: Funding cost shock - USD/SGD rates shifting

Most Southeast Asian fintechs fund themselves through a mix of: taking deposits (cheap) + corporate debt (medium cost) + securitization (selling a portion of the loan book to investors). When the Fed raised rates in 2022-2023, funding costs jumped across every tier. If Monee funds 50% via USD corporate debt at SOFR+spread, every 100bp jump adds ~$45M/year in cost on a $9B book - eating straight into NIM.

The same applies to Rakuten: the BOJ is gradually exiting zero rates (already raised to 0.5% in 2025, expected to reach 0.75-1% in 2026). Rakuten Card's funding costs will rise - but it can only pass on part of that to customers, since Japanese card-revolving rates are capped at 18%/year under the Money Lending Business Act.

Risk 4: Customer concentration - lose Shopee, Monee is nothing

An estimated 80-90% of Monee's borrowing users come from the Shopee ecosystem. If Shopee loses market share (TikTok Shop is steadily eating into it) or hits a crisis (regulatory, scandal), Monee's acquisition channel disappears and the loan book shrinks. This is a risk of concentration on a single source.

Grab is similar: GXS depends on Grab's user base. If Singapore's tax policy shifts (as commission caps did in 2023), or if MAS requires Grab to separate its relationship with GXS - the model cracks.

Risk 5: The SVB lesson - duration mismatch at scale

GXS Bank and Rakuten Bank take deposits (withdrawable anytime) and lend/invest them out. Once the book gets large enough ($88B for Rakuten Bank), even a short bank run could force asset sales at a loss. The SVB lesson from March 2023 - 24 hours from "healthy" to FDIC takeover - applies to every digital bank.

Rakuten Bank is fortunate to sit on 25+ years of relationships with Japanese customers (sticky), but GXS, just 3 years old with a gig-economy customer base (not sticky), is a much more bank-run-prone profile. One bad headline on TikTok could trigger a mass withdrawal.

Risk summary to hold onto: A super app's fintech arm is NOT free money. The economic logic is sound (sunk CAC, high NIM, data flywheel), the growth numbers look great (Grab +120%, Sea +80%), but there has not yet been a single real stress test. Every consumer credit model in history has gone through an NPL blowup cycle - from the 2008 subprime mortgage crisis, to the 2010 Andhra Pradesh microfinance crisis, to Klarna in 2022. The question isn't "will it happen" but "when, in which market, and how bad." Monee's $9.2B book and Grab's $1.18B book will be the first real test case for Southeast Asia's generation of digital lenders once a genuine recession arrives.

8 · Zooming Out - Is This "Financialization" At The Corporate Level?

This story isn't new. The four classic markers of corporate financialization that Greta Krippner described back in 2005 played out across a wave of American conglomerates from 1980-2010. Grab, Sea, and Rakuten are now repeating the pattern - just faster, with different tools.

Four markers

  1. Profit from the financial arm exceeds the core business - the contribution ratio inverts.
  2. Capital allocated to finance exceeds core capex - cash flows toward wherever ROIC is higher.
  3. Wall Street prices the company as a fintech, not as the core business - management's incentives reinforce the pattern.
  4. When finance collapses, the whole company collapses - the shrunken core business can no longer carry the weight.

The US precedent, 1980-2010

CompanyCore businessFinancial armPeak contributionOutcome
General Electric Turbines, generators, motors GE Capital (1932 → 1990s boom) ~50% of group profit (2007) The 2008 GFC pushed GE to the brink of collapse; split into 3 companies in 2021.
General Motors Automobiles GMAC (1919) GMAC alone went bankrupt in 2008-2009 GM went bankrupt 6/2009, $50B bailout. GMAC split off into Ally Financial.
Ford Motor Automobiles Ford Motor Credit (1959) ~30-40% of group profit in the 2000s The only one of the Big 3 that did NOT need a 2009 bailout - because it had tightened up FMCC beforehand.
Sears Roebuck Retail Discover Card, Allstate, Dean Witter ~60% of profit in the mid-1980s Sold off the entire financial arm in 1993 - too late. Went bankrupt in 2018.

The common pattern: the financial arm starts out supporting the core business (financing car purchases / appliances), then grows independent, then surpasses the core in profitability, and finally drags the whole conglomerate down when a credit crisis hits. The logic pushing this along: GE Capital generated ~25% ROIC, aviation ~12%, power ~8%. Any board has to allocate capital to the highest-return segment - that's a rational decision, not laziness.

Mapping Grab / Sea / Rakuten against the four markers

MarkerGrab HoldingsSea LimitedRakuten Group
1. Fin profit > core Not yet - fin is still loss-making. Shifting fast. Already there. Monee = 33% of group adjusted EBITDA in Q4/25, over $1B for the full year. Already there, strongly. FinTech consistently profitable, Mobile loss-making. Fin literally kept the group alive 2021-2024.
2. Capital to fin > core capex Shifting. Capital allocated to GXS exceeds ride/food capex. Already there. Monee's loan book grew +$4.1B in a year - larger than Shopee+Garena combined capex. FinTech now self-funds without further injection - already mature.
3. Wall Street prices as fintech Shifting - GRAB has risen to 5.2x sales as the market notices fin growth. Already there. SE trades at ~5x sales. A Monee spinoff could add $20-30B in market cap. Already spun off - Rakuten Bank IPO'd 4/2023 at P/B ~1.8x, valued at $7.4B separately.
4. When fin collapses, company collapses Untested. Ride/food would still cushion a GXS loss. Already dependent. If Monee's NPL rises to 5%, group EBITDA would fall 40-50%. Fin is the cushion. If Bank/Card blows up, the whole group (already carrying Mobile's debt) collapses with it.

