Can't outlast the interest rate spread, what will banks rely on in the next decade?

Ask AI · How can banks use digital intelligence to solve the problem of narrowing net interest margins?

Intro: Use digital intelligence to comprehensively rebuild the logic of bank operations, asset structure, and service models.

After the “golden decade,” China’s banking industry has entered a period of deep transformation.

With macroeconomic structural adjustments, further deepening of interest rate liberalization, tightening financial regulation, and the shock of a technology revolution all converging, four pressures—low interest rates, low net interest margins, high risk, and strong regulation—have continued to loom within the industry.

As of March 29, the 13 banks that have already disclosed their financial reports saw their 2025 average net interest margin narrow to around 1.5%, down 10 bps year over year, and by nearly 60 bps compared with 2020. Data from the National Financial Regulatory Administration shows that in 2025, Chinese commercial banks cumulatively achieved net profits of about RMB 2.4 trillion, up about 2.3% year over year.

Against the backdrop of overall industry pressure, bank differentiation is intensifying, and the traditional scale-expansion model can no longer be sustained.

Some banks cling to traditional businesses and get trapped in dilemmas such as a shortage of assets, high liability costs, and rising risk. Others attempt to break the deadlock through digital transformation.

By May 2025, large and medium-sized banks had set up dedicated technology-finance departments at both headquarters and provincial sub-branches, with 2,178 technology branches nationwide.

According to industry research, nationwide commercial banks have generally formulated digital strategies, and most large and medium-sized banks have established dedicated digital departments. However, many banks’ digital initiatives still remain at the early stage of “replacing human work with tools”—online channels, mobile office, and electronic approvals.

McKinsey’s 2025 Global Banking Report shows that currently, financial institutions using AI agents at scale worldwide are still less than one-tenth. But as AI technology becomes deeply embedded in core business processes, forming an intelligent closed loop of “sense-decision-execution,” leading banks are expected to reduce operating costs by 15%–20% and increase return on shareholders’ equity by 4 percentage points.

Some banks have begun to explore deeper transformation paths. Digital intelligence transformation has moved from a “multiple-choice question” to a “mandatory question.”

In 2024, China Merchants Bank launched its “AI + Finance” strategy, applying artificial intelligence to scenarios such as intelligent investment advisory and risk control. Ping An Bank, leveraging the group’s technology strengths, built an AI smart service system to promote deep integration of artificial intelligence and financial services. Industrial and Commercial Bank of China has built a “Digital ICBC” ecosystem to drive end-to-end business digitization.

Industrial Bank, meanwhile, uses digital intelligence as the core engine—reconstructing business logic, optimizing the asset structure, and laying out emerging tracks—to achieve resilient growth across cycles. Its 2025 revenue and net profit have shown consecutive positive growth for two years, total assets have surpassed RMB 1.1 trillion, ranking it second among joint-stock banks. Its non-performing loan ratio has remained stable at a low 1.08%, making it a typical example of digital intelligence transformation among joint-stock banks.

Bid farewell to the “golden era”

Over the past decade, China’s banking industry relied on the dividends from urbanization and industrialization, achieving rapid growth by depending on credit deployment, net interest margin income, and traditional corporate business. This formed an operational inertia of “heavy scale, heavy collateral, heavy offline presence, heavy traditional industries.” As the economic development model shifted, this mature model quickly became ineffective, and the narrowing of net interest margins became a common challenge across the industry.

Since 2020, the LPR has been cut multiple times and deposit interest rates have advanced in marketization, pushing bank asset-side yields steadily downward. Meanwhile, the liability-side cost has remained relatively rigid, and net interest margins have been shrinking year by year.

Among them, state-owned large banks have relatively stronger margin resilience thanks to their branch networks and customer base. Joint-stock banks and smaller banks lack scale advantages, so their margin declines have been more pronounced.

