Behind Apple and Google's handshake on AI: Why are domestic mobile phone manufacturers afraid to revolutionize themselves?

The handshake between Apple and Google is not the end but the beginning—it proves that in the AI era, open cooperation is more viable than working in isolation.

At the beginning of 2026, the most shocking news in the tech circle was undoubtedly the AI alliance between Apple and Google—Apple will deeply integrate Google's Gemini core architecture into the next - generation Siri of the iPhone. The two long - time rivals have put aside their ecological barriers and injected a new AI soul into their flagship phones through technological cooperation.

The essence of this cooperation is that Apple has given up its obsession with developing AI in isolation and chosen pragmatism for growth. In contrast, in the domestic mobile phone market, Huawei, Xiaomi, and vivo are sticking to their own closed ecological loops. Newcomers like Doubao Mobile and Alibaba Qianwen are tentatively innovating on the periphery. The entire industry is caught in a dilemma of "wanting to revolutionize but not daring to make radical changes." Where exactly is the bottleneck in the path of integrating the underlying AI of mobile phones? And who will define the ultimate form of the next - generation AI phones?

01 Apple's "Reluctance to Self - Disrupt"

Some outsiders interpret Apple's partnership with Google as an "AI defeat and surrender," but in fact, it is the most shrewd business decision of this tech giant—It's not that it doesn't want to disrupt itself, but it is reluctant to pay the price of ecological collapse for the revolution.

Since Siri was launched in 2011, Apple has had a head - start in mobile AI. However, the slow technological iteration in a closed ecosystem has turned Siri from an industry benchmark into a "dumb assistant." Under the impact of ChatGPT and Gemini, the AI capabilities of the iPhone have become the biggest shortcoming for users when considering a phone upgrade. Data shows that in 2025, the shipment of Android AI phones accounted for more than 45%, while the high - end market share of the iPhone declined for the first time due to the lack of AI features.

Apple's dilemma is that its AI technology reserves lag far behind Google and Microsoft. Developing its own large - scale model not only requires tens of billions of investment but also needs to break the underlying logic of iOS, which prioritizes local processing and privacy—This means reconstructing the system architecture and opening up data interfaces, which is equivalent to disrupting the iOS ecosystem. In contrast, by cooperating with Google, Apple only needs to pay a licensing fee, and Siri can directly acquire the complex reasoning and multi - modal interaction capabilities of Gemini 3 Pro. At the same time, it can safeguard its privacy through the "local pre - processing + privacy audit" mechanism.

More importantly, this cooperation safeguards Apple's core interests: Apple has not ceded any control over the iOS ecosystem. Users still consume within the Apple ecosystem, and Google is just a "technology provider." Google, on the other hand, has achieved a monopoly on the mobile version of Gemini through 1.5 billion iPhone users, counterbalancing the alliance between Microsoft and OpenAI. This kind of cooperation, where both parties get what they want without breaking the closed loop, is essentially an exchange of interests between giants rather than a real self - revolution.

Apple has filled its AI shortcoming at the lowest cost, reminding the industry that the core of revolution is "value reconstruction" rather than "self - destruction." However, domestic mobile phone manufacturers have regarded the "closed loop" as a "life - saving charm" and fallen into another extreme.

02 The Dilemma of Domestic Giants

The AI strategies of Huawei, Xiaomi, and vivo have always been trapped in the double - closed loop of "self - developed model + self - developed system"—Huawei has the Pangu large - scale model + HarmonyOS, Xiaomi has the Pengpai large - scale model + Pengpai OS, and vivo has the Lanxin large - scale model + OriginOS. Each company is building full - stack capabilities from "chip - model - system - application," seemingly well - fortified but actually confining themselves.

They are not lacking in technical strength: The Mate 70 series of Huawei has a terminal - side AI computing power of up to 30 TOPS, enabling local multi - modal reasoning; vivo's Lanxin large - scale model is deeply integrated with OriginOS 6, and the XiaoV Circle Search 2.0 can complete the entire "recognition - execution" closed loop; Xiaomi's Pengpai OS connects mobile phones with IoT devices, and AI can link the whole - house smart devices. However, these capabilities are trapped within the ecological walls and cannot form a cross - platform general AI service—Users using a Huawei phone cannot call Alibaba Qianwen to handle e - commerce tasks; users using a Xiaomi phone cannot use Doubao's cross - application proxy to book a taxi or order a meal. The ecological closed loop has become an isolated island of experience.

There are three main reasons why domestic giants are reluctant to make radical changes:

First, they are too deeply bound by ecological interests. Take Huawei as an example. The HarmonyOS ecosystem bears the hope of transformation after the setback of its mobile phone business. Opening up the model interface may lead to the loss of applications and shake the foundation of HarmonyOS. Xiaomi's Pengpai OS is deeply integrated with the Mi Home IoT. The closed loop of AI services is directly related to the user stickiness of the IoT ecosystem. Once opened, the ecological resources invested in the early stage may be wasted.

