AI healthcare in China has crossed the 1.5 trillion yuan market mark, with capital flooding into a sector that regulators and hospitals are increasingly treating as a standard part of medical practice. Three categories of players - tech giants, medical device manufacturers, and AI-native startups - are competing to lead a market growing at 40-50% annually, according to data from Analysys.
The latest signal came in July 2026, when SenseTime Healthcare completed a Series B round of over $100 million at a valuation exceeding 10 billion yuan. The subsidiary of the AI vision company is sprinting toward an IPO on the strength of its "Medical World Model," with insiders saying it could become the first listed stock in that category.
Tech giants are placing large bets. ByteDance has invested 6 billion yuan to build what it calls the country's first "AI-native hospital." Ant Group's health app Afu has passed 30 million monthly active users. Tencent has released a full-stack AI healthcare offering. Baidu-backed BioMap, an AI pharmaceutical company, has reportedly filed for listing.
This is not the sector's first boom. Around 2021, four star AI medical imaging companies - Keya Medical, Infervision, Deepwise, and Airdoc - all rushed toward IPOs. Only Airdoc made it, and its stock has since fallen from HK$75 to single digits. The others are still operating, but the collective setback exposed a core problem: AI products struggled to generate large-scale revenue, and high R&D costs produced continuous losses.
What AI actually does in hospitals today
The difference this round is technological. After DeepSeek's emergence in early 2025, vertical medical large models entered a period of rapid iteration. Industry insiders say the generalization and multimodal abilities of large models have, for the first time, let AI evolve from a tool into an assistant - a shift from the previous round's imaging-only focus.
Policy support has also arrived. In 2024, the National Healthcare Security Administration added an "extended item" for AI-assisted diagnosis to its medical service pricing guidelines, and national-level signals have encouraged local experimentation with AI applications.
AI-assisted diagnosis is the most mature direction. A practitioner told Dingjiao One that "single and relatively easy-to-standardize modules such as imaging, medical record assistants, and medication review are relatively mature, but complete assisted decision-making has not been fully implemented, and no large model can independently complete the medical consultation process so far."
The principle behind AI medical imaging is straightforward: feed labeled CT scans, X-rays, and fundus photos to an algorithm, let it learn what lesions look like, and have it flag suspicious areas for doctor review. China's AI medical imaging market exceeded 150 billion yuan in 2025 and is expected to reach 235.7 billion yuan in 2026. As of June 2026, the NMPA had approved 134 Class III AI imaging diagnostic software products.
Clinicians see a more nuanced picture. Dr. Hu, a neurosurgeon at a Beijing tertiary hospital, told Dingjiao One that AI's real clinical value lies in quantitative analysis. For cerebral infarction, AI can stack tomographic images into a 3D model and calculate infarct volume accurately - a key parameter for treatment plans where visual estimation often fails. He estimates 70-90% of tertiary hospitals have such systems.
But he cautioned that some "AI imaging" products still fall short of true intelligence. "Quite a few of them are essentially advanced digital image processing technologies, which are products of the previous round of deep learning image recognition."
Drug R&D and the limits of AI
AI pharmaceutical development targets the "double ten dilemma" - new drugs take ten years and $1 billion, with high failure rates. AI is involved in target discovery, molecular design, and protein structure prediction, compressing lab trial-and-error into computer simulations.
As of mid-2026, more than 170 AI-designed or AI-optimized drug pipelines had entered clinical trials worldwide, with over ten candidates reaching Phase III. The industry calls 2026 the "year of clinical verification" for AI pharma.
No drug designed entirely by AI has been approved for market. The human body's physiological mechanisms are too complex - AI can calculate elegant formulas but cannot predict real reactions once a drug enters the body. Dr. Hu's assessment is blunt: "The medical field will not open a fast track for new technologies." He acknowledges AI may compress drug development cycles by three to five years, but "the rigid safety verification link of clinical trials cannot be replaced."
Surgical robots and health management
AI in disease treatment spans three layers: treatment plan matching, surgical path planning, and robot-assisted surgery. The first two are preoperative and have seen some implementation. The third - surgical robots like the da Vinci system - faces different obstacles.
