Alipay, Baidu, and iFlytek Win Big—As the Government Spends $1.1 Billion on AI Healthcare, Is This a Lifeline or a Feeding Frenzy for Tech Giants?

When iFlytek, Alipay, and Baidu each secured government healthcare AI contracts worth over RMB 100 million within a single month, the entire industry seemed to be wheeled out of the ICU and into a VIP recovery suite—breathing easier, pulse stabilizing, eyes gleaming with hope. After more than two years of agony, developers of medical large language models finally exhaled: it turns out the government was willing to pay all along—it just needed the “right” approach.

But a closer look at these three blockbuster deals reveals a subtle absurdity: RMB 427.6 million to iFlytek for a “National AI Application Pilot Base,” RMB 206 million to Alipay (Ant Group) for Zhejiang’s “Medical AI Innovation Service Platform,” and RMB 169 million to Baidu for Guangzhou’s “Respiratory Infectious Disease Prediction and Early Warning System.” Not a single yuan was spent on an AI model that can directly diagnose pneumonia, nor was any contract tied to algorithmic accuracy above 95%. Instead, they all point to a grander mission: build the roads, not the cars; erect the stage, don’t perform the play.

Is this the “historic inflection point” for AI healthcare commercialization—or just a policy-driven “digital infrastructure binge”? The answer may lie in an industry inside joke: “G-side money is easy to win but hard to profit from; B-side work is tough but truly valuable; C-side dreams are beautiful but burn cash like wildfire.”

For the past two years, medical large models have been the “most tortured” sector in tech. On one hand, technology raced ahead—Baichuan launched a 32B-parameter medical model, Zhipu AI integrated with tens of thousands of hospitals, and iFlytek’s Spark Medical LLM iterated to version X1.5. On the other, the market froze: hospital budgets shrank, doctors refused to adopt, patients distrusted outputs, and investors clutched their wallets tight. While general-purpose LLMs grabbed headlines in consumer apps with chat and poetry, vertical medical AI remained trapped in a death loop: “great technology, no use cases; solid products, no payment mechanism; real demand, zero trust.”

The turning point came in November 2025, when China’s National Health Commission and four other ministries jointly issued the Opinions on Promoting and Regulating the Application of “AI + Healthcare,” explicitly stating: by 2027, a batch of high-quality clinical specialty LLMs will be developed; by 2030, AI-assisted diagnosis will be widely deployed in all secondary-and-above hospitals. This wasn’t a suggestion—it was a directive. Local governments swiftly followed: Guangdong rolled out “Yueyi Zhiying,” Jiangxi covered every village clinic, Shanghai built AI diagnostic zones—a state-led “AI New Infrastructure Campaign” had officially begun.

Enter iFlytek, Baidu, and Ant Group—the natural “general contractors” with massive computing power, data assets, and deep government ties. They’re no longer selling point solutions; they’re building entire “digital highways”: unifying data standards, breaking down hospital silos, deploying foundational models, and operating platforms. What the government wants isn’t an AI that reads X-rays—it’s an “operating system” where all AIs can run.

As one investor put it: “This mirrors Alibaba and Tencent’s early cloud strategy—pave the roads first, and only then can AI vehicles drive.” But new problems emerge: road-building is grueling, capital-intensive, and slow. Projects drag on for 2–3 years, payments arrive six months after acceptance, gross margins dip below 30%, and vendors bear unlimited liability for data security, system stability, and long-term maintenance. This isn’t an AI business—it’s informationization engineering 2.0.

The table below exposes the fundamental differences across the three healthcare AI markets:

This table makes one thing clear: G-side is a game for giants, B-side is a battlefield for specialists, and C-side is a gamble for dreamers.

For iFlytek and peers, this wave of government contracts is undeniably a lifeline. In H1 2025, iFlytek Health’s G-side revenue (from grassroots and regional solutions) already exceeded 50% of total income, with year-over-year growth peaking at 178%. These projects bring not just cash flow but “national team” credibility—leveraging government trust to unlock hospital and insurer partnerships. But for startups, the door is slamming shut. One founder admitted: “We can’t even afford the bid deposit, let alone promise three years of on-site maintenance.”

A deeper crisis looms: when AI becomes infrastructure, innovation gets standardized into oblivion. Government tenders don’t ask, “Can your model reduce misdiagnosis?” but “Does it support domestic chips?” and “Can it integrate with the provincial health data platform?” Technical merit yields to compliance and integration capability—AI is downgraded from “intelligent engine” to “digital pipeline.”

And the core commercialization bottleneck—who will pay for AI continuously?—remains unsolved. Hospitals won’t pay separately for diagnostic assistance, national insurance doesn’t reimburse AI services, and patients won’t shell out for a “maybe useful” health tip. So companies pin hopes on the future: once platforms are built, data flows, and workflows stabilize, a business model will magically appear. Yet history repeatedly shows: infrastructure enables applications—but doesn’t automatically generate profits.

So, is this RMB 8 billion splurge a turning point or an illusion?

The answer may be: it’s a structural inflection, not a full-blown boom. The government is using real money to declare: AI + healthcare isn’t optional—it’s mandatory. It buys the industry precious breathing room and creates testbeds for data accumulation, model refinement, and standard setting. Guangdong’s “Yueyi Zhiying” already connects 2,146 grassroots facilities; Jiangxi’s AI system covers over 20,000 village clinics—real-world feedback from these deployments is far more valuable than lab data.

But we must stay sober: G-side contracts are “policy backstops,” not “market validation.” They solve the “existence” problem, not the “usability” or “willingness-to-pay” questions. True commercialization still hinges on frontline clinical adoption—doctors must want to use it, hospitals must dare to assume liability, and patients must be willing to pay.

As one healthcare strategist bluntly asked: “Do you dare take responsibility for AI’s decisions? What happens when something goes wrong?” Until that question is answered, AI will remain on the periphery of care—acting as an “efficiency tool,” never a “decision-maker.”

The government’s RMB 8 billion isn’t buying AI—it’s buying time, confidence, and order. It tells the market: “We believe in this path. Keep going.” But how far you go, how you navigate, and whether you reach the finish line—that’s entirely up to you.

For giants, this is a chance to fortify moats; for startups, a wake-up call to reposition. Forget fantasies of overnight success with a breakthrough model. The winners will be the “all-rounders”—those who can chew through G-side heavy lifting, embed deeply in B-side micro-scenarios, and plant seeds of trust in the C-side.

Spring may have finally arrived for AI healthcare—but the first things to bloom might not be flowers, but grass: the quiet, resilient, soil-ready pioneers waiting for real ground to grow.