China’s AI healthcare push hits funding problem

China’s drive to apply artificial intelligence to healthcare is facing obstacles over fragmented medical data, limited data-sharing and uncertainty over who will pay for AI-powered services, according to policy experts and industry practitioners.

China has made healthcare a priority under its nationwide “AI+” initiative, launched by the State Council in 2025. Leading hospitals have accumulated large health-data resources, while companies have developed healthcare-focused large language models and AI health applications, Foreign Policy reports.

But few AI healthcare products have secured approval and reimbursement through the public medical insurance system, limiting their commercial prospects in China.

Practitioners at AI and health-policy forums in Beijing and Shanghai this year identified usable hospital data, institutional data-sharing and the ability of AI to reduce healthcare costs as major bottlenecks.

Chinese hospitals hold large volumes of medical information, but patient records are often scattered across departments, stored in incompatible formats and recorded using different coding systems. National interoperability standards are recommendations rather than mandatory requirements, making it difficult to exchange data between institutions.

Data quality is another challenge. AI systems require longitudinal records that follow patients over time, while Chinese medical records are largely point-in-time snapshots.

Privacy requirements also discourage some hospitals from sharing information. Health data contains sensitive personal information, and administrators can face liability risks if it is leaked. The National Health Commission has encouraged approaches such as federated learning, which allows AI models to be trained across institutions without pooling raw data, but such systems require technical expertise and can increase costs.

Much existing data-sharing takes place between hospital departments and external research teams, partly because academic publications are important for the promotion of Chinese physician-scientists. Such transfers are generally governed by ethics approvals and agreements that prohibit commercial use.

The commercial case for healthcare AI remains uncertain. Companies are increasingly focusing on narrower, single-purpose applications that can be incorporated into medical equipment or devices rather than relying solely on large medical foundation models.

AI can reduce diagnostic, operational and research costs, but those savings do not necessarily translate into higher revenues for hospitals or pharmaceutical companies.

The problem is particularly significant for public hospitals, many of which are under financial pressure from debt and cost-control measures. Doctors also do not necessarily earn more when they work more efficiently, meaning efficiency gains do not automatically create additional revenue.

China’s public medical insurance system, known as 医保, is therefore seen as the key potential payer. But regional insurance funds are required to balance their income and expenditure, making a separate reimbursement pool for AI healthcare devices difficult to establish.

Experts argue that AI products will need to demonstrate that they reduce overall healthcare costs, for example by preventing missed diagnoses or unnecessary procedures. One approach is to incorporate AI into services and equipment that are already covered by the healthcare system. Imaging AI, for example, can be integrated into scanners or reporting systems and reimbursed as part of the equipment.

Efforts to create a “data flywheel” — in which better data produces better models and improved care generates more data — are also being hindered by fragmentation.

Hospitals often rely on proprietary IT systems, creating vendor lock-in and making it costly and difficult to convert historical data into compatible formats. Models can also perform substantially differently across hospitals, complicating their validation and commercialisation.

China’s vertically organised hospital system adds another layer of difficulty. Hospitals operate through separate administrative structures and may bear the risks associated with data sharing without receiving clear benefits from making their systems interoperable.

Policy proposals include making interoperability standards a requirement for hospital software certification and establishing legal protections for hospitals that share data through secure systems such as trusted data spaces.

Other approaches include using synthetic datasets generated from properly licensed medical records. China’s National Data Administration has also established a health data track under its Data Element X competition to encourage the development of such methods.

Commercial insurance could provide another source of funding. A commercial-insurance “category C” catalogue introduced in 2025 offers a potential starting point, although insurers would need shared actuarial data on health risks and costs to price AI products accurately.

Ultimately, policymakers face the broader question of where AI can deliver genuine improvements in healthcare. The challenge is not only developing the technology but aligning hospitals, pharmaceutical companies, insurers and regulators around incentives for data-sharing, investment and the delivery of better and more accessible care.

By Aghakazim Guliyev