文 | 锦缎
过去28个月，市场对“医疗AI诊断”热门股完成一轮完整的压力测试：Tempus自历史高点回撤约35%；估值大幅回撤引发我们再审视“医疗AI诊断”的现实商业卡点。
本文基于2020年初至2026年9月国内公开披露的212张AI相关医疗器械注册证，以及政策原文，沿“产品化—注册证—价格立项”三道商业化闸门，逐层解析AI诊断的核心商业化卡点。
产品化：监管要求AI产品设计逻辑必须依据医学指南且具体到垂直适应症领域。DeepSeek模式的通识大模型产品难被监管认可，背后是对医学严肃性和清晰责任边界的敬畏。
注册证：独立软件，或嵌入医疗器械的软件组件（注册主体仍是医疗器械）为监管唯二认可的AI产品注册形态，AI诊断商业化只能卖软件或卖设备硬件。DeepSeek式以API接口为主流的商业模式还拿不到医疗行业的入场券。
价格立项：没有独立的医疗服务产出做支撑，监管不会为AI诊断设立独立的价格，价格体系中AI诊断的最高待遇仅是“不额外加价”的扩展项。AI诊断为企业打造第二创收曲线的通道被关闭。
即便抛开前述政策限制，市场“体感”中医院仍较难给AI诊断单独付费，因为AI诊断不是提供科室“从0到1”的新服务，而是设备厂商既有软件能力的智能化升级：院端视其为谈判筹码，厂商视其为竞争壁垒，AI诊断的真实角色是厂商守住市场份额和价格水平的“防守资产”，而非创收工具。
医疗AI诊断的长期破局，或许仍需等待行业更多积累后的DeepSeek时刻。不过短期或可调整思路，把“不显著的提效”做成“可量化的生产力”：
要么把效率换算成收入，用以显著缩短报告周转时间、提升检查通量等，让AI的贡献直接进入科室的收入报表；要么去医生稀缺的场景做“生产力本身”，用以基层、体检、大规模筛查，在那里AI不是辅助，而是唯一可负担的生产力；要么跳出医保支付方，到体检机构、药企、商保的预算里，寻找为数据和服务付费的客户。
01 背景：AI诊断热门股的价值回归
过去28个月，资本市场用真金白银，给热门股Tempus的“医疗AI诊断”商业叙事做了一轮完整的压力测试。
2024年6月，Tempus以37美元上市，头半年即大起大落。8月盘中翻倍至77美元，11月逼近80美元，12月单月又暴跌46%。
真正的引爆点是2025年1月：美国“国会山股神”佩洛西家族1月14日买入看涨期权被披露、AI健康助手Olivia同期发布，1月21日单日暴涨35.5%，创上市以来最大单日涨幅；2月14日盘中摸到91.45美元，随后急速回吐至56美元。
此后靠阿斯利康与Pathos的2亿美元肿瘤基础模型协议、连续超预期的季度业绩，在2025年10月冲上历史高点104.32美元，市值约180亿美元。后虽披露EBITDA季度转正以及段永平建仓等利好信息，股价仍持续下跌，直到9月逐渐披露PRISM2论文登上Nature Medicine、与Moderna及默克达成合作后，股价才回涨至68美元附近，仍较历史高点跌约35%。
以下则为英伟达收盘价走势参考：
02 医疗AI应用现状：浮华之下的结构显著分化
从公开资料梳理来看，医疗AI的商业化应用方向主要涉及以下七类：
1.医学影像AI：影像辅助诊断；
2.诊断AI：检测辅助诊断；
3.AI制药：靶点挖掘等；
4.医疗大模型：病历质控、循证问答等；
5.手术机器人AI：手术定位导航等；
6.办公AI：医生/医院办公协助等；
7.患者慢病管理：C端体检解读、居家慢病管理等。
以上仅是案例分类视角下的应用方向；倘若换到获证视角，则会看到很不一样的结果。
我们统计了2020年初至2026年9月公开披露的产品名称或产品内容含AI（或人工智能）的医疗器械注册证，共计212张（其中三类证174张）——诊断相关的证占到91%，177张应用于影像诊断AI，独占84%，全是软件产品；其次为体外诊断AI应用，14张，占比7%。
03 医疗AI诊断的三道商业化闸门
前一章我们看到，从案例分类看，医疗AI有七条赛道，热闹而分散；但换到“获证”视角，212张注册证中有177张落在影像诊断，独占84%。这个84%并非偶然的技术偏好，而是医疗行业政策引导的结果。
要理解它，首先要理解医疗产品的市场推广逻辑链条。
医疗赛道是一个强政策干预的行业，任何医疗产品的商业化路径都逃不脱制度早已框定的标准化路径——产品化（企业与监管）→产品证（药监局）→价格立项（医保局）（价格立项后不一定同步进入医保清单，进入医保以扩大商业化规模的环节本文不作展开，亦不影响本文对行业底层逻辑的解析）。
医疗AI的价值必须以标准化医疗产品为载体，穿过三道闸口，才能在医疗行业实现规模变现：
产品化是所有商业推广的前提，未完成标准化包装的医疗服务，无法进入后续产品注册与价格立项流程（ICL不在本文讨论范畴）；
产品只有取得药监局批准的医疗器械注册证，企业才能合规向医院销售；
产品只有纳入医保局的价格立项指南，医院才能合规向患者收费并合理采购企业产品。
一、产品化：遵循垂直领域医学指南逻辑的辅助性产品设计
产品化方面，药监局始终坚持AI诊断的辅助定位与“医学指南逻辑支撑”的产品设计原则。
2026年6月，国家药监局医疗器械技术审评中心在《人工智能辅助诊断医疗器械临床评价注册审查指导原则（征求意见稿）》中再次明确：“（人工智能）软件仅可辅助医师进行诊断……医师对临床诊断结果负责”、“产品设计必须依据医学指南”。
医学的严肃性，不允许一个责任边界模糊、功能内涵不清的东西在医疗体系里成为产品——AI能力必须被封装为能力边界清晰、责任风险可控、可重复交付的标准化产品，而不是一项服务、一个模型接口或一段提示词。
二、注册证：独立软件或嵌入医疗器械的唯二形态
注册证方面，药监局始终限定唯二的产品注册形态——独立软件，或嵌入医疗器械的软件组件（注册主体仍是医疗器械）。
早在2022年3月，国家药监局《人工智能医疗器械注册审查指导原则》即明确AI医疗器械的申报适用范围——“包括第二类、第三类人工智能独立软件和含有人工智能软件组件的医疗器械（包括体外诊断医疗器械）”。AI产品的获证形态由此被锁定为两种：独立软件申报，或嵌入医疗器械中以医疗器械主体进行申报。
直到2025年药监局发布《关于发布优化全生命周期监管支持高端医疗器械创新发展有关举措的公告》，监管口径有所松动——“开展医用大模型、人工智能医疗器械、医学影像前处理、流程优化软件等新功能、新技术、新模态产品的管理属性和类别研究，依据技术发展成熟度及时动态调整产品管理类别”。
但态度松动不等于原则松动——按2026年9月最新发布的《人工智能医疗器械注册审查指导原则（2026年修订版）（征求意见稿）》，AI产品的注册形态依旧锁定在独立软件与嵌入组件两种。
三、价格立项：无独立收费资格的扩展项地位
2024年11月，AI首次被写入价格立项指南，但文件明确不赋予其独立收费、额外加价的地位。
当时，国家医保局同时印发《放射检查类》《超声检查类》《康复类》三份医疗服务价格项目立项指南（试行），首次在立项指南中写入AI内容——三份指南均统一在主项目下补充“人工智能辅助诊断”扩展项，并明确“扩展项指同一项目下以不同方式提供或在不同场景应用时，只扩展价格项目适用范围、不额外加价的一类子项，子项的价格按主项目执行”。
同期，国家医保局发布官方解读文章，给出理由：“在尚无独立的医疗服务产出、辅助诊断质量效用难以确定的情况下，在已收取诊断相应项目检查费用后，不宜单就人工智能辅助诊断再向患者额外收费。”
2025年12月，国家医保局发布的《病理类医疗服务价格项目立项指南（试行）》中，“人工智能辅助诊断”同样被列为病理诊断的扩展项。
值得注意的是，2026年8月发布的体外诊断收官价格指南——《检验类医疗服务价格项目立项指南（试行）》中，“人工智能辅助诊断”在征求意见稿阶段曾被列入扩展项，正式稿中却被删除，转而归入主项目的价格构成，失去了单列于“扩展项”的地位——即从价格项目设计看，AI辅助诊断甚至没能在更大范围的体外诊断项目中守住“扩展项地位”，而被纳入主项目的基础能力构成。
至此回看影像AI诊断在获证上的“独树一帜”，便能从政策逻辑得到解释：
其一，AI辅助诊断的“扩展项”地位，最先在影像类检查中获得价格立项指南的确认；其二，177张影像AI注册证全部为软件产品——因为政策目前只认可两种注册形态：独立软件，或嵌入医疗器械的软件组件（后者的注册主体仍是医疗器械，而非AI）。
04 医疗AI诊断的商业化卡点：政策的严肃性与市场化的实效性
梳理医疗AI诊断商业化的政策链条后不难看出，政策端的核心限制在于产品设计形态与收费资格；但即便抛开政策影响，医疗AI诊断在院端市场化路径上同样存在自身卡点。
一、政策维度：产品形态和收费能力的设限
产品化闸口，药监局锁定AI诊断的辅助定位与垂直产品设计原则。药监局要求企业必须依据医学指南设计垂直领域的AI产品，甚至细分垂直到具体适应症，DeepSeek这类通识型AI能力无法直接迁移应用——监管需要的是能力边界清晰、责任风险可控、可重复交付的标准化产品，而不是一项服务、一个模型接口或一段提示词。
价格立项闸口，医保局按“独立的医疗服务产出”定价，而AI诊断恰恰没有独立产出，因而不具备独立收费资格；医疗企业只能将AI嵌入医疗器械中以设备硬件的价格条码来试图“增收”。
从更宏观的视角来理解政策深意，政策其实并没有“打压”医疗AI诊断——它只是严格遵循“以人民健康为中心、以临床价值为导向、控制人民群众医药费用负担”的总原则（2021年八部门《深化医疗服务价格改革试点方案》），稳步推进医疗AI的技术进步。尤其2023年医疗反腐以后，医保基金收入不再保持过去的高速增长；存量时代下，医保资源的配置必须更为审慎。
二、市场维度：AI诊断不是新付费对象，是企业竞争力的基本项
特别说明：以下内容仅为作者对真实医院市场的接触体感，受限于接触范围，难免存在明显偏误与较强的主观性，望各位注意甄别。
如果说政策维度回答的是“收不了钱”，市场维度回答的则是——就算能收，也很少有医院愿意为AI诊断多付钱。
先看推广方式。医疗AI诊断当下的商业化路径有两条：独立软件销售与嵌入医疗器械销售，后者占绝对主导——前者不具备收费通道（第三章已述），绝大多数只能搭设备的“车”进院。证是软件的证，钱是设备的钱。
再看医院如何看待AI诊断。事实上，AI出现以前，影像和检验设备本身就配套分析软件——照出的片子需要软件重建与处理，样本检测结果需要软件分析基础生物信息。医院采购设备，从来不是只买硬件，而是买软硬件一体的整体方案。AI出现以后，厂商把长期积淀的软件分析能力做了智能化升级，并在此基础上申报AI软件的三类证。从供给角度看，这无疑是厂商的服务能力增量；但从需求方医院的角度看，这不是一次从0到1的改变——过去买设备就包含分析软件，如今厂商把分析软件升级为AI，医院更多将其视为厂商服务能力的增强，而非新增了一项应该单独付费的服务。
结合市场竞争环境，这种院端态度还会被进一步强化。对成熟医疗器械而言，行业一旦迈入激烈竞争态势，降价便不可避免，甚至出现恶性降价——送设备、换试剂消耗的打法屡见不鲜。在这样的市场里，医院天然倾向于把厂商设备的AI和自动化增值能力，理解为厂商应对激烈竞争、保住市场份额与价格而做的服务增值。买方把它当作谈判筹码（不配AI，我可以把市场给到配AI的厂商），卖方把它当作竞争壁垒（我追随行业节奏做了AI增值，我可以再谈判维持设备原价并守住原来份额）——双方都无法把它当作一个独立的付费商品。
所以，从医院端的实际体感看，即便没有政策干预的限制，现阶段的市场化谈判中也很难把当前的AI诊断能力推入创收阶段；它只能成为厂商竞争力增强的重要工具，用以保持现有设备和试剂的价格水平、避免过度降价。
05 破局的不成熟思考
医疗AI诊断的长期破局或许仍需等待医疗行业更多积累后的DeepSeek时刻，不过短期亦有一些可参考的思路：把“不显著的提效”做成“可量化的生产力”。
要么把效率换算成收入，显著缩短报告周转时间、提升检查通量，让AI的贡献直接进入科室的收入报表；要么去医生稀缺的场景做“生产力本身”，基层、体检、大规模筛查，在那里AI不是辅助，而是“规模化医疗关口前移”可负担的生产力；要么跳出医保支付方，到体检机构、药企、商保的预算里，寻找为数据和服务付费的客户。
只有当AI的价值能以医院看得懂的报表语言呈现，付费意愿才会从“锦上添花的犹豫”变成“主动买单的坚定”。
这也是本文开篇那句话的另一面：回归业务本身，成为生产力本身，政策和市场都会重新给AI定价。
本文由 AI宝库自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
♪ Man, ♪
Over the past 28 months, the market has completed a full set of stress tests for the medical AI diagnostics unit: Tempus retreated about 35 per cent from historical heights; and a significant valuation retreat triggered a review of the real business card points of the medical AI diagnostics.
This document is based on 212 publicly disclosed AI-related medical equipment registrations from the beginning of 2020 to September 2026, as well as the original text of the policy, and analyzes the core commercialized points of the AID diagnosis by level, along the three commercial gates of the Productization-Registration-Pricing Line.
Productization: Regulation requires that the design logic of AI products must be based on medical guidance and be specific to vertical adaptation. The LLM product, known as the DeepSeek model, is difficult to regulate and is backed by awe of medical seriousness and clear boundaries of responsibility.
Registration certificate: Independent software, or software components embedded in medical devices (the subject of which is still a medical device), regulates the register of second-approval AI products, which commercializes only software or equipment hardware. The DeepSeek business model, which mainstreams the API interface, does not yet have tickets to the medical sector.
Price lines: Without the support of an independent output of health services, regulation will not establish an independent price for an AI diagnosis, and the best treatment for an AI diagnosis in a price system is an extension of “no extra price”. The AI diagnostic route to create a second income-generating curve for businesses was closed.
Even outside the above-mentioned policy constraints, it is still more difficult for the market “body sense” hospital to give AI a separate diagnosis, paid, because it does not provide a new “0 to 1” service for the section, but rather an intelligent upgrading of the software capabilities of the equipment manufacturer: The court considered it a bargaining chip and the manufacturer considered it a competitive barrier, and the real role of AI's diagnosis was that the manufacturer kept the market share and price level “defensible assets” rather than an income-generating tool.
The long-term breakdown of the medical AI diagnosis may still have to wait for DeepSeek after more industry accumulation. However, in the short term, it may be possible to adjust the approach to “unvisible efficiency gains” to “productivity of quantization”:
Either efficiency convert is incomed to significantly reduce reporting lead times, increase inspection fluxes, etc., and allow AI's contribution to go directly to the income statement of the section; or go to the scarce scene of a doctor to do “productivity itself” for grass-roots, medical examination, mass screening, where AI is not auxiliary, but only affordable; or jump out of the health insurance payout to the budget of a medical examination body, a pharmaceutical company, a commercial insurance, to look for clients for data and services paid.
Background: The return of value to the AID diagnostic hot unit
Over the past 28 months, the capital market has conducted a full set of stress tests for the commercial narrative “Medical AI Diagnosis” in Tempus, a popular unit.
In June 2024, Tempus was listed at $37, the first half of which was a big drop. It doubled to $77 in August, approached $80 in November and fell by 46 per cent in December alone.
The real point of detonation was January 2025: on January 14, the Pelosi family of the United States of America “Congressional Monuments” bought to see an increase in their options was disclosed; during the same period of publish, an AI health assistant, Olivia, went up by 35.5 per cent on 21 January, the largest single-day increase since he came to the market; and on February 14, they touched $91.45 and then threw back to $56.
This was followed by a $200 million tumour base model agreement between Asliccan and Pathos, a continuous, over-anticipated quarterly performance, which in October 2025 hit an all-time high of $104.32 with a market value of about $18 billion. Following the disclosure of good information on the EBITA Quarterly Reform and the Yongping War, stock prices continued to decline, and only after the gradual disclosure of the PRISM2 paper on Nature Medicine in September, in cooperation with Moderna and Merk, did the stock price rise to about US$ 68, or about 35 per cent higher than in history.
The following is a reference to the trend of Ying Weidar closing prices:
02 Current status of medical AI application: a significant fragmentation of structures under the floating
In terms of public information, the commercialization directions of medical AI are mainly in the following seven categories:
1. Medical video AI: video-assisted diagnosis;
2. Diagnosis AI: test-aided diagnosis;
3. AAI Pharmaceuticals: Target excavation, etc.;
Medical LLM: medical records, evidence-based questions and answers;
5. Surgeon robot AI: Surgeon positioning navigation, etc.;
6. Office AI: Medical/hospital assistance, etc.;
Patient management: C-end medical interpretation, home-based management of chronic diseases, etc.
These are only the application directions in the case classification perspective; if they are changed to the certificated perspective, a very different outcome is seen.
A total of 212 (of which three types of certificates 174) - The number of diagnostic-related certificates was 91 per cent, 177 for video-diagnosis AI and 84 per cent for software products, followed by in vitro diagnostic AI applications, 14 or 7 per cent.
03 Three commercial gates for medical AI diagnosis
As we have seen in the previous chapter, from the case classification, medical AI has seven popular and dispersed tracks; in the “certification” perspective, however, 177 of the 212 registration certificates fell on video diagnostics, accounting for 84 per cent. This 84 per cent is not an occasional technological preference, but rather a result of policy guidance in the health sector.
It is understood, first and foremost, by the logical chain of market promotion of medical products.
The medical track is an industry of strong policy intervention, and any path to the commercialization of medical products will not escape the standardized path that the system has set for itself - – Productization (business and regulation) product certification (MAE) price setting (MSA) (prices are not necessarily added to the health insurance list at the same time as they are set, and the chain of access to health insurance to increase the scale of commercialization is not developed, nor does it affect the analysis of the industry's bottom logic).
The value of medical AI must be carried by standardized medical products, through three gates, in order to realize a scale in the medical industry:
Productization is a prerequisite for all commercial promotion, and medical services that do not complete standardized packaging do not have access to subsequent product registration and price setting processes (ICL is not covered by this paper);
Only by obtaining a certificate of registration of medical equipment approved by the Office of Drug Control and Drug Control can an enterprise comply with the requirements for selling to hospitals;
Only if the product is included in the price-setting guidelines of the Health Insurance Board can the hospital comply with the patient ' s fees and reasonably procure the enterprise ' s products.
I. Productization: support product design that follows the logic of the medical guidelines in the vertical field
In terms of productization, the Agency has always adhered to the principles of product design, which support the orientation of AID diagnostics and are “backed by the logic of medical guidance”.
In June 2026, the Technical Review Centre for Medical Devices of the National Drug Authority again clarified in the Guidelines for the Registration Review of Clinical Evaluation of Artificial Intelligence Assisted Diagnosis Medical Devices (AIS) (Advisory Drafts) that: “Software (AIS) can only assist a physician in his or her diagnosis... The physician is responsible for the clinical diagnosis results and the product must be designed in accordance with a medical guide”.
The seriousness of medicine does not allow something that is vague and insufficiently functional to become a product in the health system. - AI capacity must be sealed as a standardized product with clear capability boundaries, controllable liability risk and re-deliveryable, rather than a service, a model interface or a prompt.
II. REGISTRATION : INDEPENDENT SOFTWARE OR A SECOND-SEQUARTERS PACKAGE OF INSTRUMENTS
With regard to registration certificates, the Office of Drug Control and Drug Control always limits the form of product registration to only two - independent software or software components embedded in medical devices (the subject of registration remains a medical device).
As early as March 2022, the National Drug Authority's Guidelines for the Registration Review of Artificial Intelligence Medical Devices, i.e., clarify the scope of application for the declaration of the AI medical device - “including Category II, Category III AII.S. software and medical devices (including in vitro diagnostic medical devices) containing components of artificial intelligence software”. The certificated form of the AI product is thus locked in two categories: an independent software declaration, or a declaration of the subject of the medical device embedded in the medical device.
Until 2025, the UNDCP publish Bulletin on Measures to Optimize the publish Life Cycle Regulation to Support the Innovative Development of High-end Medical Devices, the regulatory calibration was relaxed - “to carry out new functions such as medical LLM, artificially intelligent medical devices, medical pre-image processing, process optimization software, management attributes and types of new technologies, new modelled products, and to adapt product management categories to technological maturity in a timely and dynamic manner”.
However, a loose attitude is not equivalent to a relaxed principle - the registration of AI products remains locked in independent software and embedded components in accordance with the Guidelines for the Registration Review of Artificially Intelligent Medical Devices (2026 Revision) (Referral Draft) of September 2026 publish.
Price setting: Extension status without eligibility for independent fees
In November 2024, for the first time, AI was included in the price-setting guide, but the document clearly did not give it the status of a stand-alone fee or additional mark-up.
At that time, the National Health Insurance Agency simultaneously published three (pilot) guidelines for the development of health-care price projects for the radiology category and the ultrasound category and the rehabilitation category, and for the first time included AI in the development guide - – The three guides harmonize the addition of the “Advisoryly Assisted Diagnosis” extension under the main item and clarify that “extension items are only those sub-items that extend the scope of application of the price item and are not subject to an additional increase, when provided in different ways or applied in different scenarios under the same item, at the price of the main item”.
During the same period, the official publish article of the National Health Insurance Institute gave reasons for the following: “In the absence of independent medical service output and uncertainty as to the quality of the diagnostic assistance, it is not appropriate to charge the patient an additional fee for the artificial intelligence aided diagnosis alone, when the cost of the examination of the corresponding project has been charged.”
In December 2025, publish of the National Institute of Health and Social Security, entitled “Advanced Intelligence Aided Diagnosis” was also included as an extension of pathological diagnosis in the Guidelines for the Setting of Prices for Pathological Services Project (Pilot).
It is worth noting that in August 2026, publish's In vitro Diagnosis Receiver's Price Guide — “Guide to the Establishment of Project Prices for Inspection Medical Services (Preliminary)” — “Advisory Intellectually Assisted Diagnostics” were included in the extension at the consultation stage, but the official draft was deleted and moved to the main project's price component, losing its status as a single “extension” - • That is, in terms of price project design, AI-assisted diagnostics do not even hold the “extension status” in the larger in vitro diagnostic project, but are part of the base capacity structure of the main project.
So, looking back at the "one-size-fits-all" of video AI's diagnosis, it can be explained by policy logic:
First, the “extension” status of AAI-assisted diagnostics, first confirmed by the price-setting guide in video type checks, and second, the 177 ICA registrations are all software products - because the policy currently recognizes only two forms of registration: independent software, or software components embedded in medical devices (the latter are still medical devices, not AI).
04 Commercialization of medical AI diagnostics: seriousness of policy and effectiveness of market-based
It is not difficult to see from the policy chain that streamlines the commercialization of medical AI diagnostics that the central limitation at the policy level lies in product design patterns and fee eligibility; but even without the policy implications, medical AI diagnostics have their own card points on the institutional market path.
Policy dimensions: product form and fee-capability limits
Product-chemical gates, where the Agency locks down the AIS-assisted positioning and vertical product design principles. The DEA requires that companies design vertical AI products based on medical guidance, even subdivided vertically into specific adaptive conditions, and that general AI capabilities such as DeepSeek cannot be directly migrated - – Regulation requires standardized products with clear capacity boundaries, manageable liability risks and repermissible delivery, rather than a service, a model interface or a prompt.
