等了一年多，国行iPhone用户终于能用苹果的AI功能了。
7月15日，网信部门发布备案名单，苹果智能（Apple Intelligence）正式完成端侧生成式AI服务备案。同批过审的还有华为、OPPO、vivo、努比亚等7款手机端侧AI服务——但最受关注的还是苹果。
原因很简单：这是苹果第一次在核心AI功能上明确使用第三方大模型。国行版Apple Intelligence将接入阿里巴巴旗下通义千问的部分能力，同时百度也提供了部分技术支持。那个「所有核心技术必须自研」的苹果，在中国市场妥协了。
为什么是阿里和百度？不是一个，是两个
苹果选合作伙伴的思路很有意思：不是all in一家，而是同时用两家。
据传阿里千问主要提供文本处理、图像生成和Siri增强的核心能力，百度则负责搜索增强和部分知识问答。双供应商策略，这是苹果的老传统了——就像它的芯片不只用台积电、屏幕不只用三星一样，关键供应链永远不能单点依赖。
更深层的考量是风险分散。中国的AI监管环境变化快，如果只用一家，万一那家出了问题，苹果整个AI功能都要停摆。用两家，东边不亮西边亮，至少能保证基础服务不中断。
对阿里和百度来说，这也是一个巨大的胜利。能成为苹果的供应商，本身就是一种技术背书——而且还是「国家级认证」的背书。更实际的好处是：数以亿计的国行iPhone用户，每天都在调用你的模型。这个调用量级别的训练数据和反馈，是花多少钱都买不来的。
妥协的不是技术，是市场
很多人说「苹果不行了，连AI都要靠别人」。这个判断太表面了。
苹果不是做不出来AI模型——Apple Intelligence的核心模型是苹果自己的，只是针对中国市场做了「本地化适配」。苹果的妥协，不是技术层面的，是商业和监管层面的。
中国市场对AI内容有特殊的监管要求：数据必须在境内、内容必须符合国内规范、模型必须通过安全评估。如果苹果坚持只用自己的海外模型，要么过不了审，要么体验差到没法用。与其给中国用户一个阉割版的AI功能，不如找本地合作伙伴，给用户一个完整的体验。
这不是苹果第一次这么干了。在中国，苹果的云上贵州数据合作、Apple Pay接入支付宝微信、App Store的内容审核……每一项都是「全球战略」对「中国市场特殊性」的妥协。
"苹果的中国策略从来不是『要不要妥协』，而是『妥协到什么程度才能既保住原则又保住市场』。"—— Dawn Vision编辑部
但这次妥协的分量不一样。以前的妥协都是「边缘业务」——数据存储、支付、内容审核。这次不一样，AI是下一代计算的核心入口。把核心入口的一部分交给第三方供应商，这在苹果历史上是罕见的。
这也从侧面说明了一件事：中国市场太重要了，重要到苹果愿意在核心战略上让步。iPhone在中国有几亿用户，中国的供应链支撑了苹果90%以上的产能。如果因为AI功能缺失导致中国用户换机意愿下降，那损失不是节省一点研发成本能补回来的。
明天见。
After waiting over a year, Chinese iPhone users finally get Apple's AI features.
On July 15, Chinese cyberspace regulators released a filing list — Apple Intelligence officially completed its on-device generative AI service registration. Six other phone AI services also passed, from Huawei, OPPO, vivo, and Nubia — but Apple was the one everyone watched.
The reason is simple: this is the first time Apple has explicitly used third-party LLMs for a core AI feature. The China version of Apple Intelligence will integrate some capabilities from Alibaba's Tongyi Qwen, with Baidu also providing partial technical support. The Apple that "everything core must be built in-house" has compromised in the China market.
Why Alibaba and Baidu? Not One, But Two
Apple's approach to choosing partners is interesting: not all-in on one, but using two simultaneously.
Reportedly, Alibaba's Qwen provides core capabilities for text processing, image generation, and Siri enhancement, while Baidu handles search augmentation and some knowledge Q&A. Dual-supplier strategy is classic Apple — just like it doesn't use only TSMC for chips or only Samsung for displays, critical supply chains never have single points of failure.
The deeper consideration is risk diversification. China's AI regulatory environment changes fast. If Apple used only one provider and that provider ran into issues, its entire AI functionality would grind to a halt. With two, there's always a backup — at minimum, basic services stay online.
For Alibaba and Baidu, this is also a huge win. Becoming an Apple supplier is itself a technical endorsement — and a "nationally certified" one at that. The more practical benefit: hundreds of millions of Chinese iPhone users calling your model every single day. That volume of training data and feedback can't be bought at any price.
The Compromise Isn't About Technology — It's About Market
Many people say "Apple is washed up, it can't even do AI on its own." That's a surface-level take.
It's not that Apple can't build AI models — Apple Intelligence's core model is Apple's own, just with "localization adaptations" for the China market. Apple's compromise isn't technical — it's commercial and regulatory.
China has specific regulatory requirements for AI content: data must stay within the country, content must comply with domestic standards, models must pass security assessments. If Apple insisted on using only its overseas models, it either wouldn't pass review or the experience would be too poor to ship. Rather than giving Chinese users a crippled AI feature, better to find local partners and deliver a complete experience.
This isn't the first time Apple has done this. In China, Apple's Guizhou Cloud data partnership, Apple Pay integrating Alipay and WeChat Pay, App Store content moderation — every one is a compromise of "global strategy" to "China market specifics."
"Apple's China strategy has never been 'whether to compromise' — it's 'how much to compromise to keep both principles and market.'"— Dawn Vision Editorial
But this compromise carries different weight. Previous compromises were all "peripheral" — data storage, payments, content moderation. This time is different. AI is the core entry point for the next generation of computing. Handing part of that core entry point to third-party suppliers is rare in Apple's history.
It also demonstrates one thing from the side: the China market is too important — important enough for Apple to concede on core strategy. Apple has hundreds of millions of iPhone users in China, and Chinese supply chains support over 90% of Apple's production capacity. If missing AI features reduces Chinese users' upgrade intent, the loss can't be recouped by saving a bit on R&D costs.
See you tomorrow.
苹果的中国策略从来不是『要不要妥协』，而是『妥协到什么程度才能既保住原则又保住市场』。
—— Dawn Vision编辑部
Apple's China strategy has never been 'whether to compromise' — it's 'how much to compromise to keep both principles and market.'
— Dawn Vision Editorial
Apple Intelligence · China approval · Alibaba Qwen · Baidu AI · dual-supplier strategy · localization · China market compromise
Sources · 信源 Sources
本文基于 Dawn Vision 认知引擎处理的 12 个源信号生成，经编辑部人工审核。素材来源：TechCrunch、爱范儿、环球网。
This article was generated by the Dawn Vision cognitive engine processing 12 source signals, with editorial review. Sources: TechCrunch, iFanr, Global Times.
