Apple Taps Alibaba’s Qwen for China AI Launch

Apple will launch Apple Intelligence in China using Alibaba’s Qwen model, with testing from Nov. 4 and a full rollout in December. The move ends a two-year delay as local Android rivals use AI to drive upgrades.

The partnership gives Apple a compliant path to offer generative AI in one of its most important markets. It also gives Alibaba prominent access to Apple’s system-level features and large installed user base in China.

November testing, December rollout

Apple said Apple Intelligence will become available in China on Nov. 4 through the latest iOS developer beta. It plans a full release via a software update in December.

Like the overseas version, which uses OpenAI’s ChatGPT, the China version will rely on an external foundation model. Alibaba Cloud’s Qwen will be embedded into Siri, system-level writing tools and visual intelligence features.

The model will support generative text, image understanding and cross-modal content creation.

Compliance entity capitalized at $35 million

Corporate records show Apple prepared its local compliance structure well before the launch.

Tianyancha, a corporate records database, shows Apple Technology Development (Shanghai) was established in January 2025 with registered capital of $35 million. It is wholly owned by Singapore investment entity APPLE SOUTHASIA PTE.LTD.

The company’s business scope covers software development, big data and data-processing services. It is set to act as the domestic entity for algorithm filing and data operations for Apple Intelligence.

Tianyancha App also shows the partner, Alibaba Cloud, has registered capital exceeding 1.01 billion yuan (about $142 million). The source describes it as operating China’s largest public-cloud computing cluster.

The arrangement reflects the compliance steps foreign technology companies must take under China’s generative AI rules and data-export security review requirements.

Why Apple chose on-device technology plus Qwen

Overseas, Apple uses its Private Cloud Compute servers, custom chips and an agreement with OpenAI to create an on-device-and-cloud AI system.

In China, providers of public generative AI services must complete algorithm filings and security assessments. Rules also cover training-data compliance, content-generation filters and local storage of sensitive personal information.

Apple’s Shanghai entity provides a local legal entity for algorithm filing and creates an organizational compliance boundary from its overseas parent.

Alibaba Cloud and Qwen have completed the relevant filings and provide government- and enterprise-level security credentials. The source describes the partnership as Apple’s lowest-friction route into China’s generative AI market.

AI is now central to phone upgrades

Apple’s China launch also responds to pressure on high-end hardware sales from domestic rivals.

Since the iPhone 16 line went on sale, the absence of Apple Intelligence in China reduced the upgrade to a camera-control button and routine chip improvements. That weakened its appeal to business users and technology enthusiasts.

Huawei’s Celia on HarmonyOS NEXT, vivo’s BlueLM and Honor’s platform-level AI already support system-level intent recognition, cross-app tasks and document summaries.

With AI becoming a required feature rather than an optional extra, Apple’s position in phones priced above 5,000 yuan has come under pressure.

The November test and December rollout are timed for the year-end shopping season, including follow-on Singles’ Day promotions and the New Year and Spring Festival consumption period. Apple needs Qwen-powered Apple Intelligence to restore demand for upgrades.

Privacy controls versus Alibaba’s distribution gain

The deal also balances Apple’s control over its operating system with Alibaba’s pursuit of system-level traffic.

Apple has set strict privacy and permission terms. Users must approve each Qwen request through a pop-up, and requests and generated data cannot be used for secondary training of Alibaba’s models, according to the source.

That means Apple is not handing over system scheduling or core personal data from photos and notes. Qwen is effectively a cloud-based model used for tasks requiring greater computing power.

For Alibaba, becoming the preferred China partner for iOS system-level AI is still a major strategic gain. IPhone users will encounter Qwen through Siri and camera-based visual recognition, increasing its exposure against Baidu’s ERNIE Bot, Tencent’s Hunyuan and ByteDance’s Doubao.

User experience and cost risks remain

Apple still faces execution risks, starting with the hybrid architecture.

Overseas Apple Intelligence is designed to understand requests deeply across Mail, Messages and third-party apps. In China, Apple’s on-device model and Alibaba’s larger cloud model come from different technical systems.

Complex cross-app, multi-turn conversations could suffer latency, misunderstanding or inconsistent behavior in early testing. The repeated pop-up authorization required for Qwen may also reduce the smoothness of the interaction.

Computing costs are another challenge. If tens of millions of Chinese iPhone users can use Qwen-powered writing and visual recognition for free, token consumption could create large inference costs.

Unlike standard enterprise API contracts based on token usage, the long-term division of infrastructure costs between Apple and Alibaba will depend partly on whether the service drives Qwen memberships or other distribution revenue.

Real-world test begins in November

The $35 million Shanghai entity and Alibaba’s cloud infrastructure mark Apple’s latest step to fit its AI strategy into China’s regulatory system. Tim Cook has paired a local compliance entity with a major Chinese model provider to bring AI back to the iPhone.

The industry’s remaining question is whether the China version can deliver a seamless experience despite the split architecture, permission prompts and heavy computing costs. November’s test will provide the first practical answer.