OpenEvidence is by no means the endgame for domestically produced medical AI in China.

The competition surrounding medical AI presents starkly contrasting scenarios on both sides of the ocean.

In the past six months, American tech companies have continuously broken records in terms of financing scale and market valuation for healthcare AI. The popular Abridge completed a $300 million financing round, reaching a valuation of $5.3 billion; Tala Health followed closely, achieving a $1.2 billion valuation with a $100 million super seed round; Open Evidence reached a valuation of $12 billion after receiving $250 million from Thrive Capital and DST Global, equivalent to nearly 100 billion RMB.

Similarly focused on the exploration of AI within hospitals, China's medical IT companies appear somewhat overshadowed.

Winning Health's 2026 Early Preview of 2025 Performance Shows a Shift from Profit to Loss, with Non-GAAP Net Profit Attributable to Shareholders Expected to Lose RMB 304 Million to RMB 374 Million. Market Value Continues to Decline, Failing to Meet Expectations Supported by AI.

B-Soft's ownership change at the end of the year fails to reverse its decline, with projected losses of 305 million to 405 million yuan in 2025, and its market value unlikely to return to a ten-digit figure.

Similar technical paths, but completely different progress, have led many startups to begin "localizing" the American model, seeking the possibility of rapid monetization in medical AI.

But it needs to be clear: Is AI healthcare in the U.S. overestimated? Will the improved OE be capable of meeting the needs of doctors in China and help companies achieve a comeback in the limited future?

Looking solely at the business level, the models of Abridge, Tala Health, and Open Evidence are not unique, as there are numerous medical IT companies in China with similar scenarios and converging technologies.

Abridge Focuses on AI Clinical Documentation Automation Platform, Which Can Transcribe Doctor-Patient Conversations in Real Time, Accurately Extract Key Information Such as Chief Complaints, Medical History, Medication, and Diagnosis, and Automatically Generate Structured Medical Records Compliant with SOAP Standards. In This Process, Abridge Addresses the Industry Pain Points of Heavy and Time-Consuming Medical Record Keeping by Doctors, and the Generated Medical Record Process Is Traceable for Verification, Which Can Reduce Hallucinations to a Certain Extent.

In China, many companies are currently comparable to Abridge, such as Fuxin Innovation, WholeMed, iFlytek Healthcare, and Unisound, which possess technologies for converting speech to text and automatically generating structured medical records. Moreover, companies like United Imaging Intelligence and Deepwise Healthcare can automatically extract key information from medical imaging reports to generate structured imaging reports. Overall, their capabilities are slightly superior to those in the U.S. market.

Tala Health Focuses on Full-Process Healthcare Navigation, with Its Key Product Being the AI-Powered Multi-Agent Collaboration Service Platform, Bridging the "Last Mile" Gap in Medical Services. Unlike AI products that solely focus on the consultation stage, this platform covers the entire chain from initial patient consultation to post-recovery follow-ups. After patients describe their symptoms through a conversational interface, the platform’s AI conducts systematic inquiries, generates standardized medical records, and coordinates with licensed medical staff to complete tasks such as review, triage, examination scheduling, medical insurance pre-authorization, and referral, while continuously tracking the patient's recovery progress.

In China, domestic companies such as WeDoctor, Ping An Good Doctor, and Ark Health also focus on full-link medical coordination and chronic disease closed-loop management. However, due to the insufficient smoothness of some hospital systems, there may be breakpoints in the entire chain, requiring patients to manually input data. Therefore, the actual experience is slightly weaker compared to the U.S. market.

OpenEvidence: An AI Engine Focused on Clinical Evidence-Based Decision-Making, Specializing in Empowering Authoritative Medical Knowledge. It primarily addresses the challenges of doctors lacking authoritative references for diagnosis and treatment decisions and low efficiency in literature review. Unlike traditional decision-making systems, it integrates tens of thousands of top-tier global medical journals, authoritative treatment guidelines, and drug instructions, relying on retrieval-augmented generation technology to provide licensed physicians with precise and traceable clinical evidence support. Due to its accuracy and authority, it has become the most frequently used clinical AI tool by U.S. practicing physicians.

