China’s artificial intelligence market has crossed from experiment to infrastructure. The country’s core AI industry passed 1.2 trillion yuan (roughly $174 billion) in 2025, spread across more than 6,200 AI companies, according to Minister of Industry and Information Technology Li Lecheng (MIIT, March 2026).[1] On the demand side, China passed 700 million generative AI users in the first half of 2026, taking penetration above 50%, up from 515 million a year earlier (CNNIC, September 2026).[2][3] By September 2026, AI-native apps alone reached 546 million monthly active users, up 82.1% year on year (QuestMobile, November 2026).[4]
Who leads it? ByteDance’s Doubao is the runaway consumer champion at 382 million MAU in May (QuestMobile, July 2026);[5] by September, Alibaba’s Qwen app was second at 158 million and DeepSeek third at about 123 million (QuestMobile, November 2026).[6] Beneath the apps sits a five-way platform fight — ByteDance, Alibaba, Tencent, Baidu and DeepSeek — plus a startup tier (Zhipu and MiniMax, now listed in Hong Kong, and Moonshot AI, which has filed to follow them) that in July 2026 pushed Chinese open models closer to the global frontier than they have ever been. This is DIA’s market deep-dive: sizing, adoption, players and money. For policy and model strategy, see our China AI strategy pillar.
How big is China’s AI market?
The headline government figure: China’s core AI industry exceeded 1.2 trillion yuan (~$174 billion) in 2025, with the country home to over 6,200 AI enterprises (MIIT, March 2026).[1] That “core industry” measure covers chips, models, platforms and applications — it is a production-side number, not enterprise spend, so treat it as the ceiling rather than the software market.
The fastest-moving slice is model-as-a-service. China’s MaaS market is on track for roughly 18.6 billion yuan in revenue and 40,000 trillion token calls in 2026, growing at a staggering 1,154.9% CAGR from 2024 to 2030 (IDC, May 2026).[7] Token prices keep falling, so usage is exploding far faster than revenue — the same dynamic that defines China’s gaming market, where scale precedes monetisation.
Regionally, China anchors an Asia-Pacific AI and generative AI market that IDC expects to reach $175 billion by 2028, compounding at 33.6% (IDC, April 2025).[8] Our Asia AI market country breakdown puts China’s weight in that total in context: no other market in the region comes close on users, models or capital deployed.
Which AI companies dominate China?
The platform tier: ByteDance, Alibaba, DeepSeek, Tencent, Baidu
ByteDance’s Doubao assistant leads China’s consumer AI market with 382 million monthly active users as of May 2026, adding 13.78 million users in June alone (QuestMobile, July 2026).[5] ByteDance is now moving to monetise, with plans to introduce Doubao subscriptions (Caixin, May 2026) — a notable shift in a market where consumer AI has been almost universally free.[9]
Alibaba’s Qwen app is the momentum story: 158 million MAU in September 2026, up an extraordinary 5,077% year on year after Alibaba relaunched its consumer assistant under the Qwen brand (QuestMobile, November 2026).[6] DeepSeek, the Hangzhou lab that detonated the market with its R1 model in January 2025, holds about 123 million MAU.[6] It closed its first outside funding round in June 2026, raising more than $7.4 billion at a valuation above $50 billion, and by October was nearing completion of a second round of about 100 billion yuan (roughly $15 billion) at a pre-money valuation of about 500 billion yuan.[10][11] Tencent’s Yuanbao, Baidu’s ERNIE Bot and iFlytek’s Spark share the long tail — a striking reversal for Baidu, which had China’s first-mover ChatGPT answer in 2023.
The startup tier just went public
China’s “AI tigers” hit liquidity in January 2026. Zhipu AI (now Z.ai) and MiniMax raised a combined $1.2 billion in Hong Kong IPOs, with MiniMax doubling on debut,[12][13] and Zhipu climbing so relentlessly that on 22 June 2026 — days after its GLM-5.2 model launched — its market capitalisation reached a record HK$1 trillion (roughly US$128 billion), with the shares closing about 2,400% above their IPO price (Caixin, June 2026).[14] The shares have since given back much of that gain, but the company used the rally to raise around $9 billion more in July and September.[15][16] Moonshot AI, maker of the Kimi assistant, raised $3.9 billion in the six months to May 2026 at a $20 billion valuation, which had reportedly climbed to $35 billion by July, with annual recurring revenue rising from roughly $200 million in April 2026 to around $1 billion by August, and in September it filed confidentially for its own Hong Kong IPO.[17][18][19] SenseTime and iFlytek, the pre-LLM incumbents, remain relevant in vision and enterprise verticals but no longer set the agenda.
July 2026 was the month Chinese open models reached the frontier
On 16 July 2026, Moonshot released Kimi K3 — a 2.8-trillion-parameter mixture-of-experts model with the full weights published on Hugging Face on 27 July under a custom Kimi K3 licence.[20][21] Independent evaluators placed it second on the Vals AI Index and third on Artificial Analysis, behind only the very newest American flagships — at a fraction of their cost (Interconnects, July 2026).[22] Analyst Nathan Lambert called it “the strongest open model ever released” and estimated the gap between open and closed frontier models has shrunk from six-to-nine months to roughly three-to-five. Demand was so intense that Moonshot paused new subscriptions within days, with daily sales up at least six-fold.[23]
Markets read it as a second DeepSeek moment. In the week of the K3 launch the Philadelphia Semiconductor Index fell 12.5% — its worst week since April 2025.[24] The pipeline behind it is relentless: Zhipu’s GLM-5.2 (June 2026) is an MIT-licensed 753-billion-parameter model[25] priced at roughly one-sixth the per-token cost of the equivalent US flagships.[26] It was followed in August by GLM-5.3, an extended post-training pass on the same base.[27] DeepSeek’s V4 began reaching general availability at the end of July, and Alibaba previewed its 2.4-trillion-parameter Qwen3.8-Max within days of K3.[28][29] Four frontier-adjacent releases in roughly six weeks — that is the cadence the rest of the market now has to price in.
How do Chinese consumers and enterprises actually use AI?
Chinese consumer adoption is the fastest large-scale technology diffusion since mobile payments. Generative AI users passed 700 million by mid-2026, and 76% of them use it to get answers to questions (CNNIC, September 2026); a year earlier, CNNIC found 74.6% of users were under 40 and more than 90% preferred domestic models.[2][3] Engagement is deepening, not just widening: AI-native app users averaged 92.7 uses and 183 minutes per month by mid-2026 (QuestMobile, July 2026).[5] Add AI plugins inside super-apps (644 million MAU) and device-maker assistants (755 million MAU), and AI touches essentially every Chinese smartphone user (QuestMobile, July 2026).[30] Foreign models remain largely absent — ChatGPT and Claude are not officially available, as we detail in our guide to AI access in China — so this is a domestically supplied market end to end.
Enterprise adoption is broad but shallower. Over 60% of leading Chinese enterprises have integrated generative AI into core business processes (IDC, May 2026), while more than 30% of manufacturers had adopted AI by end-2025 (MIIT, March 2026).[7][1] Baidu’s AI Cloud alone claims collaborations with 65% of China’s state-owned enterprises (Caixin, June 2025).[31] The monetisation gap is the caveat: usage is measured in hundreds of millions, revenue in single-digit billions.
Who is funding the boom?
Capital has come roaring back. Chinese AI startups raised more than 110 billion yuan ($16.2 billion) in Q1 2026 alone, a 185% year-on-year surge concentrated in LLMs and embodied AI (SCMP, April 2026).[32] Big-tech capex is the larger flow: Alibaba has committed at least 380 billion yuan (~$53 billion) to AI and cloud infrastructure over three years and now expects to exceed it, while Tencent’s Q1 2026 capex jumped 63% quarter on quarter to 31.9 billion yuan (Yicai, May 2026).[33][34]
Chinese open-source models are winning the download war
Open weights are China’s export strategy. Alibaba’s Qwen family had passed 700 million downloads on Hugging Face by January 2026,[35] and has over 113,000 derivative models — more than Google and Meta combined (Hugging Face, “State of Open Source”, March 2026).[36] Chinese models accounted for 41% of all Hugging Face downloads in 2025, with China overtaking the US in monthly download volume, and MIIT says China ranked first globally in open-source model downloads over the past year (MIIT, March 2026).[36][1]
The usage numbers now follow the downloads. On OpenRouter, the model-routing platform that serves as the market’s live scoreboard, Chinese-origin models carried 46.4% of routed tokens by mid-2026, against 35.7% for US models (Forkast, July 2026).[37] DeepSeek alone handles 17.6% of routed tokens and Qwen 13.9%.[37] The shift is reaching Western business software too: DeepSeek became the top trending vendor on Ramp’s corporate-spend index — meaning it is showing up in American expense reports, not just experiments — and a16z partner Martin Casado has said that when a startup pitches with an open-source model, the odds it’s a Chinese one are around 80%.[38][39] Where that export wave lands regionally is mapped in our guide to Chinese AI models in Southeast Asia.
Beijing may yet narrow that channel itself. On 16 October 2026 the commerce and science ministries proposed adding 23 AI technologies to China’s export-control catalogue, including the weights of models above certain performance thresholds, though weights that are already published openly fall outside the draft.[40]
The West can’t decide what to do about it
That usage shift is now a live policy problem, and the uncertainty is itself a market factor. After Kimi K3 launched, Axios reported the White House was weighing restrictions on Chinese models, from Entity List designations to procurement rules, while analysts noted that downloadable open weights make an outright ban nearly impossible to enforce (Axios; Tom’s Hardware, July 2026).[41] What has actually happened is narrower. On 19 October 2026 the Department of Homeland Security made public an order requiring every federal civilian agency to remove DeepSeek products, including third-party tools built on its models, and to stop buying them.[42] It doesn’t touch private or consumer use, DeepSeek still isn’t on the Entity List, and the EU has imposed no bloc-level restriction at all.
The contradiction is the story: American startups are adopting Chinese open models for cost reasons faster than Washington can decide whether to restrict them. Every option on the table — bans, entity listings, procurement rules — collides with the fact that the weights are already on a million hard drives. We unpack what Washington can and can’t do in our explainer on whether the US can ban Chinese AI models. For the full strategic contest, see our analysis of the China vs US AI race.
What is the government’s AI target?
Beijing’s “AI+” Action Plan, issued by the State Council in August 2025, targets adoption of AI agents and intelligent terminals exceeding 70% across key sectors by 2027, on a path to an “intelligent society” by 2035 (State Council, 2025).[43] A follow-on plan issued in April 2026 leans on AI to grow China’s service sector into a 100 trillion yuan (about $14 trillion) industry by 2030 (Caixin, April 2026).[44] The state is also stacking the inputs: China had filed 1.576 million AI-related patent applications by April 2025 — 38.58% of the global total — and local governments such as Shanghai are subsidising computing access with hundreds of millions of yuan.[45][46] The full policy architecture is covered in our analysis of China’s AI strategy.
Chips are the binding constraint
Compute, not talent or data, caps China’s AI market. US export controls keep Nvidia’s China-bound accelerators in regulatory limbo, and Beijing has responded by prioritising domestic silicon in state-funded data centres. Huawei planned to roughly double output of its Ascend 910C AI chips to around 600,000 units in 2026, from about 300,000 in 2025 (Bloomberg, September 2025) — still short of demand, which is why every hyperscaler reports full utilisation.[47] The market’s centre of gravity is shifting to inference, which IDC expects to account for over 70% of intelligent computing demand by 2027 (IDC, May 2026) — a workload domestic chips can serve more credibly than frontier training.[7] Expect compute scarcity to shape model design (efficiency-first, DeepSeek-style) for at least another two years.
Robots are the next battleground
Embodied AI is where China’s manufacturing base compounds its model progress. IDC forecasts China’s spending on embodied intelligence will grow from $1.4 billion to $77 billion within five years — a 94% CAGR (IDC, May 2026).[7] Humanoid output is projected to grow 94% in 2026 alone, with Unitree Robotics and AgiBot capturing nearly 80% of shipments (TrendForce, April 2026).[48] Morgan Stanley doubled its China humanoid shipment forecast to 50,000 units as commercialisation accelerated, and Chinese firms released more than 300 humanoid robot models in 2025 (MIIT, March 2026).[49][1] For a market piece, the takeaway is simple: China’s next AI revenue pool is hardware-shaped.
China AI Market 2026: FAQ
How big is China’s AI market in 2026?
China’s core AI industry surpassed 1.2 trillion yuan (~$174 billion) in 2025 across more than 6,200 companies (MIIT, March 2026).[1] The model-as-a-service segment alone is heading for roughly 18.6 billion yuan in 2026 revenue (IDC, May 2026).[7]
How many people use AI in China?
China passed 700 million generative AI users in the first half of 2026, a penetration rate above 50% (CNNIC, September 2026).[2] AI-native apps reached 546 million monthly active users by September 2026, up 82.1% year on year (QuestMobile, November 2026).[4]
What is the most popular AI assistant in China?
ByteDance’s Doubao leads, with 382 million monthly active users in May (QuestMobile, July 2026).[5] In September, Alibaba’s Qwen app followed at 158 million and DeepSeek at about 123 million (QuestMobile, November 2026).[6] Over 90% of Chinese users prefer domestic models (CNNIC, October 2025).[3]
How much funding are Chinese AI startups raising?
Chinese AI startups raised over 110 billion yuan ($16.2 billion) in Q1 2026, up 185% year on year (SCMP, April 2026).[32] Zhipu AI and MiniMax listed in Hong Kong in January 2026, and DeepSeek raised more than $7.4 billion in its first outside round in June before moving to close a second of about 100 billion yuan (roughly $15 billion) in October.[10][11]
What is China’s official AI target?
The State Council’s “AI+” Action Plan (August 2025) targets over 70% adoption of AI agents and intelligent terminals across key sectors by 2027, and a fully “intelligent society” by 2035 (State Council, 2025).[43]
Are Chinese AI models as good as American ones?
Months behind, not years — and closing. Moonshot’s Kimi K3 (July 2026) ranks behind only the newest US flagships on independent indices, and analysts estimate the open-closed frontier gap at roughly three-to-five months (Interconnects, July 2026).[22] Chinese-origin models now carry 46.4% of tokens routed through OpenRouter (Forkast, July 2026).[37]
Will the US ban Chinese AI models?
Not outright. In October 2026 the Department of Homeland Security published an order requiring federal civilian agencies to remove DeepSeek products and stop buying them, and more than a dozen states already ban it on government devices.[42][50] There is no restriction on private or consumer use, and freely downloadable open weights make any broad ban hard to enforce.
Can you use ChatGPT or Claude in China?
Not officially — Western frontier models are unavailable through normal channels, which is why the market is served almost entirely by domestic players. See our full guide to which AI models you can use in China.
Sources & Further Reading
- MIIT minister: Value of China’s AI industry hit 1.2 tln yuan in 2025 — CGTN, 2026-03-05
- Generative AI users in China exceed 700 million — CGTN (Xinhua), 2026-09-30
- Generative AI users in China reach 515 million by June 2025, report says — China Daily, 2025-10-21
- QuestMobile 2026 China Mobile Internet Autumn Report — 36Kr, 2026-11-03
- QuestMobile: China’s AI-native apps reach 499 million monthly active users — TechNode, 2026-07-14
- 红果用户规模超越豆包,千问领跑亿级 App 增速榜 — iFanr (ifeng Tech), 2026-11-04
- China Is Leading the AI Supercycle, and the Distance Is Growing — IDC, 2026-05-21
- Asia/Pacific AI spending to reach $175bn by 2028, driven by GenAI boom, says IDC — TelecomTV, 2025
- ByteDance Plans Subscriptions for AI Chatbot Doubao — Caixin Global, 2026-05-05
- DeepSeek Completes First Funding Round, Raises Over CNY 50 Billion at Valuation Above CNY 330 Billion — TrendForce, 2026-06-17
- DeepSeek doubles funding round to 100bn yuan as valuation holds steady — The Standard, 2026-10
- Startups go public in litmus test for Chinese AI — The Star, 2026-01-08
- Chinese AI unicorn MiniMax surges 109pc in Hong Kong IPO, nets US$619m — Malay Mail, 2026-01-09
- New Model Propels Zhipu AI’s Market Value to Record HK$1 Trillion — Caixin Global, 2026-06-23
- China’s Zhipu AI launches US$4 billion Hong Kong shares offering, term sheet shows — The Standard (Reuters), 2026-07
- China’s Z.AI raises US$5 billion from new share, convertible bond sales, filing shows — The Standard (Reuters), 2026-09-13
- China’s Moonshot AI raises $2B at $20B valuation as demand for open source AI skyrockets — TechCrunch, 2026-05-07
- Moonshot AI targets $2B in annualized revenue by year-end after ARR triples in two months — Crypto Briefing, 2026-09-11
- Chinese AI firm Moonshot files confidentially for Hong Kong IPO: report — DealStreetAsia (Reuters), 2026-09-03
- Moonshot releases 2.8-trillion-parameter Kimi K3 — Tom’s Hardware, 2026-07
- moonshotai/Kimi-K3 — Simon Willison’s Weblog, 2026-07-27
- Kimi K3: The open-weights escalation — Interconnects, 2026-07-20
- China’s Moonshot AI pauses Kimi subscriptions as demand surges sixfold ahead of Hong Kong IPO — Crypto Briefing, 2026-07-20
- Kimi K3 wipes $3.3T from chip stocks as Moonshot moves toward Hong Kong IPO — Tech Times, 2026-07-20
- Z.ai’s open-weights GLM-5.2 beats GPT-5.5 on multiple long-horizon coding benchmarks for 1/6th the cost — VentureBeat, 2026-06
- GLM-5.2 Just Beat GPT-5.5 at a Sixth of the Cost — Labellerr, 2026-06-22
- Z.ai launches GLM-5.3 with claimed 50% gain on coding benchmark — TechNode, 2026-08-17
- Alibaba previews Qwen3.8, claims it is second only to Claude Fable 5 — SiliconANGLE, 2026-07-19
- DeepSeek V4: Specs, Pricing, Benchmarks, and How to Run It (2026 Guide) — Morph, 2026-08-21
- China’s AI App Semi-Annual Report: Douyin’s Doubao Leads with 382 Million MAUs, Alibaba’s Qianwen Surges 58-Fold — BigGo Finance, 2026-07-15
- Baidu Targets Enterprise Cloud Market with a Raft of Industry-Specific AI Agents — Caixin Global, 2025-06-10
- China’s AI start-up funding triples to US$16 billion in first quarter amid bets on LLMs, robotics — South China Morning Post, 2026-04
- Alibaba commits US$53 billion for AI infrastructure in largest private computing project — South China Morning Post, 2025-02
- Tencent’s First-Quarter Profit Jumps 21% as Capex Tops USD4.4 Billion on AI Spending — Yicai Global, 2026-05-13
- Alibaba’s Qwen leads global open-source AI community with 700 million downloads — eHangzhou, 2026-01-13
- State of Open Source on Hugging Face: Spring 2026 — Hugging Face, 2026-03-18
- Chinese AI Models Now Capture Up to 46% of US Enterprise Token Usage — Forkast, 2026-07-10
- DeepSeek topped Ramp’s trending software vendors in June 2026 as US companies chase cheaper AI — The Decoder, 2026-07
- 80% Of Startups Using Open-source Models Are Using Chinese Models: Andreessen Horowitz Partner — OfficeChai, 2025
- China Considers Export Controls on AI Models, Data and Chips — TechRepublic, 2026-10
- Trump administration reportedly reviving push to ban Chinese AI models following Kimi K3 launch — Tom’s Hardware, 2026-07-20
- DHS bans Chinese DeepSeek AI from federal civilian agencies — The Hill, 2026-10-19
- China to promote AI Plus initiative (State Council guideline) — State Council of China (gov.cn), 2025-08-27
- China Unveils AI-Driven Plan to Build $14 Trillion Service Sector by 2030 — Caixin Global, 2026-04-22
- China leads world with over 1.57 million AI patents — WAM, 2025-04
- Shanghai to expand AI application by subsidizing computing power, LLMs, and more — Shanghai Municipal People’s Government, 2025-07-29
- Huawei to Double Output of AI Chip as Nvidia Wavers in China — Bloomberg, 2025-09-29
- China’s Humanoid Robot Output to Rise 94% in 2026; Unitree and AgiBot to Hold Nearly 80% Share — TrendForce, 2026-04-09
- Morgan Stanley raises China humanoid robot shipment forecast to 50,000 units — South China Morning Post, 2026-06
- These States Have Banned DeepSeek — StateTech Magazine, 2025-04