China's cloud market: saying goodbye to the 'rock-bottom prices' era

Price hikes mark an end to the race to the bottom

Alibaba Cloud (阿里云) and Baidu Intelligent Cloud (百度智能云) on March 18 announced simultaneous price increases for AI compute and file-storage products — compute-card lines up 5–34% and parallel file storage around 30% in Alibaba's case; Baidu's AI compute reportedly rose about 5–30% with similar storage adjustments. The coordinated moves signal a clear break from the deep discounting that defined China’s cloud market just a year ago. What was a market-building price war is now morphing into a phase focused on monetizing scarce compute capacity.

From "burn money" to scarce resource economics

In 2024 China’s largest cloud vendors — Alibaba Cloud (阿里云), Tencent Cloud (腾讯云) and Baidu (百度) among them — cut prices aggressively (Alibaba alone averaged more than a 20% cut, with some services down as much as 55%) to capture developers and ecosystems during the early big‑model boom. That strategy treated compute as a commoditized, depreciating input. The macro has flipped. Demand for AI has exploded: Sullivan reported that enterprise-level large‑model daily token calls in China jumped from 10.2 trillion in H1 2025 to 37.0 trillion in H2 2025 — a 263% rise — as AI moves from chat to action‑oriented agents that consume orders of magnitude more tokens per task.

Supply, costs and geopolitics are tightening the market

Supply has not kept pace. High‑end GPUs and data‑center capacity are heavy‑capex and long‑cycle items; it has been reported that delivery windows for NVIDIA H100/H200 chips are pushed out into 2027. Memory and NAND prices are also rising sharply — Counterpoint reported roughly 90% increases in Q1 2026 — directly raising storage and I/O costs for cloud providers. US export controls and broader semiconductor trade frictions have been cited as factors that complicate the global high‑end GPU supply picture, adding a geopolitical dimension to what otherwise looks like pure market economics.

Winners, losers and what customers should expect

This recalibration is global: AWS and Google Cloud have already announced price increases for AI and data‑transfer services, so China’s move is part of a wider industry reset from subsidized growth to "compute monetization." For startups and token‑hungry applications the bill will rise — those without clear revenue models or high‑value use cases face real risk. For cloud vendors, self‑developed chips and full‑stack optimization become stronger moats; second‑ and third‑tier providers lacking hardware autonomy may be forced to choose between raising prices and losing key customers, or absorbing losses until capacity and price pressures ease. Who benefits? Those who can turn scarce compute into paid, sticky value.