In terms of technical approaches alone, there are a dozen projects in China that can rival OpenEvidence. However, in terms of valuation, the combined total of these projects might just approach that of OpenEvidence.

The problem lies inDoctor's HabitsAndPayerFu Xin Chuangke CEO Wu Di believes: For American doctors, all clinical decisions must be based on authoritative and traceable medical evidence. Open Evidence fully adheres to these standards, only including content from top-tier peer-reviewed medical journals, official clinical guidelines, information disclosed by pharmaceutical regulatory agencies, and authoritative medical databases.

When doctors perform diagnostic activities, Open Evidence can not only provide corresponding decision support to improve the accuracy and efficiency of diagnosis but also leave real-time traces, becoming one of the pieces of evidence for doctors to follow the evidence-based concept during diagnosis and providing a basis for potential disputes.

By contrast, although there are also a large number of clinical guidelines and medical literature in China, the standards from different sources have not been fully unified. Doctors need to rely on their experience to assist in decision-making due to the lack of authoritative standards to support their decisions. Therefore, for doctors in China, AI engines for clinical evidence-based decision-making can only be regarded as an efficiency-enhancing tool.

Even in many hospitals with outdated computer hardware and software, forcibly implementing AI engines will only exacerbate system lag, becoming an additional burden.

The payer also has similar differences. Although at this stage, OpenEvidence's AI serves clinical doctors and medical students with formal physician qualifications, its core paying clients are pharmaceutical and device companies (a small part of the functions also charge hospitals and doctors). It delivers targeted advertisements based on doctor inquiry scenarios, such as pushing compliant hypoglycemic drug information when a doctor searches for diabetes treatment plans, or matching relevant targeted drug data when querying cancer therapies.

However, domestic regulatory authorities strictly prohibit pharmaceutical and medical device companies from directly placing advertisements with doctors, which has completely locked down OpenEvidence's unique business model. For now, domestic healthcare IT companies can only target hospitals and the government as payers for AI services, and their valuation cap can at most match the current figures of Abridge and Tala Health.

Nowadays, the primary auxiliary decision-making capabilities of the medical system have been basically established, and the willingness of payers to pay has further declined. In the absence of new flagship products being launched, the valuation range of China's medical IT may further decrease in the short term.

From a macro perspective, the technical levels and market sizes of China and the U.S. are similar, with no differences in magnitude. However, the absence of pharmaceutical and device payment entities in China's medical AI sector creates a gap, making rapid monetization unlikely.

Such defects are serious, but not fatal.

Returning to OpenEvidence's Series D financing, the reason investors are willing to offer it a sky-high valuation is partly because it can quickly monetize through its existing business model, and partly because it also has potential in B2B sectors beyond pharmaceuticals and medical devices. Simple pharmaceutical and medical device marketing is not enough to build a competitive barrier for a company valued at hundreds of billions.

"The model of OpenEvidence has its leading edge, but from an engineer's perspective, its agent capabilities do not stand out significantly compared to other agent products. The entire system can be quickly replicated by other IT enterprises," said Wu Di.

"For example, if Google and Oracle were determined to do this, they would certainly have the ability to integrate these authoritative journals into their own systems and provide services to doctors in an embedded rather than add-on manner. Moreover, if Google were to sell pharmaceutical and medical device advertisements, it would most likely perform better than a startup."

In other words, if relying solely on pharmaceutical and device marketing, OpenEvidence's model is easy to replicate, and its ceiling is also clearly visible.

In the face of such a situation, OpenEvidence's practical approach is to use advertising as a stepping stone to monetize the value of AI in a more direct way.