In 2025, net interest margins at state-owned large banks generally fell below 1.5%, while joint-stock banks’ net interest margins generally stayed above 1.6%. However, net interest income still faces downward pressure. To stabilize margins, most banks have been forced to compress liability costs. But the trend of deposits becoming more time-based and longer-term is clear. Even a “rate-increase-for-deposits” form of internal competition has emerged among some smaller banks, further squeezing profit space.

During the economic transition period, financing demand from traditional sectors such as infrastructure investment, real estate, wholesale and retail has slowed down, and the supply of high-quality assets has been insufficient, leaving banks trapped in an asset shortage where “there is money but it is hard to invest.” At the same time, traditional credit models that rely on “reviewing statements and reviewing collateral” struggle to fit the characteristics of emerging industries that are light-asset, high R&D, and high-growth. As a result, financial supply in national strategic areas such as technology, green initiatives, and high-end manufacturing is severely lacking.

If you have high-quality, stable, low-risk real-economy assets, you can take the initiative in industry reshuffling. But traditional asset operation models heavily depend on human experience, making it difficult to achieve precise pricing and dynamic risk control.

The traits of science and technology innovation enterprises—light assets, lack of collateral, and high risk—create a natural conflict with traditional credit evaluation systems. Many early-stage tech startups have historically been unable to obtain bank financing due to insufficient collateral. On the other hand, risks continued to be released in traditional areas such as real estate and local financing platforms. Some banks saw their non-performing loan ratios rise, increasing pressure for risk resolution, resulting in a situation where “asset shortages coexist with high risk.”

Although nationwide commercial banks have all formulated digital strategies, the vast majority still remain in the initial stage of tool substitution for human work, focusing only on process optimization and cost reduction and efficiency improvement, without embedding digital intelligence into the core logic of business operations.

For example, some banks have rolled out iterations of mobile banking and intelligent customer service. But credit approval, risk assessment, and customer management still rely on human experience. Some banks have invested heavily in building technology systems, but data-silo issues are serious. Corporate and commercial, retail, and interbank data cannot be fully connected, making it hard to achieve precise marketing and intelligent risk control.

Compared with overseas peers, international banks such as JPMorgan and Goldman Sachs have already applied AI to core areas including transaction pricing, risk modeling, industry research, and sentiment analysis. Domestic banks’ digital intelligence transformation still shows a certain gap.

Bank business structures are highly convergent. Corporate business focuses on large state-owned enterprises and real estate, while retail business focuses on mortgages and credit cards. Intermediary business depends on payment and settlement and agency sales, lacking differentiated competitiveness. With high-quality customers limited, banks fall into price wars and resource wars, and overall earnings continue to decline.

Meanwhile, emerging financial institutions such as internet banks and consumer finance companies, leveraging lightweight operations and digital intelligence advantages, are taking market share in retail credit and small and micro finance, further diverting customers from traditional banks.

The industry has generally recognized that transformation is no longer a “multiple-choice question,” but a “mandatory question” tied to survival. However, how to break the deadlock has become a challenge facing every bank.

Breaking the deadlock on “asset shortages”

The 2026 National Two Sessions have pointed the way forward. In the government work report, the “Report on the Work of the Government” first proposed “building a new form of smart economy,” and wrote “computing-power and electricity coordination” into new infrastructure projects. “The 15th Five-Year Plan” grid investment is expected to reach RMB 4 trillion, and green hydrogen, new energy storage, and other initiatives were included for the first time among six major future industries.

In the field of technology finance, the government work report clearly calls for “strengthening finance for the full chain and full lifecycle of technological innovation,” requiring financial institutions to shift from “seeking short-term gains” to a patient-capital model of “long-term accompaniment.” For science and technology enterprises in key core technology areas, a normalized “green channel” is to be implemented for listing and financing, as well as merger and acquisition restructuring.

Under this policy direction, the banking industry continues to explore how to use digital intelligence tools to empower industrial finance and to reshape banks’ understanding and methods of pricing industrial assets.

Taking technology finance as an example, more and more banks are converting indicators of light assets—such as patents, R&D, technological barriers, and research teams—into “technology credit assets” that can be quantified, credit-enabled, and risk-controlled, truly enabling full-lifecycle financial support for technology innovation enterprises.

