JPMorgan’s Alex Yao says China’s “hundred-model” AI war now bets on enterprise value

China’s AI market used to feel like a competition for who could ship the most impressive models. Now, JPMorgan’s Alex Yao says the fight is changing shape. In his view, the drive for AI monetisation among Chinese tech companies is turning into a battle over “delivering measurable business value” instead of raw technical performance. That difference sounds semantic, but in markets it is everything. Benchmarks are easier to market. Business value is harder to prove. And the companies that cannot show the second end up paying for the first with real capital.

Yao, head of China equity research at JPMorgan, also points to another shift underway. China’s AI landscape is rapidly consolidating, moving from a fragmented “hundred-model” battle toward a smaller group of globally competitive leading players. The underlying logic is straightforward: not every model wins, but the winners are those that can turn consumer-facing features into reliable enterprise-grade infrastructure. In other words, the monetisation path is moving upstream, from demos and consumer engagement to workflows, integrations, and spending that shows up in revenue rather than hype.

If you have spent any time around software markets, you know why this matters. Early AI investment often optimizes for capability, because capability is what attracts talent, users, and attention. But as the market matures, buyers start demanding proof. Enterprises, especially, do not buy “smart.” They buy outcomes that reduce cost, increase throughput, or lower risk. When Yao frames the shift as measurable business value, he is basically describing the moment where CFOs and procurement teams become the real gatekeepers. That is a different buyer with different timelines and a different tolerance for uncertainty.

This is also where the “hundred-model” era runs into physics. Many approaches can look strong in isolation. But only a few will survive contact with real product requirements such as uptime, governance, latency, integration with existing systems, and the ability to support multiple customer use cases at scale. Consolidation is what you get when investors and customers refuse to fund endless experiments. JPMorgan’s framing implies that the market is not just narrowing because developers are quitting. It is narrowing because the commercial bar is rising, and companies that cannot clear it are getting pushed out.

There is another reason this pivot from technical performance to business value is consequential. It changes how companies allocate capital and how boards measure progress. In a “model war,” leadership can hide behind training runs and research milestones. In an “enterprise value” war, the internal scorecard shifts to deployment metrics, retention, customer expansion, and the ability to generate revenue from infrastructure rather than only usage. That tends to favor teams that are already building distribution channels, partnerships, and enterprise-ready tooling, not only teams that are publishing papers.

Regulation and policy environment also tends to reinforce this kind of monetisation filtering, even when the exact rules differ across sectors. Compliance-heavy industries and data-governance requirements make it harder to monetize quickly with purely consumer-first approaches. That pushes vendors to tailor solutions for organizational buyers who need stronger controls and clearer accountability. The enterprise-grade framing in Yao’s comments aligns with that reality: enterprise buyers want systems they can trust inside their own operational constraints.

Second-order effects follow quickly for peers. If a smaller set of leading players increasingly looks “globally competitive” and can transform consumer-facing features into enterprise-grade infrastructure, then fundraising and partnership strategies will likely tilt accordingly. Competitors will face pressure to either differentiate through vertical integrations and reliability, or accept that their cost of capital rises as the market discounts them for lacking measurable enterprise traction. For investors and operators, the question becomes less “which model is best” and more “which product reliably sells, deploys, and expands.” That is a very different evaluation lens, one that rewards execution and distribution.

For executives across Chinese tech, the takeaway is not subtle. Yao’s JPMorgan perspective suggests the center of gravity is moving. The winners are likely to be those that can convert impressive AI capabilities into dependable enterprise infrastructure and, crucially, into results that can be tied to business value. If you are building or investing, this is the moment to ask whether your roadmap is winning on capability alone, or whether it also has a credible monetisation engine that can survive scrutiny. The market is consolidating. The business value bar is rising. And as the “hundred-model” battle fades, the companies that can prove impact will be the ones still standing after the hype cycle moves on.