Financial Sector Modernises Core Systems as OceanBase Rises

Hong Kong Fintech Week kicked off this week, focusing on “Finance+”, “AI+” and overseas business expansion. In the latest Global Financial Centres Index, Hong Kong has once again topped the rankings as the world’s leading fintech hub. Trading, cross-border payments, digital assets and artificial intelligence applications have all taken turns in the spotlight. Yet behind the buzz, a more fundamental question is emerging: as financial innovation moves closer to core business operations, are the underlying systems ready?

Speaking in a Monday conversation with Bosco Fung, OceanBase’s General Manager for Hong Kong and Macau, Charlie Dai, a Hong Kong-based Vice President and Principal Analyst at Forrester, noted that financial innovation in Hong Kong has accelerated markedly over the past two years and is becoming increasingly embedded in core business. AI is moving from peripheral functions such as customer service and knowledge management into risk management, daily operations and even end-to-end business process redesign, while digital assets are becoming more institutionalised and scaled. These trends are ultimately flowing down to the infrastructure layer. “Modernising core systems is no longer just an internal IT matter; it is a prerequisite for financial institutions to capture the next wave of growth,” Dai said.

Three Tests Beneath the Boom

In Dai’s view, there are three lenses through which to judge whether a core system is fit for purpose.

The first is always-on business operations: financial services are shifting from batch, offline processing to real-time, always-on operations. Systems must therefore support longer service hours and handle critical transactions more promptly.

The second is continuity: once accounts, payments and trading run around the clock, any system failure can quickly spill over to customers and the business. Core system stability is no longer just a technical metric; it is part of business continuity.

The third is intelligence: as AI penetrates deeper into business processes, demand for real-time data access is rising sharply. Systems must not only record every transaction reliably, but also support immediate data reading and analysis.

These pressures are being amplified by sharp swings in transaction volumes and sustained data growth. Dai pointed out that financial business does not grow in a straight line; market conditions, unexpected events or new lines of business can create hard-to-predict spikes in a short time. At the same time, the volume, variety and real-time requirements of data are all climbing. He summarised enterprise demands in three areas: reliability, to reduce the impact of single points of failure, ensure data accuracy and consistency, and allow upgrades without disrupting business; scalability, to make capacity expansion routine and achievable without interrupting operations; and elasticity, to flexibly allocate resources according to actual load and prevent one busy workload from dragging down the core system.

He also highlighted an easily overlooked risk: past capacity planning assumed that future growth could be predicted with reasonable accuracy. In an environment where financial innovation is accelerating and AI agents are widely deployed, those forecasts may well break down. IT infrastructure often takes time to build, while financial products must be launched quickly, creating a clear mismatch in pace.

OceanBase, which originated from Alipay, has an architecture that maps closely to these three demands. Fung said that from the extreme traffic of Alibaba’s Singles’ Day shopping festival to today’s support for more than 4,000 financial and public-sector mission-critical scenarios, OceanBase’s architecture has been honed through solving its own toughest engineering challenges. Its native distributed design allows customers to scale capacity up or down on demand without downtime, while cross-datacenter and cross-region redundancy arrangements aim to ensure no data loss and second-level service recovery in the event of an incident. This echoes Dai’s repeated point that stability and data accuracy are both non-negotiable.

Transactions and Analytics Moving Closer Together

Beyond handling massive transaction volumes, the financial industry is intensely focused on risk management. Scenarios such as risk monitoring, customer analytics and know-your-customer checks all depend on timely data access. Dai observed that by the time a payment is completed, systems are often already expected to identify risk and even deliver personalised services, yet in reality data is frequently scattered across different systems.

He argued that traditional architectures separate transaction processing from data analytics, relying on repeated data movement to connect the two. This introduces synchronisation delays, lengthy pipelines, duplicated storage and the cost of maintaining consistency. “When a business needs to make real-time decisions, the distance between a transaction occurring and the data becoming available directly weakens service capability.” The value of risk assessment is highly dependent on data freshness; if data cannot reach the risk system immediately after a transaction, even the most accurate model may miss the optimal intervention window. The same applies to customer insight: firms need to see what a customer is doing now, not an outdated historical portrait.

