I keep thinking about a number I read last week: over 70 banks in China have shut down their credit card centers since the start of 2025. Guangfa Bank’s Zhengzhou center went from three or four hundred people at its 2022 peak to less than a third of that by early 2026. The rest were not transferred. They were just gone.
This is not a story about a bad quarter. It is about what happens when an entire industry realizes it no longer needs its own workforce.
The 2025 annual reports for all 42 A-share listed banks are out. Total revenue grew 1.38%. Net profit grew 1.44%. But total assets expanded 9.54%. The banks are getting bigger while earning barely anything extra. The phrase analysts use is “expanding the balance sheet without expanding income.” What they mean is: the old model is broken and everyone knows it.
The old model was simple. Take deposits at low rates. Lend at higher rates. Pocket the difference. That difference, called net interest margin, has collapsed. In 2010 it accounted for 81% of bank income. By the end of 2025 it was 72.62%. The latest reading puts it at 1.40%, down from 1.43% a year earlier. Three-year fixed deposits at major banks now pay less than 2%. Meanwhile people are borrowing less, not because rates are too high, but because they do not trust what their income will look like next year.
I noticed something in the annual reports that felt personal. Multiple banks announced they would keep cutting marketing budgets in 2026. Close more out-of-town branches. Freeze or reduce salaries. The language is always the same: “tighten our belts,” “operational efficiency,” “digital transformation.” What it means is that a job at a bank is no longer the “golden rice bowl” my parents’ generation chased.
The AI part is what makes this different from a normal downturn. AI loan review and AI customer service now handle more than 90% of basic interactions at some banks. The counter staff I remember from my childhood, the ones who stamped passbooks and counted cash by hand, have been replaced by systems that do not need lunch breaks or housing allowances. The banks that still have them are closing physical locations as fast as they can.
There is a strange irony here. The big state banks are using AI to squeeze out the very middle-class stability they once symbolized. A job at ICBC or China Construction Bank used to mean a decent apartment, a reliable pension, and social respect. Now those same institutions are deploying the technology that makes those jobs unnecessary. The 2025 reports show that the largest banks are using tax records, utility bills, and transaction data to build risk models that approve loans without a human ever looking at the application. Many smaller banks still rely on “the branch manager’s judgment.” The gap is not closing. It is widening.
The retail banking story is especially bleak. For a decade the industry mantra was “retail is king.” Credit cards, consumer loans, personal business loans, mortgages. Banks chased consumers because retail paid higher rates and had lower default rates. That story is over. In 2025 corporate loans grew 10.33%. Personal loans grew 1.03%. Household debt leverage peaked at 62.3% in March 2024 and has fallen to 59% as of March 2026. Chinese families are actively reducing debt. They are not “unwilling to borrow.” They are unwilling to bet on a future they cannot see clearly.
The risk is showing up in the numbers. Among 29 banks that disclosed it, the average non-performing rate on personal loans was 1.71% in 2025. Twenty-seven of those 29 saw increases. Zheshang Bank’s retail business is a stark example: 103.4 billion yuan in revenue, negative 5.2 billion yuan in profit. The retail portfolio that was supposed to be the growth engine is now a drag.
I do not know why but I keep returning to that Guangfa Bank credit card center in Zhengzhou. Three hundred people to less than a hundred in four years. The building is probably still there. The desks and chairs might still be there. But the work they used to do, approving card applications and fielding customer complaints, is now mostly handled by algorithms that run overnight and do not ask for overtime pay.
The government has a phrase for this: “financial institutions should reduce quantity and improve quality.” In 2026 alone, 72 rural village banks have been dissolved or merged. Fourteen disappeared in April alone. The policy logic is sound. Consolidate weak institutions. Let the strong get stronger. But the human cost is measured in people who showed up to work one morning and were told their branch no longer existed.
What strikes me is the quietness of it all. There are no protests. No viral social media campaigns from laid-off bank workers. Just a steady, almost polite contraction. One quarter the branch is open. The next quarter it is a self-service kiosk. The quarter after that, nothing.
Maybe this is what a mature digital economy looks like. The platforms that once created millions of delivery driver and customer service jobs are now replacing the white-collar positions that were supposed to be safe. The banks that financed China’s construction boom are now using the profits from that boom to buy servers that make their own employees redundant.
I do not have a neat conclusion. The banking data for 2025 is what it is: growth without prosperity, expansion without hiring, efficiency without mercy. If you know someone who still works at a bank, maybe check in. Their job might be in an AI training dataset already.