Tencent Cloud has lifted the monthly price of CodeBuddy Enterprise Flagship from 78 yuan to 198 yuan, a 154% increase in one move.
This was not a new feature bundle or a reshaped plan. The price tag simply changed. The Enterprise Exclusive edition was raised at the same time, from 158 yuan to 316 yuan, a clean 100% jump. It marks Tencent Cloud's third price increase of 2026: in March it raised external token prices for the Hunyuan model family by as much as 4.63 times, and on May 9 a separate round of increases took effect for AI compute, TKE container nodes and EMR big-data products, all up 5%.
The pace: three increases in half a year
Tencent Cloud's 2026 pricing moves now form a fairly aggressive sequence:
| Time | Products affected | Increase |
|---|---|---|
| March 11 | Hunyuan model token pricing | Up to +463% |
| Effective May 9 | CVM-GPU, TKE Native nodes, THPC, EMR | +5% |
| Effective May 15 | CodeBuddy / WorkBuddy Enterprise editions | +100% to +154% |
Tencent is not alone. On March 18, Alibaba Cloud announced price adjustments for AI compute and selected storage products, ranging from 5% to 34%. The same day, Baidu AI Cloud rolled out what it called a structural pricing optimization. Overseas, AWS raised EC2 ML capacity pricing by roughly 15% in January, while Google Cloud adjusted pricing for several services on May 1.
JD Cloud is the main outlier, choosing instead to cut prices by 16% to 40%. One side is discounting to win new customers; the mainstream cloud vendors are moving in the opposite direction.
The models are not the expensive part. The shovels are.
The core explanation is simple: Nvidia supply is not enough, and demand has been ignited by AI agents.
Market analysts point to several concrete drivers behind China's 2026 compute shortage:
- Agent applications are taking off. A single agent task can move from thousands of tokens to hundreds of thousands or even millions, pushing token consumption up exponentially.
- Open-source agent frameworks such as OpenClaw are spreading. They turn people who rarely touched APIs into heavy users.
- High-end Nvidia chips remain bottlenecked. Delivery times for H100 and B200-class hardware are still unusually long.
- Inference workloads are overtaking training. Inference needs steady capacity, not just peak compute, which puts more pressure on cloud cash flow.
One industry forecast says China's annual AI token consumption could rise from the 10 x 10^15 range in 2025 to 3,900 x 10^15 by 2030, a 370-fold increase in five years. At that scale, lower cost per token and lower total cost are two entirely different things.
A useful contradiction
Tencent Cloud vice president Li Qiang has said publicly that price increases without technological innovation are difficult to sustain over the long term.
In plain terms: Tencent does not want price hikes to be the whole business model.
Dowson Tong, CEO of Tencent's cloud and smart industries group, put it in more formal language, saying the company should create value for customers through products and services rather than rely on aggressive pricing.
The message is the same: prices can rise, but the rise has to be backed by something customers can actually feel.
What it means
Open-source model companies have spent the past two years selling the story of cheap inference: DeepSeek V3 at $0.14 per million tokens, Qwen's aggressive value-for-money pitch, and a constant drumbeat that tokens are getting cheaper.
But the model layer is what is getting cheaper. The underlying compute layer is not. The part of the cloud business still pinned down by Nvidia ultimately gets passed on to customers.
The sharp increase for SaaS products such as CodeBuddy sends another signal: coding-agent workflows consume far more tokens than many buyers expected. An enterprise user can make thousands or tens of thousands of code-generation requests a month, and the included 2,000 credits will not stretch very far. Anything beyond that has to be bought at the new price.
The next thing to watch is the next pricing move from Alibaba Cloud and Baidu. If both raise prices again in Q3, the rest of 2026 will look less like a story about AI getting cheaper and more like one about compute getting more expensive, with different public-relations wording attached.
Sources: Tencent Cloud's third AI compute price increase, CodeBuddy raised 154% (BigGo Finance); AI Compute Prices Rise Across China: Tencent Joins Alibaba, Baidu in Hikes (TrendForce); CocoLoop; Alibaba, Tencent Hike Cloud Prices as AI Boom Drives Up Hardware Costs (Caixin Global)