2025 Healthcare AI Industry Report: Exploring Value Measurement and Payment to Overcome the Challenges of Medical AI

The widespread adoption of AI, from deep learning to large language models, has already transformed the revenue structures of many leading companies, making it a key driver of innovation and productivity. However, the medical field seems largely disconnected from this shift. Despite multiple waves of technological change and countless enterprises working tirelessly, none have achieved large-scale profitability in healthcare.

Does the healthcare sector genuinely require AI? Are the limitations found in data, computing power, or algorithms? What can we expect in both the short and long term? Where does the key to commercial breakthroughs lie?

In this report, we have identified ten medical specialties where AI has been deployed successfully, each with its unique implementation approach. Through in-depth interviews with over 30 doctors and 40 enterprises, we analyze AI's application models, commercialization progress, and future development trends from the perspectives of hospitals, physicians, and patients. Our goal is to find clarity amid uncertainty and establish a solid foundation for moving forward.

What Is the Value of Medical Artificial Intelligence?

The market size for medical AI solutions in China reached 16.4 billion in 2024. As a key technology reshaping production methods across industries, medical AI has maintained high growth momentum even as other sectors experienced downturns due to economic cycles. According to Frost & Sullivan's estimates, the core market for medical AI solutions is projected to reach $ 35.3 billion by 2030, with a compound annual growth rate (CAGR) of 13.63%. Over the next five years, factors influencing the market size of medical AI will include its scope of application, hospitals' willingness to adopt AI solutions, the costs associated with AI evaluation and approval, challenges in accessing medical data, and the competitive landscape of products in the industry pipeline.

Small-scale profitability for medical AI may be achievable within the next five years, driven by physician involvement and policy support. In the early stages of medical AI development, challenges such as the lack of large-scale, standardized medical data, low reuse rates of processed data, a shortage of interdisciplinary talent bridging medicine and engineering, and hospitals' reluctance to pay separately for AI solutions led to persistently high training costs, limited generalizability of certain intelligent solutions, and constrained application value, causing many AI companies to take detours. However, with hospitals’ growing understanding of AI, increased proactive participation of physicians in AI research and development, multiple policies promoting the adoption of medical AI, and the efficiency improvements in data governance brought by large models, medical AI companies are expected to control costs and achieve small-scale profitability within the next five years.

Reduced difficulty in accessing clinical data and cost control drives breakthroughs in both application scenarios and product effectiveness. Since the establishment of the National Data Administration, the process of assetizing medical and health data has accelerated, with multiple transactions of health data occurring on exchanges. If the critical challenge of data can be overcome to enable large-scale data transactions, the most significant cost component in medical AI development is expected to decrease substantially. This would facilitate a shift from quantitative to qualitative improvements in the output and effectiveness of intelligent applications.

Contradictions in the beneficiaries of medical AI raise concerns about the large-scale commercialization of this technology. For most medical devices, their value can be precisely measured in terms of health economics, such as efficacy, efficiency, and cost. However, at present, the value brought by medical AI varies across different stakeholders. For example, after the introduction of AI in a specific department, workflow optimization may allow procedures that previously required collaboration with other departments to be conducted within the department itself. While this reduces treatment time and medical expenses for patients, the department's revenue may decline due to DRG (Diagnosis-Related Group) constraints, as a single department now performs a procedure that originally involved two departments. This creates a conflict between the department's and the patients' needs for AI. In other words, even when AI capabilities meet medical requirements, the widespread existence of such value contradictions remains a primary reason why the technology has yet to achieve large-scale commercialization.