Singapore’s five digital banks filed their 2025 accounts last month, and between them the four retail and SME banks lost S$291m. GXS lost S$132m, MariBank S$55.6m, Trust Bank S$53.6m and ANEXT S$49.8m. The fifth, Green Link Digital Bank, made S$16.1m, its first profit, on a business that lends to small and medium-sized enterprises and never set out to gather retail deposits.
That split is the story of the licences. Three of the four loss-makers have solved the problem that was supposed to be hard, which is persuading Singaporeans to move money to a bank with no branches. What none of them has solved is what to do with it.
The balance sheets
Trust Bank holds close to S$4bn of deposits, the largest base of the group, from around one million customers. Its loans and advances crossed S$1bn during the year. For every dollar a customer has deposited, Trust has managed to lend roughly a quarter, and it took until March 2026 to record its first profitable month. ANEXT, Ant’s wholesale bank, passed S$1bn of deposits for the first time and finished the year at S$1.15bn, while its loan book shrank to S$636m, the first decline since it opened. MariBank, Sea’s bank, has S$1.9bn of deposits and income of S$37.3m, which covers about two-thirds of its annual loss.
GXS, the Grab and Singtel bank, is the one that grew its book fastest. Loans more than tripled to S$814m and total income rose by half to S$44m. It still lost S$132m, the largest loss in the group, because a book of that size does not earn enough to pay for a bank built to serve millions. Grab reports US$2.5bn of deposits across GXS, its Malaysian bank and Indonesia’s Superbank, and does not break Singapore out, so GXS’s own deposit figure is not public. The pattern across the other three is enough to infer it.
Deposits arrived because the banks paid for them and because Grab, Shopee and FairPrice put a savings account in front of customers who were already in the app. Loans did not arrive because the borrowers the banks wanted were already somewhere else. Singapore’s unsecured market is small, MAS caps unsecured credit at a multiple of monthly income, and DBS, OCBC and UOB hold the salary accounts, the credit limits and the revolving balances that make a retail bank profitable. A customer who is already fully limited at an incumbent has nothing to bring to a challenger except a deposit, and a deposit that cannot be lent is a cost that arrives every month.
The bank that lent first
Green Link Digital Bank is the counter-example inside the same regime. Its shareholders are Greenland and Linklogis, its customers are MSMEs in supply-chain finance, and its 2025 operating income was S$71.8m, of which S$54.3m was net interest income. It went from a S$5.1m loss to a S$16.1m profit in a year, on a business that starts with an asset and then funds it, in that order.
ANEXT is the closer comparison, since both hold wholesale licences and both lend to SMEs. ANEXT’s loan book went backwards in 2025 while its deposits kept growing, and it now says it will break even in 2027. Same licence, same market, same borrower segment; what separates them is whether the bank has an origination channel for loans that works as well as the one it has for deposits.
Cards, and what comes after
All three retail banks reached the same conclusion this year. GXS launched a credit card and Trust relaunched its existing card with a stockback option within days of each other in August, and MariBank, which launched its card two years earlier, cut its cashback rate in January. A card is the only consumer product that is a payment, a loan and a daily habit at the same time, and it is the fastest way to turn a deposit customer into a revolving borrower. The rewards being offered are at or above what the transactions earn, so the cards are acquisition budgets for a lending book that does not yet exist. Whether they build one is the question for 2026 accounts.
Hong Kong’s virtual banks reached year six with three at breakeven and five still losing money. Indonesia’s digital banks are mostly profitable because they are lending arms of platforms that already had borrowers. Singapore is the slowest of the three markets to reach profit, and the reason is visible in the loan-to-deposit ratios above. Hong Kong’s banks at least had a large unsecured market to attack; Indonesia’s had captive borrowers; Singapore’s have neither.
What the shareholders decide next
The capital behind these banks belongs to Grab, Sea, Ant, Standard Chartered and Greenland, and each of them now has four years of evidence. Sea’s response is already visible: MariBank’s Singapore business was going backwards before either competitor launched a card, and the improvement in Sea’s banking numbers is coming from its new Philippine bank. ANEXT has given itself until 2027. Standard Chartered has a bank with a million customers, S$4bn of deposits and one profitable month.
For the incumbents, the accounts confirm what the loan books already showed. Four years of digital bank competition has moved deposits, at a price the challengers are paying, and has barely touched the credit relationships where the profit sits. A card book that revolves would change that. On the 2025 numbers, though, the digital banks’ problem in Singapore is the mirror image of the one they set out to solve. Nobody needed to teach them how to take money in. What they still have to learn is how to lend it out.