China’s cloud market has finally stepped away from the era of price wars. On March 18, Alibaba Cloud and Baidu Intelligent Cloud both announced price hikes.

Alibaba Cloud’s official website revealed that prices for computing power card products, such as the T-Head Zhenwu 810E, would climb by 5% to 34%, while file storage CPFS (Smart Computing Edition) would surge by 30%. Baidu Intelligent Cloud also issued a pricing adjustment notice, stating that prices for AI computing power-related products would rise by approximately 5% to 30%, with parallel file storage seeing an increase of about 30%.

The transition from price wars to widespread price hikes signifies the end of an era of price competition in China’s cloud computing industry, with the sector now officially entering a monetization phase focused on the value of computing power.

I. Farewell to Price Wars: Cloud Computing Steers Toward a Pro-Cyclical Phase

Rewinding the clock by a year, the Chinese cloud computing market presented a starkly different landscape.

In 2024, cloud providers were still embroiled in price wars. In February of that year, Alibaba Cloud announced at its 2024 strategic launch event that it would slash official prices for cloud products across the board in China, with an average reduction exceeding 20% and a maximum reduction of 55%, earning the moniker of the ‘largest price cut in history.’

Subsequently, major players like Tencent Cloud and Baidu Intelligent Cloud almost all jumped into the fray, propelling the entire industry into a competitive phase of ‘trading price for volume.’

During the initial phase of large-scale model adoption, cloud providers generally regarded computing power and model capabilities as entry-level resources, utilizing extremely low prices to attract developers and enterprises, prioritizing the occupation of ecological niches.

At the time, the industry consensus was that computing power would become cheaper and might even transform into a continuously depreciating foundational resource, akin to bandwidth.

However, this assessment was swiftly revised.

This March, Alibaba Cloud and Baidu Intelligent Cloud nearly simultaneously announced price increases for AI computing power and storage products, with some products seeing hikes of more than 30%. Such synchronized price hikes by leading cloud providers are rare in recent years.

From a ‘maximum price cut of 50%’ to a ‘maximum price hike of 30%,’ China’s cloud computing market underwent a pricing logic reversal within a year. Behind this shift lie changes in supply-demand dynamics.

On one hand, demand continues to soar.

Over the past year, AI applications have evolved from conversational tools to intelligent agents capable of performing tasks. Tools like OpenClaw exemplify this transformation, where AI no longer merely generates text but can directly invoke software and complete complex tasks across applications. Consequently, computing power consumption has skyrocketed from thousands or tens of thousands of tokens per query-answer interaction to hundreds of thousands or even millions.

A report released by Frost & Sullivan in February 2026 disclosed that the average daily invocation volume of enterprise-grade large models in China surged to 37.0 trillion tokens in the second half of 2025, up 263% from 10.2 trillion tokens in the first half, achieving nearly threefold expansion in just six months.

On the other hand, the expansion of computing power supply has not been as rapid as anticipated. High-end GPUs, servers, and data center construction all entail heavy asset investments with long cycles and high costs.

Storage costs have risen sharply. According to a report by Counterpoint, prices for memory and NAND flash storage both increased by 90% in the first quarter of 2026, with the upward trend likely to persist, directly driving up cloud providers’ cost structures for data storage and transmission.

Meanwhile, the supply of high-end GPUs remains constrained, with delivery cycles for core computing chips like NVIDIA’s H100 and H200 already extending to 2027. Computing power expansion is no longer a problem that can be resolved in the short term.

Given the rapid release of demand and the difficulty of synchronizing supply growth in the short term, computing power remains a ‘relatively scarce resource,’ making price hikes inevitable.

Moreover, computing power differs from traditional cloud resources in that its core costs are highly reliant on investments in high-end chips and infrastructure. During the period of continuous price cuts, cloud providers were essentially trading profits for scale. Once AI demand transitions from user acquisition to large-scale usage, such a low-price-driven strategy becomes difficult to sustain long-term.

Thus, this price hike represents an industry-wide ‘course correction.’ When leading providers simultaneously raise prices, it signals the arrival of a new phase: cloud computing is no longer just about competing for users through low prices but is now rebuilding competitive barriers around computing efficiency, model capabilities, and solution capabilities.

II. Post-Price Hike: Cloud Computing Enters the ‘Computing Power Monetization’ Phase

This round of price increases is not confined to the Chinese market; the global cloud infrastructure market is also experiencing a noticeable price rebound.

In January 2026, Amazon AWS fired the ‘first shot’ in AI computing power price hikes, raising prices for its EC2 Machine Learning Capacity Blocks (Capacity Blocks for ML) by approximately 15%. A few days later, Google Cloud announced that it would comprehensively raise global data transfer service prices starting May 1, marking its first such price increase for foundational cloud services in nearly two decades.

The synchronized moves by global tech giants indicate that the cloud computing industry has officially transitioned from ‘burning money to capture market share’ to a ‘computing power monetization’ phase.

Over the past decade, the main theme of China’s cloud computing market has been price wars. Providers like Alibaba Cloud and Baidu Cloud used sustained price cuts to educate the market, expand user bases, and squeeze competitors. However, the strategy of capturing market share through low prices is no longer sustainable, and cloud providers now need reasonable profits to support underlying technology R&D and long-term service stability.

This industry shift will quickly transmit downstream to customers, who will face direct pressure from rising costs.

Over the past year, many AI startups have grown accustomed to low-price subsidies from cloud providers. However, with computing power prices on the rise, ‘shell’ applications that lack self-sustaining capabilities and rely solely on token consumption to prop up valuations will face significant setbacks.

The barriers to entry for AI entrepreneurship, particularly for foundational model R&D, will be further raised. Entrepreneurs must now consider whether the value generated by their applications can outpace the rise in computing power costs. This will force the market to abandon the illusion of ‘burning money for scale’ and shift toward pursuing high-value, high-stickiness scenarios.

For cloud providers, this ushers in a higher-dimensional competition.

In-house R&D capabilities have become the deepest moat for major players. Cloud providers with self-developed chips and full-stack optimization capabilities can still employ flexible cost-hedging strategies and resource scheduling efficiencies even after raising prices in line with the market.

In contrast, second- and third-tier cloud providers lacking hardware autonomy face a dilemma. Raising prices may drive core customers to migrate to leading platforms, while not raising prices would result in massive losses amid soaring chip and storage component costs. This cost inversion will further squeeze their survival space to the limit.