NextFin News — China’s market for artificial intelligence in healthcare has surpassed 1.5 trillion yuan, according to Analysys, and has been expanding at an annual rate well above the broader medical sector. AI-assisted diagnosis systems are now present in a majority of tertiary hospitals and a substantial share of secondary ones. Regulators have cleared more than 120 Class III AI medical devices, many of them focused on medical imaging. Capital has followed: specialist firms have raised new rounds, and large technology groups have expanded health and hospital projects.

This is not the first wave of enthusiasm. Several imaging-focused startups that pursued listings around 2021 struggled to convert regulatory clearances into durable, scaled revenue. One listed. Others remained private and continued to operate under tighter conditions. The earlier episode left a clear lesson: technical approval and pilot deployments do not automatically produce a business.

What has changed is the capability of the underlying models and the breadth of attempted applications. Large multimodal systems can handle text, images and structured records in ways earlier narrow algorithms could not. Policy signals have also become more explicit, including the creation of pricing categories that contemplate AI-assisted diagnosis. The practical question is no longer whether AI can perform discrete tasks inside a hospital. It is which tasks generate reliable payment, from whom, and at what margin.

Where the technology is already at work

The most mature use is assisted imaging. Algorithms trained on large volumes of labeled scans flag regions of interest or produce quantitative measurements that doctors review. The Chinese market for AI medical imaging was reported above 15 billion yuan in 2025 and is projected higher for 2026. Dozens of Class III products cover lungs, cardiovascular structures, cerebrovascular findings and other specialties.

Clinicians who use these tools often describe the value in narrower terms than marketing materials. Quantitative volume measurement, three-dimensional reconstruction and consistency on repetitive tasks matter more than autonomous diagnosis. Experienced physicians still read primary images themselves; the software reduces variability and saves time on metrics that are tedious to calculate by eye. Products that over-promise autonomous judgment tend to be treated cautiously.

Drug discovery is the highest-visibility research application. AI shortens early stages of target identification and molecule design. Global tallies count many AI-influenced candidates in clinical trials, yet no fully AI-designed drug has completed the path to approval solely on algorithmic origin. Human physiology and trial requirements remain binding constraints. Time can be saved; safety validation cannot be skipped.

Surgical robotics and intraoperative guidance attract attention but face high capital cost, training burden and liability. Remote surgery adds network-latency risk that most institutions are unwilling to accept for critical procedures. Consumer-facing health assistants—symptom checkers, report interpreters, chronic-disease nudges—reach the largest audiences and the weakest willingness to pay. Engagement is high; conversion to sustained subscription or premium service remains low.

Three types of competitors

Large technology groups bring capital, cloud infrastructure and consumer distribution. Some have invested in hospitals or built consumer health apps with tens of millions of monthly users. Their strength is scale and model capability; their weakness is depth of clinical workflow and the long regulatory and procurement cycles of the medical system.

Traditional device manufacturers start from installed equipment, sales channels and existing registrations. Adding AI features to scanners or monitors creates an upgrade path that pure software vendors lack. Their constraint is speed of software iteration and the need to integrate intelligence without disrupting hardware reliability and service contracts.

AI-native firms focus on algorithms, data and specialized models. Several hold multiple Class III certificates in imaging niches. A smaller number have reached or approached profitability. Many still depend on “bolt-on” integrations with hospital information systems—an approach that becomes harder as institutions tighten interfaces and prefer native functionality inside core platforms.

Practitioners across these groups increasingly agree on one point: single-point tools are becoming harder to sell. Hospitals and payers prefer capabilities that fit into broader clinical or operational systems rather than standalone applications that require separate logins, separate validation and separate budget lines.

The payment problem

Four potential revenue sources exist: public insurance reimbursement, hospital capital or operational budgets, pharmaceutical partnerships, and direct consumer payment. Only the second and third are currently reliable for most vendors, and even those are under pressure.

Public insurance inclusion of AI-assisted services would create the most stable demand, yet pricing and coverage rules are still evolving. Hospital budgets face cost-control and volume-based procurement disciplines that limit what software can command. Pharmaceutical companies pay for tools that demonstrably compress discovery timelines; that market is real but belongs mainly to computational-chemistry and biology specialists. Consumers download health apps readily and pay for them reluctantly.

As a result, the firms with the clearest near-term economics tend to be those that attach AI to hardware already being purchased, or those that sell measurable research productivity to drug makers. Pure software plays that once expected rapid hospital-wide adoption are adjusting expectations downward or seeking adjacent revenue streams—insurance, population health, or platform fees—that can absorb cost and share infrastructure.

The next phase of the market will be decided less by model benchmarks than by procurement and reimbursement design. Until payment pathways are clearer and more predictable, capital will continue to flow toward the companies that can either ride existing device budgets or demonstrate concrete savings and revenue to a defined buyer. Everyone else will remain in the familiar position of proving clinical utility while still searching for a scalable customer who is both willing and able to pay.