"It's not that the technology can be implemented immediately once you have it," Dr. Hu said, citing high equipment costs, expensive consumables, and long surgeon training cycles. On telesurgery, he noted the first barrier "is not even the AI algorithm, but the network delay and signal stability - which is a fatal risk in the surgical scenario."
What he is optimistic about is the combination of medicine and engineering: "People who do engineering and material design do not understand the human body structure, and doctors know the clinical pain points but do not understand engineering mechanics and fluid mechanics. Only when the two sides are connected can the invention and innovation of medical devices be possible."
AI health management is the lightest track. Ant's Afu has over 100 million cumulative users and processes more than 10 million health consultations daily. JD Health reports 80% AI consultation penetration. Ping An Good Doctor's AI handles 4 million consultations per day. These consumer-facing services have lower technical thresholds than imaging or pharma - the challenge is willingness to pay.
Dr. Hu has observed a specific problem: some AI health consultation products have "a strange circle of overstating minor illnesses and understating serious illnesses," over-outputting suggestions for simple symptoms while failing to warn about complex conditions due to knowledge boundaries.
Three types of players, one track
Tech giants hold traffic and capital. Tencent, Ant, JD Health, ByteDance, and Alibaba Health have users, money, and technical infrastructure, but lack deep medical scenario experience. Their strategy: scale first through C-end health management, then push into hospital scenarios.
ByteDance has gone heaviest, acquiring hospitals and winning approval for a tertiary general hospital in Beijing's Chaoyang District. Ant takes a pure C-end platform approach. JD Health's 2025 revenue of 73.4 billion yuan came mostly from medical product sales, with AI as a conversion tool.
Some industry observers are skeptical of the giants' medical contributions. Ye Shiqi, a long-time AI healthcare observer, said Tencent and Ant "have contributed very little to the development of medical AI," and that JD Health and Alibaba Health are essentially traditional e-commerce businesses. Giants can scale quickly, but they struggle with the medical industry's heavy regulation, long cycles, and slow monetization.
Medical device manufacturers take a different route. United Imaging, Mindray, and Cofoe Medical embed AI as value-added modules in existing hardware. Their moats are channels, hardware, and regulatory certificates - United Imaging Intelligence alone holds 20 NMPA Class III certificates and has deployed AI in over 4,000 medical institutions.
Lv Yizhi, executive dean of Cofoe Medical's AI research institute, told Dingjiao One that device manufacturers face pressure to integrate AI with their equipment, but "device manufacturers have competitive advantages and industry barriers that pure software manufacturers do not have."
AI-native companies - Deepwise, Airdoc, Yidu Cloud, and Baichuan Intelligent - bet on models, algorithms, and data. Deepwise holds 19 Class III certificates in cardiovascular and cerebrovascular fields. Yidu Cloud achieved its first full-year profit in fiscal 2026 at 78.77 million yuan. Baichuan Intelligent shifted from general-purpose AI to medical focus in March 2025, launching its medical-enhanced model Baichuan-M4.
Why this matters for healthcare professionals
AI is already embedded in clinical workflows. The penetration rate of AI-assisted diagnosis systems exceeds 65% in tertiary hospitals and 40% in secondary hospitals. For working clinicians, the practical question is no longer whether AI will arrive - it is which tasks it will take over and which it will leave alone.
Single, standardized tasks like imaging analysis, medical record assistance, and medication review are the most mature. Complete diagnostic decision-making and autonomous care remain out of reach. The professionals who adapt fastest will be those who treat AI as a quantitative tool - for volume calculations, 3D reconstruction, and pattern matching - rather than as a replacement for clinical judgment.
For those in administrative or revenue-cycle roles, the economics matter. AI-assisted diagnosis now has a billing code, device manufacturers are selling AI as a value-add, and the market is growing at 40-50% annually. The tools that survive will be the ones that prove they can generate revenue, not just demonstrate capability. That is a standard the previous wave of AI imaging companies failed to meet, and it remains the industry's central test.
Your membership also unlocks:
