Original · Unique Research · 2026-03-23 · Shanghai
Editor's note: This full English edition preserves the original March 2026 interview and its reporting voice. Company financial, regulatory and clinical-performance claims are statements reported in that source, not independently verified findings. The source does not provide registration numbers, approved indications or a testing protocol supporting the Med-PaLM comparison; selection for an industry project program is not itself a medical-device registration. Statements about improved diagnosis, reduced risk, market demand and 2026 as the industry's beginning remain attributed claims or opinions, not medical guidance or demonstrated safety. The original twice describes 合规付费客户 as five Chinese characters; the phrase has six. That wording is retained below rather than silently corrected.
Unique Research · Guest Interview
Six years without a penny from institutional investors,
yet this company says, "We're together because we're happy."
2026 is the true beginning of AI in healthcare.
Company interviewed: ProcenAI
Interviewee: Wang Pu (Co-founder)
Founded: 2019 · Xi'an
Founded in 2019, ProcenAI has not taken any institutional investment to date.
In AI healthcare, other companies have expanded aggressively with money from dollar-denominated funds, burning through one funding round after another. What about this company? It quietly reached break-even.
Stranger still is its slogan. Not "change the world," not "AI-powered healthcare," but—
"We're together because we're happy."
It sounds like a decorative sticker on an internet company's wall. But when Wang Pu says it, there is no boasting in his voice, only the matter-of-fact conviction that this is simply how things should be.
Six years. No institutional cash injections. No frantic expansion. No fundraising on the strength of a PPT deck.
How did they survive?
01 Xi'an, of All Places
Everyone thinks medical AI should be built in Beijing or Shanghai. Wang Pu chose Xi'an.
His reason is simple: "Combining Xi'an Jiaotong University with the Fourth Military Medical University brings together a technical foundation, clinical gold standards and a rich base of data."
Jiaotong provides foundational algorithms and technical talent. The Fourth Military Medical University provides highly standardized clinical data and diagnostic and treatment protocols.
"
From the outset, we built products on clinical gold standards, not on algorithm engineers' imagined assumptions.
Beijing and Shanghai have lively capital markets, but hospital resources are stretched, data is hard to obtain, and deployment takes a long time. Xi'an, by contrast, turned out to be an underestimated location.
While others chased the next hot opportunity, they sought the ability to set the terms: to define clinical data and to define diagnostic and treatment workflows.
02 Two Product Lines Are Really One Tree
ProcenAI is working on both AI dentistry and AI pathology.
Dentistry takes 70% of resources, with mature commercialization and a clear cash flow profile. Pathology takes 30%, with high technical barriers and substantial long-term value.
Many people ask: When a startup's resources are tight, why split the effort in two?
Wang Pu smiles: "We're not splitting our forces. We have one trunk that has grown two branches."
They built a shared AI platform: a general-purpose imaging data engine, a data governance platform and a medical knowledge graph.
The platform refined through dentistry is used directly for pathology,
making R&D much more efficient than at companies focused on a single field.
These are not two business lines. They are one tree: when the roots run deep, the branches and leaves naturally flourish.
03 Registration: Let Class II Fund Class III
What is the biggest hurdle for a medical AI company?
Obtaining registration.
Class II registration is easier, but leaves limited room for pricing. Class III registration is more valuable, yet can easily take two or three years and burn through enormous sums.
Many startups have failed at this point. They cannot wait, or they run out of money.
ProcenAI's strategy sounds a little like doing "whatever it takes":
First monetize products with Class II registration, then use the revenue to cover the costs of Class III registration.
Two Class II dental registrations have already been approved. Commercialize quickly; get into hospitals quickly.
Two Class III registrations are under application, with one selected for the Ministry of Industry and Information Technology's challenge-based project program.
Use the commercial cycle enabled by Class II registration to fund the clinical trials for Class III registration. This is making one campaign sustain the next.
04 2026 Is the Real Beginning
When it comes to the industry's bubble, Wang Pu's view is even more unexpected:
2026 is the true beginning of AI in healthcare.
Layoffs at IBM Watson Health, valuation corrections at Chinese AI healthcare companies... Outsiders read these as an "industry winter." To Wang Pu, they mark a watershed.
