From Burning Money to Self‑Sustaining: How Chinese AI Navigates the Value Verification Period

A different adjustment in China

The global generative AI wave has entered a value‑verification period. But while the United States is enacting a rapid purge — bankruptcies, asset fire‑sales and outright talent absorptions — China’s industry is quietly pivoting from headline‑grabbing consumer experiments to steadier, revenue‑oriented plays. Burning cash for growth? Or building foundations for lasting business models? In China, the answer increasingly looks like the latter: strategic refocusing, deep vertical bets and infrastructure plays rather than theatrical exits.

What the West’s “fast clear” looks like

In the U.S., the market has punished non‑commercial narratives quickly. Companies such as Humane and Olive AI either saw product lines shuttered or businesses wound down, while Inflection AI’s core team was absorbed by a tech giant — all signs of a liquidation‑style market discipline. At the same time, massive capital is being re‑concentrated at the top: OpenAI’s huge recent financing and deep partnerships with cloud and chip providers underline a Silicon Valley logic of concentrating resources on a few winners and moving on.

China’s “value deepening”

China’s top AI players — for example Zhipu AI (智谱AI), MiniMax (MiniMax), Kimi (月之暗面), Jieyue Xingchen (阶跃星辰), Baichuan Intelligence (百川智能) and Lingyi Wanwu (零一万物) — are pulling back from noisy consumer fronts into B‑side and government contracts, industry‑specific models and infrastructure. Zhipu’s prospectus shows the tradeoff plainly: strong revenue growth but far larger R&D and compute spending (roughly RMB 6.85bn revenue vs RMB 44bn invested through mid‑2025). Venture funding persists, but it is shifting into verticals and base layers — LiblibAI, Wuwen Xinqiong (无问芯穹) and others have raised sizable rounds — reflecting an industry aiming to convert scale into sustainable unit economics rather than headline reach.

Pressure points: route choices, giants and geopolitics

Challenges remain. Firms face sunk‑cost risks as technical routes diverge; commercialization often looks more like bespoke projects than high‑margin SaaS; and China’s tech giants — Alibaba Cloud (阿里云), Baidu Intelligent Cloud (百度智能云), Tencent and others — can simultaneously supply the infrastructure startups need and squeeze margins through aggressive price competition. Exit routes are narrower than in the U.S., pushing more firms to seek creative public‑market or industrial‑takeover paths and to pursue overseas expansion. Finally, geopolitics is complicating matters: it has been reported that Anthropic accused several Chinese models of large‑scale probing and model‑distillation attacks, prompting renewed calls for tighter AI chip export controls and sharpening the trade‑policy risks Chinese firms face as they globalize. The result is an ecosystem that is less about dramatic failures and more about hard, incremental institutional learning — and a test of whether patience can replace pivot‑fuelled hype.