For example, China Construction Bank has established a dynamic innovation capacity evaluation system centered on four-dimensional factors: talent, technology, funding, and market. It has launched products such as “Good Innovation Loans” and “Good Technology Loans.” Industrial and Commercial Bank of China leverages the ICBC Group’s “debt + loan + guarantee + equity” advantages to build a full-lifecycle service matrix for technology-oriented enterprises. China CITIC Bank has rolled out a fully online pure-credit “Tech-e Loan,” using enterprise innovation points and technology innovation qualifications as key bases for credit granting.

Industrial Bank uses a “technology flow” evaluation system as its core approach, starting from technology innovation capabilities to provide precise credit to technology enterprises. By the end of 2025, Industrial Bank’s “technology flow” had served more than 365k technology enterprise customers.

In the manufacturing sector, Industrial Bank’s manufacturing loans are nearly RMB 1 trillion, far exceeding the industry average. At the same time, its corporate real estate financing outstanding balance declined by more than RMB 50 billion year over year, with funds flowing precisely to the core areas of the real economy.

These achievements in structural optimization all reflect the precision pricing and dynamic risk-control capabilities enabled by digital intelligence.

From digital intelligence to smart intelligence

Faced with industry challenges, leading banks have all started transformation explorations. The core directions focus on four dimensions: digital intelligence upgrades, optimization of asset structure, layout of emerging tracks, and integrated operations.

State-owned large banks, leveraging their scale advantages, are pushing digital intelligence across the entire spectrum. Industrial and Commercial Bank of China, China Construction Bank, Agricultural Bank of China, and other state-owned large banks, backed by strengths in funding, customers, and branches, use digital intelligence as the core lever for end-to-end upgrades.

Retail-focused joint-stock banks concentrate on retail digitization, building strengths in wealth management. Joint-stock banks with retail at the core, such as China Merchants Bank and Ping An Bank, tightly bind digital intelligence with wealth management and retail credit.

Industrial finance-oriented banks go deep into the real economy, laying out technology and green tracks.

Different from the end-to-end coverage of state-owned large banks and the C-side focus of retail banks, the transformation core of this type of bank is to use digital intelligence to address pain points in industrial finance, achieving differentiated competition through featured tracks.

Industrial Bank, represented by an industrial finance specialty bank, has stepped out of the traditional corporate-business framework and the “shallow-ization” misconception. With digital intelligence as the bridge, it connects emerging industries such as technology, green sectors, and manufacturing, and reconstructs the underlying logic of operations.

The leap from “digitization” to “smart intelligence” is the high point of competition for the banking industry.

In the view of Lü Jiajin, Chairman of Industrial Bank, digital transformation is a fight for life and death. The bank has elevated digital intelligence to a strategic core, established an AI + Action Leading Group, and has invested about RMB 8 billion in technology for three consecutive years, with a technology workforce of over 8,000 people. Currently, enabled by digital intelligence, growth in high-net-worth customers is over 12%, the IT delivery cycle has been reduced by over 30%, and cumulative AI smart marketing has reached 21.39 million people.

The results of digital intelligence are also directly reflected in financial indicators.

Industrial Bank’s cost of interest-bearing liabilities has decreased by 43 bps year over year, and its net interest margin has remained at 1.71%, with a decline far better than the industry average. Net interest income has shown consecutive positive growth for three years.

On the asset side, digital intelligence has driven deep optimization of its credit structure. It has said goodbye to the traditional reliance on “real estate + infrastructure,” shifting to three strategic areas: green, technology, and manufacturing. In 2025, its green loans reached RMB 1.1 trillion, up 19.05% year over year; its manufacturing loans were nearly RMB 1 trillion, up 15.10%; and loans in the real estate sector declined significantly.

In the field of technology finance, to address the fact that traditional credit models are difficult to adapt to technology enterprises with “light assets” and “high growth,” Industrial Bank developed a “technology flow” evaluation system, assessing enterprises across 15 dimensions including invention patents, research teams, and technological advantages.

