China’s MaaS Market: From 16× Growth to 18.6 Billion RMB — What’s Next?

1. The Growth Trajectory: 16× in One Year

The numbers behind China’s MaaS boom are difficult to grasp without seeing them side by side. According to IDC’s May 2026 tracking report, publicly accessible large model API calls in China jumped from 114 trillion tokens in 2024 to 1,944 trillion tokens in 2025 — a year-over-year increase of roughly 16×.

IDC expects 2026 to deliver another ~20× surge, pushing total consumption to approximately 40,000 trillion tokens. On the revenue side, the public cloud MaaS market reached 3.07 billion RMB in 2025 and is forecast to hit ~18.6 billion RMB in 2026.

Three forces are driving this curve: multi-modal model maturity, the large-scale deployment of Agent-class applications, and the expansion from text generation into multi-modal understanding and automated execution. Each interaction now consumes an order of magnitude more tokens than simple chat. If you are evaluating a Chinese AI partner today, the single most important thing to understand is that this market is not growing — it is exploding, and the competitive dynamics of even 12 months ago no longer apply.

2. Market Structure & Extreme Concentration

For foreign procurement teams, the most striking finding from IDC’s 2026 report is not the growth rate — it is the concentration. Three vendors control nearly 90% of all public cloud MaaS token calls in China.

Source: IDC China Enterprise MaaS Market Tracking, 2025. Shares by external token consumption, excluding vendors’ own internal usage.

Alibaba Cloud holds a solid second place at 28%, leveraging its massive enterprise customer base and open-source Qwen ecosystem. Baidu AI Cloud sits at 10% in the public cloud race but leads in private deployment. Tencent Cloud’s cautionary tale: despite enormous consumer reach, Tencent Cloud has dropped out of the top three in public cloud MaaS call volume — a stark reminder that having a consumer AI app does not automatically translate into enterprise MaaS dominance.

For foreign firms, this concentration creates both opportunity and risk. On one hand, working with the top three gives you access to the most mature, battle-tested infrastructure. On the other hand, it creates vendor lock-in exposure that must be actively managed through multi-cloud strategies and contractual safeguards. Before entering any partnership, we recommend a thorough professional enterprise credit report on your prospective Chinese AI vendor to verify their actual market position, financial stability, and partnership history.

3. Public Cloud vs. Private Deployment: Two Different Markets

A critical insight that many foreign observers miss is that China’s MaaS market operates as two parallel universes with entirely different vendor hierarchies, pricing models, and competitive dynamics.

In the private deployment segment, the leaderboard changes completely. Baidu AI Cloud ranks among the top tier, SenseTime holds the #2 position with 11.3% share, and China Telecom AI (the subsidiary of China Telecom) has entered the top 3 in IDC’s 2025H2 ranking of private deployment vendors. This matters enormously for foreign firms operating in regulated industries: if your Chinese partner or JV requires on-premise AI infrastructure for compliance reasons, you will be dealing with a completely different set of vendors than those dominating the public cloud headlines.

4. What This Means for Foreign Buyers

If you are a procurement manager, compliance officer, or legal counsel at a multinational considering Chinese AI services, the market structure above translates into four concrete implications:

4.1 Vendor Concentration Risk Is Real

With nearly 90% of public cloud MaaS calls flowing through just three vendors, your negotiating leverage depends heavily on which tier you engage. Working directly with Volcano Engine, Alibaba, or Baidu gives you scale but limited price flexibility. Working with a smaller reseller or SI (system integrator) adds a layer of margin but potentially better service. Either way, you are exposed to the strategic direction of these three giants.

4.2 Dependency Mapping Is Essential

When your Chinese joint venture or supplier tells you they “use AI extensively,” the critical follow-up questions are: Which MaaS provider? What is their monthly token consumption? Do they have multi-cloud redundancy? A partner who is 100% dependent on a single vendor carries hidden continuity risk — if that vendor changes pricing, terms, or experiences disruption, your partner’s entire AI operation could be compromised.

4.3 You Are in a “Volume-over-Profit” Phase

IDC data shows token consumption is growing 16–20× while revenue grows only ~5×. This means unit prices are collapsing. For buyers, this is excellent news in the short term — you are buying tokens at historically low prices. But it also means your vendor may be operating at thin or negative margins, raising questions about long-term viability. A partner whose business model depends on unsustainable token pricing may not be there in two years.

4.4 Negotiation Leverage Favors Buyers Right Now

Because vendors are fighting for call volume, foreign buyers — especially those bringing meaningful scale — have unusual negotiating power. Multi-year committed use discounts, custom SLAs, and dedicated support are all on the table. But this window will close as the market matures and vendors shift from land-grab to profitability.

5. The Revenue Puzzle: Volume ≠ Profit

The most misunderstood aspect of China’s MaaS market is the gap between consumption growth and revenue growth. Token volume is growing 16–20× annually, but MaaS revenue is growing only about 5×. This means unit prices are falling at a rate of roughly 60–70% per year.

The cautionary tale here is Zhipu AI (智谱): after raising API prices by a cumulative ~83% across three rounds, their OpenRouter global ranking collapsed from #3 to #17. The market voted with its feet. Conversely, DeepSeek’s strategy of cutting prices by 75% during off-peak hours locked in developer mindshare and sustained their position as the #1 Chinese model by call volume.

For foreign firms, the lesson is clear: a Chinese AI vendor’s pricing power is not determined by their brand or technology alone, but by their engineering efficiency in cache architecture, chip adaptation, and inference engine optimization. A vendor that simply resells tokens purchased upstream is extremely vulnerable. A vendor that has invested in the full stack — from chips to inference engines — has structural cost advantages that translate into sustainable partnership potential.

6. Due Diligence Checklist for MaaS Partners

Based on the market structure outlined above, here is a practical due diligence framework for any foreign firm evaluating a Chinese AI service provider, technology partner, or portfolio company:

• Verify true market position: Which IDC tier does the vendor occupy? Is their claimed market share based on token calls, revenue, or a self-reported metric? A professional enterprise credit report reveals the real financial picture behind marketing claims.
• Map vendor dependency: Which upstream MaaS provider(s) does your partner rely on? Is there multi-cloud redundancy? What happens if their primary vendor changes terms?
• Assess chip adaptability: With domestic AI chips reaching 41% market share in 2025, does the vendor have proven adaptation capabilities across Huawei Ascend, Cambricon, and other domestic accelerators? Single-chip dependency is a red flag.
• Examine utilization rates: Industry-wide GPU utilization is below 30%. A partner with demonstrably higher utilization has superior engineering efficiency and lower cost structure.
• Check compliance posture: Does the vendor meet the 80% green power requirement for national hub nodes? Are they compliant with generative AI filing requirements? Non-compliance risks operational disruption.
• Review financial sustainability: In a market where revenue grows 5× but volume grows 16×, is your vendor burning cash unsustainably? Request audited financials and verify through independent official enterprise credit reports.
• Evaluate overseas capability: If you need cross-border AI services, does the vendor have compliant data export mechanisms and overseas entity support? Inefficient data sovereignty arrangements create legal exposure for foreign partners.