He therefore favours a direction in which transactions and analytics are handled as far as possible on the same data foundation, bringing analysis closer to where the data is generated and shortening the “transaction–judgement–action” loop. At the same time, structured records, documents and expert experience should be integrated so that systems can see more completely and judge more accurately.

OceanBase’s integrated architecture is one attempt to follow this path. According to Fung, OceanBase aims to use a single stack to handle both transactions and real-time risk analytics, reducing reliance on multiple vendors’ technologies and data-movement tools, and cutting the instability that can arise from layering multiple systems. He added that this integrated capability will be extended further into AI scenarios, supporting unstructured data processing and providing database administrators with a simpler, more efficient management experience so that underlying data operations can keep pace with business and AI iteration.

Laying the Foundation for the AI Era

If the earlier discussion settles today's immediate challenges, the next round of competition will be determined by preparations for the AI era. Dai outlined a clear trajectory: predictive AI a decade ago, generative AI three years ago, and AI agents over the past two years, all of which are creating a paradigm shift for the global financial industry. His core argument is that financial institutions’ infrastructure priorities must shift from pure “model infrastructure” to the “data and context infrastructure” needed to support AI-driven decisions.

The logic is straightforward. AI agents are no longer just generating content or answering questions; they must understand tasks, plan steps and invoke tools. The quality of their judgement depends heavily on whether they have the right context. This requires underlying capabilities in several areas: integrating diverse data sources to connect business facts, documents and expert experience; ensuring real-time, monitorable data freshness so that AI works from the current state rather than an outdated copy; and enforcing identity-based, least-privilege access and audit controls, with Forrester’s zero-trust framework serving as the guiding principle. He stressed that the starting point must remain business value: begin with high-value scenarios where data is already ready, build reusable capabilities, and only then support scaled deployment of AI agents.

OceanBase’s recent moves broadly follow this direction. Fung said that enterprise demand for AI has moved from pilots to production, and from internal use to customer-facing scenarios that require regulatory reporting. OceanBase is therefore extending its financial-grade resilient foundation into a platform that can manage both traditional data and the diverse data types required by AI, allowing customers to use one technology stack while retaining core-system stability. Complementary tools will support AI-oriented data governance and compliance, enabling business users to query the system in natural language and rapidly build their own AI scenarios.

The Bottom Line: Upgrades Must Pass an Economic Test

For financial decision-makers, the real question has never been whether to upgrade, but how to justify the return on investment. Dai pointed to two litmus tests. The first is fit-for-purpose validation: because institutions differ widely in transaction patterns, data scale and peak-to-trough characteristics, they should test application compatibility, mixed workloads, node failures and network anomalies in environments that closely resemble their own. The second is economic value assessment: benefits from resource efficiency, operational efficiency, business value and business continuity should be weighed against direct costs, migration costs and full lifecycle costs.

Citing a Forrester study, he noted that a composite model based on four enterprises showed database cost per registered user falling by about 54 per cent in the first year after migration, with further declines in the second and third years. But he cautioned that this does not mean every customer’s costs will be halved; the gains come from improved resource efficiency and business scale, not simply lower software prices. He suggested that Hong Kong financial institutions borrow the value sources and calculation methods from the study, then apply them to their own growth paths and constraints.

This assessment framework finds a practical reference in OceanBase’s experience. Fung said that OceanBase’s high compatibility with mainstream systems significantly reduces migration risk. Combined with multi-tenant resource sharing, 70–90 per cent data compression and higher computing efficiency, it is expected to save customers at least 30 per cent in total cost of ownership. In some bank migration cases, batch processing time has been cut to around one-third of its previous level. These figures are not just from discounts, but the result of improved resource efficiency and simplified operations and maintenance.

Core system modernisation in finance is a continuous journey. Dai’s conclusion carries weight: a viable upgrade path must deliver proven capabilities and real-world experience, offer measurable economic value, meet current needs for growth, efficiency and resilience, and lay a foundation of “data and context readiness” for the AI era. For Hong Kong, which aims to consolidate its position as a global fintech hub, this behind-the-scenes infrastructure question will help determine how steadily and how far it can go in the next phase.