In 2025, the “technology flow” approved credit amount was RMB 1.15 trillion, and the technology loan balance was RMB 1.12 trillion, ranking first among joint-stock banks. The non-performing loan ratio was only 0.85%. The bank has become a “growth partner” for technology enterprises by enabling linkage between AIC (financial asset investment company) equity, bond, and loan cooperation; in the same year, it deployed RMB 24k.

Digital intelligence is moving from the back office to the front desk, from a tool to an engine. This transformation trend has already become a consensus in the industry.

China Construction Bank’s “smart government services” is embedded into government services; Agricultural Bank of China’s “smart countryside” serves rural customers; and Bank of China’s “smart cross-border” helps companies go global.

And Industrial Bank’s transformation achieved a qualitative change—from tool application to paradigm reconstruction.

It uses digital intelligence to connect corporate and commercial finance, retail, and interbank data, driving the integrated development of “industrial finance + ecosystem services.” After the CRM system went live with an industry map, it enabled an upgrade from “single-point customer acquisition” to “ecosystem-based customer expansion.” It covers an objective pool including 1,800 customer lineages, 2023 core customers, and 175k customers.

This transformation paradigm of “digital intelligence + industry + ecosystem” makes joint-stock banks represented by Industrial Bank potentially break free from homogeneous competition and form core competitiveness that can endure across cycles.

The answers to a mandatory question

Digital intelligence is not just a technology upgrade—it is a comprehensive reconstruction of operating logic, asset structure, and service models. In future banking industry transformation, three major trends may emerge:

First, digital intelligence will become banks’ core competitive strength, deeply embedded in all-business-process flows. In the future, AI, big data, and cloud computing will become banks’ underlying infrastructure. From customer marketing, credit approval, and risk control to investment and trading, and operations management, the entire process will be made intelligent.

Banks that still remain online-based and tool-based will be gradually eliminated by the market. Banks that can deeply integrate digital intelligence with core business operations will gain a competitive advantage.

Second, specialized tracks will become the key to breaking the deadlock, bidding farewell to homogeneous competition. Industry differentiation will continue to intensify: state-owned large banks will focus on end-to-end services, joint-stock banks on featured tracks, and small and medium-sized banks on regional and niche market segments. Technology finance, green finance, wealth management, and industrial finance will become core areas for joint-stock banks. Only by deeply cultivating these tracks and forming differentiated advantages can they navigate through the cycle.

Third, making real-economy services a fundamental part of transformation, with value creation replacing scale expansion. The banking industry will completely bid farewell to the old logic of “scale first” and move toward a value bank model balancing “quality, efficiency, and risk.”

Around national strategies such as modern industrial systems, scientific and technological innovation, and green development, aligning economic benefits with social benefits will become the core direction of bank transformation.

At a new starting point for development, many banks have already set digital intelligence as a strategic direction for the “15th Five-Year Plan.”

China Construction Bank proposed a “Digital CCB” strategy; Industrial and Commercial Bank of China is advancing a “Digital ICBC” upgrade; Agricultural Bank of China has laid out a transformation toward a “Smart ABC”; and Industrial Bank has explicitly defined a strategic direction based on “five modernizations”—digital intelligence, green transformation, internationalization, integrated operations, and ecosystem-based development.

As AI technology accelerates iteration, the global banking industry stands at a crossroads of digital transformation. Some banks choose “tool-based” approaches, viewing AI as a means to reduce costs and increase efficiency. Others choose “track-based” approaches, viewing AI as a direction for investment and deployment.

But the truly forward-looking choice is “paradigm-based”—using digital intelligence to reshape banks’ operating logic, and making smart intelligence a foundational operating system from strategy all the way to execution.

Digital intelligence is not a multiple-choice question—it is a mandatory question; it is not icing on the cake—it is the foundation for survival.

For all banks, only by abandoning short-term thinking and deeply focusing on digital intelligence and real-economy services can they stand firm in industry change and truly build themselves into world-class value banks with the capability to endure across cycles.