"We need to distinguish genuine AI healthcare companies from traditional healthcare IT companies wearing an AI disguise."
What does genuine AI healthcare require? Wang Pu names four things: academic R&D, participation in policy, integration into workflows, and deep physician involvement. Without any one of them, it is just "technology enthusiasts pleasing themselves."
Are they being greedy when others are fearful? No. They are simply looking over a longer time horizon.
05 Grassroots Healthcare Has More Pressing Needs Than Top-Tier Hospitals
ProcenAI discovered something counterintuitive:
Grassroots healthcare institutions are more receptive to AI diagnostic tools
than top-tier, Grade III Class A hospitals.
"Dentists at the grassroots level vary widely in skill. They lack specialists, equipment and experience. AI can immediately improve diagnostic capability and reduce risk; they are willing to pay for it and promote its use."
Underserved markets will provide the biggest source of growth over the next 3 years.
While others watch tenders at top-tier hospitals, they see the pressing needs in grassroots healthcare.
06 In the Age of Large Models, They Are Betting on Small Ones
Everyone is discussing large medical models such as GPT-4 and Med-PaLM. ProcenAI's choice is unexpected:
They are more optimistic about small models specialized for individual fields.
"We use a version of GPT-4 that we have optimized ourselves as the foundation. After intensive training for clinical settings, its accuracy on core diagnostic tasks has already surpassed Med-PaLM."
But that is only the foundation. What actually reaches clinical practice is a set of highly customized, domain-specific small models.
General-purpose large models address "understanding human language." Domain-specific small models address "diagnosing and treating patients well."
It is not replacement, but collaboration across layers.
07 The Final Question
At the end of the interview, I asked Wang Pu: If you could choose only one measure of success, what would it be?
Revenue? User count? Number of registrations obtained? Valuation?
He thought for a moment.
"Compliant, paying customers."
Not GMV, not DAU, not the amount raised. Customers who truly recognize the product's value, are willing to keep paying, and fully meet healthcare compliance requirements.
In the slow-moving world of medical AI, those five Chinese characters carry more weight than a thousand words.
Closing Thoughts
Wang Pu says, "We're together because we're happy."
What he does not say is that happiness starts with seeing the path clearly—and making it work.
Six years without a penny from institutional investors, yet they reached break-even. This is not idealism; it is confidence grounded in capability.
And that single measure of success—compliant, paying customers. Five Chinese characters that capture the essence of medical AI.
Selected Q&A
Q1: How have you survived six years as a startup without institutional investment?
We achieved PMF and reached break-even. In AI healthcare, when it comes to cash burn, we may be the most smart team.
Q2: Why choose Xi'an rather than Beijing or Shanghai?
Xi'an Jiaotong University and the Fourth Military Medical University gave us resources others could not obtain. Jiaotong provides algorithms and talent; the Fourth Military Medical University provides clinical data of the highest standard. From the outset, we built products on clinical gold standards, not on algorithm engineers' imagined assumptions.
Q3: Do two product lines spread your resources too thin?
We call it "one primary and one supporting line, one fast and one steady." Dentistry takes 70%, pathology 30%. But the underlying technology is fully reused: the dentistry platform is used directly for pathology. This is not splitting our forces; it is one trunk growing two branches.
Q4: What is your plan for NMPA registration?
First monetize products with Class II registration, then use the revenue to cover the costs of Class III registration. Two Class II dental registrations have been approved; two Class III registrations are under application, with one selected for the Ministry of Industry and Information Technology's challenge-based project program.
Q5: What do you think of the argument that medical AI is a bubble?
2026 is the true beginning of AI in healthcare. We need to distinguish genuine AI healthcare companies from traditional healthcare IT companies wearing an AI disguise. Genuine AI healthcare requires four things: academic R&D, participation in policy, integration into workflows, and deep physician involvement. Without any one of them, it is just "technology enthusiasts pleasing themselves."
Q6: If you could choose only one measure of success?
Compliant, paying customers.
This article is based on an in-depth interview with Wang Pu, Co-founder of ProcenAI.