Last October, OpenEvidence and Veeva Systems established a long-term partnership to jointly develop Open Vista, which connects doctors and patients with relevant clinical trials, increasing patient access to clinical trials and thereby accelerating the development of related drugs, directly serving SaaS enterprises.

In February this year, OpenEvidence reached a cooperation agreement with Sutter Health to integrate the platform into the Epic electronic health record system, allowing doctors to perform natural language evidence searches without leaving the medical record input environment, returning to the B-end "pay-for-performance" commercial logic.

Looking back at China's medical IT enterprises, they are in a period of rapid iteration of AI technology. Every year, new AI technologies emerge that disrupt the capabilities of previous models, thereby covering scenarios that were previously difficult to reach.

Therefore, although China's medical IT is absent in seeking the direction of pharmaceutical and device payment, it is still possible to bypass this step.Seize the Opportunity of the Next Wave of AI Value Explosion Directly。

The opportunities here are related to the future development process of the medical AI era.

In the 1.0 era of medical AI, the core was the single-point "pay-per-effectiveness" model. Abridge, Fuxin Sci-Tech Innovation, and FullClinic Medicine's structured electronic medical records are typical examples of this era, where companies identified specific pain points in certain scenarios to optimize efficiency. Structured electronic medical records significantly improved doctors' daily work efficiency, with measurable benefits that made hospitals willing to pay for AI.

The 2.0 era of medical AI is a systematic "pay-for-performance" model. The AI applications in the 1.0 era were single-point and intermittent, improving efficiency at certain stages of a doctor's workflow but failing to optimize the entire process. Such improvements approach their limits once pain points are addressed one by one. Only by reconstructing the system architecture can new efficiency-enhancing opportunities be unlocked.

Currently, leading medical IT companies in China have similar plans. For instance, Winning Health proposed the "EA+AI Intelligent Architecture" several years ago, which runs GPU-based computing architecture in parallel with CPU architecture to optimize system processes and explore the potential of “pay-for-performance”. However, since single-point “pay-for-performance” is not yet well-developed and many hospitals do not have a strong demand for intelligent architecture, systematic “pay-for-performance” relies on the widespread emergence of single-point “pay-for-performance”.

As for the medical AI 3.0 era, it means adding AI decision-making on the basis of 2.0, utilizing smart management under highly coordinated human-computer interaction to achieve the optimal allocation of all medical resources in the hospital.

Recently, OpenClaw, etc.The emergence of autonomous agents has materialized many of the aforementioned imaginings, making it hopeful that the above three eras will accelerate into reality.。

For example, in the past, a certain director proposed an optimization requirement for a process, which might require a medical IT company to spend 2 months on demand research and solution communication, followed by code development, debugging, and application. Even after completing the optimization, it might still not meet the director's requirements.

By leveraging autonomous agents developed by medical IT companies, hospitals can independently request AI to meet their needs, allowing AI to complete debugging quickly. The various costs saved in the process provide room for medical IT companies to adopt a single-point "pay-per-performance" model.

Revisiting Systematic "Pay-for-Performance": The key here lies in transforming "Q&A AI" into "AI responsible for end-to-end task execution."

For instance, doctors used to have to read patient information one by one during ward rounds, but AI can optimize the entire process by reviewing patient records beforehand, automatically generating briefings from various monitoring systems, allowing doctors to grasp key points before making decisions. This shifts the value of ward rounds from identifying problems to solving them.

In general,The scenes remain the same, but the upgrade of AI capabilities enables it to truly help hospitals gain benefits, deeply integrate into doctors' workflows, and thereby create possibilities for many medical IT enterprises in China to adopt a "pay-per-performance" model., enabling medical IT companies to seek payment directly from hospital systems rather than pharmaceutical and device enterprises.

Therefore, although China's medical IT enterprises are temporarily lagging behind due to environmental limitations, such a lag will not last forever. The current business model found by OpenEvidence is not the ultimate solution for medical IT. With the rapid iteration of AI technology, we may witness the next technological explosion within 1-2 years and see medical IT reshape its landscape.