The price gate is priced by the Health Insurance Board on the basis of “independent output of medical services”, while the AI diagnostics do not have independent output and therefore qualify for independent fees; the medical enterprise can only attempt to “receipt” by embedding AI in a medical device with a price bar code for equipment hardware.
To understand the policy from a more macro-level perspective, it does not actually “press” the medical AI diagnosis - it is simply a step forward in the technological advancement of the medical AI, in strict compliance with the general principle of “the people's health-centred, clinical value-oriented, and the control of the people's medical costs” (the eight sectoral Pilot Programme for the Further Price Reform of Health Services of 2021). In particular, after the fight against corruption in health care in 2023, the revenues of the health insurance fund no longer kept up with the high growth of the past; in the age of stock, the allocation of health-care resources must be made more prudent.
Market dimension: AID is not a new paid subject and is an essential element of enterprise competitiveness
In particular, the following are the author ' s perceptions of exposure to the real hospital market, which are limited to the extent of exposure, and the apparent bias and greater subjectivity that may be expected of you.
If the policy dimension answers “no money,” the market dimension answers — and if they do, very few hospitals are willing to pay more for the AI diagnosis.
Let's see how it goes. There are two ways of commercializing the current medical AI diagnosis: the sale of independent software and the sale of embedded medical devices, which are absolutely dominant — the former do not have a fee route (as described in chapter III), and the vast majority of them are “cars” in which equipment is used. The certificate is for software and the money for equipment.
Let's see how the hospital looks at the AID. In fact, before AI came into existence, the image and test equipment itself was accompanied by the analysis software — the film produced required the software to be reconstructed and processed, and the sample test results required the software to analyse basic biological information. Hospitals purchase equipment and never buy only hardware, but a whole package of hardware and software. Following the emergence of AI, the manufacturer upgraded its long-term accumulated software analysis capability to an intelligent basis on which to declare three types of AI software. This is undoubtedly an increase in the capacity of the manufacturer from the supply point of view; but from the point of view of a demand-side hospital, this is not a change from 0 to 1 — when the equipment was purchased to include analytical software, when the manufacturer now upgrades the analytical software to AI, the hospital sees it more as an increase in the capacity of the manufacturer to provide services, rather than adding a service that should be paid alone.
This institutional attitude will be further strengthened in the context of a competitive market environment. In the case of mature medical equipment, once the industry has entered a competitive situation, price reductions are inevitable, and even a vicious price reduction — the use of equipment and reagent consumption is common. In such a market, hospitals naturally tend to interpret the AI and automated value-added capabilities of manufacturers ' equipment as adding value to the services they provide in response to intense competition, preserving market shares and prices. The buyer used it as a bargaining chip (not worthy of AI, I can give the market to the supplier of AI) and the seller used it as a competitive barrier (I added value to AI following the industry rhythm, I can then negotiate to maintain the original price of the equipment and hold the original share) - neither of the parties could treat it as an independent paid commodity.
It is therefore difficult to move the current AI diagnostic capacity into the income-generating phase of the current market negotiations, even without the constraints of policy intervention, as seen in the actual perception at the hospital level; it can only be an important tool for increasing the competitiveness of producers to maintain the price levels of existing equipment and reagents and avoid excessive price reductions.
Immaturity thinking for a bad time.
The long-term breakdown of the medical AI diagnosis may still have to wait for DeepSeek to become more accumulated in the medical industry, but there are some short-term ideas that can be taken into account: “Improveability of quantization”.
Either the efficiency of convert would be earned, the reporting lead time would be significantly reduced, the inspection flux would be upgraded, and the contribution of AI would go directly to the income statements of the section; or the "productivity itself" would be done in the scarce scenes of doctors, at the grass-roots level, for medical examinations, for large-scale screenings, where AI would not be auxiliary, but would be “scaling the health-care portal” for affordable productivity; or the health-care payout would be removed from the budget of the medical-care institution, the pharmaceutical company, the commercial insurance, to look for clients of data and services paid.
Only when the value of AI can be expressed in the language of a hospital-readable statement will the will of paid be changed from a “suspension to a “suspension to a proactive purchase”.
This is another aspect of the opening sentence: going back to business itself, becoming productivity itself, and policies and markets repricing AI.
Auto-fetched and translated by AI Gem Box. All rights belong to the original author.