A hard-to-avoid conclusion: Sea Limited has completed its financialization - sitting exactly where GE stood in 2005. Rakuten is similar but has already de-risked by spinning off Rakuten Bank as an independent listed company - a shrewd defensive design. Grab is 2-3 years from looking like Sea does today.

% of profit/EBITDA from the financial arm - historical and current conglomerates
Sears 1985 (peak)
~60%
→ bankrupt 2018
GE 2007 (peak)
~50%
→ collapsed 2008
Rakuten Group 2025
~37%
already spun off Bank
Ford 2005
~35%
survived 2009
Sea Limited 2025 Q4
33%
and rising fast
Grab Holdings 2025 Q4
~11% rev
still early in the cycle

These figures use different measures (EBITDA for Sea, profit for GE/Sears, revenue for Grab) - not directly comparable in absolute terms. But the pattern is clear: Sea Limited is sitting in "Ford 2005" territory - dangerous, but still salvageable if it proactively spins off. GE and Sears didn't spin off in time and paid dearly for it.

Two differences that make this pattern play out in 2020s Asia

Same Krippner logic, but 2020s Asia has two structural differences that make the cycle both different and more dangerous:

  • The core business still has growth left. GE in 2007 was pushed toward fintech because aviation/power had already saturated. Sears lost to Walmart. Completely different here: Shopee GMV +27%, Grab Mobility +17%, Rakuten Ichiba stable. This is a protective factor - Asia's super apps have a fallback segment if fintech shocks hit, an option GE didn't have in 2008.
  • Near-zero origination cost → the cycle moves 4x faster. GE Capital had to sell each commercial loan one by one; a super app only needs two clicks from a user to extend credit. GE took 30 years (1980-2008) to reach its fintech peak and then collapse. Sea went from zero to a $9.2B loan book in 8 years (2018-2025). The blowup cycle could move just as fast: one shock pushing NPL from 1.1% to 8-10% within 12-18 months - and Sea's 33% EBITDA contribution could turn into -10% within just four quarters.
A question for investors to hold onto. If you own shares of Sea (SE), Grab (GRAB), or Rakuten (4755.T), you don't own an "internet company." You own a fintech with a legacy internet arm attached. The valuation, risk, and stock behavior will increasingly resemble a Southeast Asian bank (credit risk) more than a growth-tech stock. Read their quarterly reports with the same lens you'd use for DBS or JPMorgan - cohort-based bad-debt ratios, net interest margin, deposit stickiness - not GMV, MAU, or take rate. That is the shift in valuation approach happening right now.

This connects to the piece "The Financialization Disease: America Today, You Tomorrow": the same mechanism operates at the level of empires. The same economic logic is entirely rational at the moment each decision is made - no one is wrong to choose the higher ROIC - but the sum of all those rational decisions pushes the system toward a fragile state. When the shock hit, GE in 2008 could not simply "go back to making jet engines"; if Sea or Grab find themselves in a similar spot by 2030, they won't be able to "go back to pure ride-hailing / e-commerce" either - because by then competitors will have eaten the entire core market share while they were focused on fintech.

9 · Summary For The Reader

Back to the original question: "why do Grab, Sea, and Rakuten all rush into finance?" - it isn't float, isn't tax, isn't hidden leverage. The reason is a completely rational four-layer economic combo:

  1. The core take rate has hit its ceiling - it can't raise prices further on ride-hailing / e-commerce without losing users.
  2. Sunk CAC arbitrage - marketing cost has already been paid by the core business, so any secondary revenue on the same user carries an enormous margin.
  3. Cross-sell flywheel - each product multiplies the value of the others (Rakuten has proven this over 20 years).
  4. 20-30% NIM - digital consumer lending margins in Southeast Asia run 5-10x higher than the core business.

The worrying part left out of the bullish pieces: the ~$15B combined loan book of the three players has an average age of under 1 year and has never been through a real recession. Monee's 1.1% NPL is pretty, but it's only half the story - the other half only shows once disbursement stalls and the denominator stops diluting the ratio. That isn't a reason to avoid these stocks, but a reason to read quarterly reports by cohort, by net NIM after loss provisions - not to treat the headline number as a proxy for the risk of the next 2-3 years.

In the end, this is a story of valuation arbitrage. Wall Street pays a P/S of 5-8x for fintech and 1-2x for gig economy - whoever shifts the most revenue into the fintech segment gains the most market cap without needing a single new user. That's why every super app is heading the same direction. And that's also why it's worth asking: when everyone is running the same playbook, where is systemic risk piling up?

Read next

More from the shelf

Jul 24, 2026Why Losses Lead to Riskier BetsJul 20, 2026Arbitrage: Why Do Markets Always Misprice?Jun 4, 2026Nassim Nicholas Taleb: Panic Early Is Better Than Panic LateApr 22, 2026Banks' Blockbuster Profits: Unfair, or Just Inevitable?

Pass it on

If it found you, share it kindly

XEmail

03 Discussion

Leave a note

A considered space for questions, counterpoints, and useful additions. Civil, on-topic, signed.

Reader notes

...

Loading notes...