Reuters reports Apple has trained a large language model built specifically for the Chinese market, with Alibaba providing the compute and technical support. It would be the first time a Western tech giant ships its own AI model in China – and a clear sign that borrowing models from local partners is no longer enough.
On Friday, August 14, Reuters – citing three independent sources – revealed that Apple has essentially completed training of a large language model designed exclusively for the Chinese market. Alibaba played a central role in the effort, supplying compute capacity, technical support, and local data expertise. Neither Apple nor Alibaba officially commented, which is standard in such cases, but the scale of the leak and the number of sources suggest the story is real.
The end of the borrowed-model strategy
Until now, Apple operated in China the way it does in most countries where it lacks its own model: integrating models from local partners into its devices. In February last year, Alibaba chairman Joe Tsai confirmed his company had been chosen as Apple’s main partner in China – hence Qwen models powering Siri and writing tools. Earlier this month, Apple’s Chinese support page briefly showed instructions for connecting Siri to Qwen on Macs, which were later quietly removed (we covered that in a separate post).
Now the strategy is shifting. According to Reuters, Apple is moving to a two-track approach: on one side its own model, trained from scratch; on the other, partner models (Qwen and Baidu technology remain part of the plan). An in-house model means Apple stops being hostage to someone else’s release schedule and can control the user experience end to end. That matters especially in a market where on-device AI quality has become a selling point.
Why Apple has to do this at all
China is the only major market where Apple must play by local rules: data stays in the country, model certification is mandatory, compliance is the foundation. OpenAI and Anthropic models (ChatGPT, Claude) simply do not exist in China – they are blocked. A foreign model also has to be adapted to Chinese content and data-security regulations.
Importantly, Apple has already cleared the first administrative hurdle: in July, Chinese authorities published that Apple’s generative AI service (Apple Intelligence) was registered on July 8 and can be used on iPhones in China. Reuters stresses that if everything goes according to plan, Apple could become the first foreign company whose own AI model is approved for the Chinese market.
A multi-billion market still missing AI
The stakes are enormous. Greater China (including Hong Kong, Macau, and Taiwan) accounts for roughly 20% of Apple’s revenue. Interestingly, the company is growing there even without AI: in Q3 of fiscal 2026, Greater China revenue reached $18.8 billion, up 22% year over year, driven mainly by iPhone sales. But analysts and Reuters itself note that the absence of AI features is one factor holding back even faster growth – especially since Chinese rivals like Huawei have been selling phones with full AI packages for a long time.
What this means for the market
This story matters for several reasons. First, it confirms that the Chinese AI market runs on its own rules – even the largest Western hardware maker must adapt its technology to local requirements and find a local partner. Second, Alibaba solidifies its position as a key infrastructure player, spanning Qwen models and compute for Apple. Third, Apple’s own model in China signals to the whole industry that the two-track strategy – own model plus local partners – is a viable path for foreign tech companies.
Apple Intelligence is expected to arrive in China with an iOS update within the coming months. No exact date was given, and the model’s technical details – size, architecture, benchmark results – remain a mystery. If Reuters’ reporting holds up, this will be one of the most important stories of the year in global AI.
This content was generated using artificial intelligence and reviewed by a human editor.
Sources
- Reuters (exclusive, Aug 14, 2026) – Apple trains China-specific LLM with Alibaba support, as cited by NetEase Tech, Sohu, Lianhe Zaobao
- Apple – Q3 FY2026 financial results (Greater China: $18.8B, +22% YoY)
- Chinese authorities – Apple Intelligence registration (July 8, 2026; published July 15, 2026)
- China AI Weekly Brief, period Aug 9-16, 2026 (job 3d03e2d4fae9)
- Earlier post on this blog: Apple Connects Siri to Alibaba’s Qwen on Macs in China – Then Pulls the Guide
This article is also available in Polish (PL).
Leave a Reply
You must be logged in to post a comment.