Second, they are locked in by their technology routes. The AI layout of domestic giants started as a way to "make up for their shortcomings." In the early days, in order to reduce their dependence on Google's GMS and Microsoft's cloud services, they had to develop full - stack capabilities independently. Now, this "self - sufficient" model has become a habit. Transforming to cross - ecological cooperation requires reconstructing the technical architecture, and the cost is much higher than sticking to the closed loop.

Third, they are afraid of losing user control. Apple dares to use Google's Gemini because it can control user data through privacy agreements and data audits. Domestic giants lack Apple's bargaining power and are worried that after opening up the interface, user data will be controlled by external large - scale model providers, and they will eventually become just "hardware providers," repeating the mistakes of the feature - phone era.

As a result, we can see that domestic giants' AI phones are caught in "involutionary innovation": One company does quantization of large - scale models on the terminal side, and another does computing power optimization; one upgrades the voice assistant, and another optimizes the imaging AI. But no one dares to break the ecological barriers and make a real underlying integration—This is not a matter of ability but of courage, a conservatism of not daring to cut off the "closed - loop interests."

03 Newcomers Breaking the Deadlock

In contrast to the conservatism of the giants, new players like Doubao AI Mobile and Alibaba Qianwen have no ecological burdens and nothing to lose. Instead, they dare to do what the giants dare not and explore the boundaries of AI phones through "disruptive innovation."

The core concept of the Doubao AI Mobile launched by ByteDance is to skip the system closed loop and directly reconstruct human - machine interaction: Users don't need to open any apps. With just natural language commands, the AI can complete tasks such as ordering meals, booking taxis, and hotels across different apps. In essence, it is an "AI operating system for mobile phones." Although the Doubao Mobile has not been widely adopted due to the fragmentation of the Android ecosystem and the difficulty of obtaining authorization from leading apps, it has proposed the core concept of "AI agent" for the first time—AI is not an accessory function of the phone but the core, which is the key direction for the next - generation AI phones.

Alibaba Qianwen has taken a different path: It doesn't make hardware but focuses on being an ecological - level AI entry point. The Qianwen APP doesn't aim to cover all services on the Internet but delves deep into the Alibaba ecosystem, connecting dozens of apps such as Taobao, Ele.me, Gaode, and DingTalk. When a user says "Arrange a trip to Sanya next Wednesday," Qianwen can link Feizhu to book flights and hotels, Taobao to buy sunscreen products, and Cainiao to send luggage, forming a service closed loop of "command - execution - feedback." This model of "first deepening the self - developed ecosystem and then opening it up in a controlled manner" avoids the ecological barriers of the giants and solves the authorization problem of the Doubao Mobile, achieving the core value of "AI handling affairs" with a single app.

The innovation of new players has hit the pain points of domestic giants: The core of an AI phone is not a simple combination of "hardware + model" but a service reconstruction centered around "user needs." Giants are concerned about "whether their ecosystem can be preserved," while new players focus on "whether users can be more convenient"; giants calculate the "existing - interest account," while new players calculate the "incremental - opportunity account." This is the essential difference between "daring to disrupt" and "not daring to disrupt."

What's more noteworthy is that the exploration of new players has begun to benefit the industry: Alibaba Qianwen has cooperated with MediaTek to deploy Tongyi Qianwen on mobile phone chips, achieving cloud - edge collaborative AI services; Xiaomi has borrowed Doubao's cross - application proxy technology to upgrade the AI assistant of Pengpai OS. The "closed - loop walls" of the giants are being gradually broken by the innovation of new players.

04 Integrating the Underlying AI

For domestic mobile phones to break through the AI dilemma, they don't need to completely abandon the closed loop—the value of the closed loop lies in consistent experience and data security, which are what users need. Instead, they need to break the "closed closed loop" and build an "open closed loop." This is the core logic of integrating the underlying AI.

Integrating the underlying AI requires three steps:

Step 1: Technologically, achieve "cloud - edge collaboration + model interoperability"

The underlying contradiction of AI phones is the contradiction between "terminal - side privacy requirements" and "cloud - side computing power requirements"—Simple tasks (setting an alarm, checking the weather) require quick responses from the terminal side, while complex tasks (data analysis, creative generation) need cloud - side computing power support. Neither a single terminal - side nor cloud - side model can meet the requirements.

The cooperation between Apple and Google is essentially the optimal cloud - edge collaboration: Siri processes simple commands locally, and for complex tasks, it calls on the cloud - side computing power of Gemini. At the same time, it protects privacy through data pre - processing; the cooperation between Alibaba Qianwen and MediaTek has achieved a "model on the chip." When the network is disconnected, the terminal - side model provides basic services, and when connected to the network, the cloud - side model upgrades the experience. Domestic giants need to break free from the obsession with "pure terminal - side" or "pure cloud - side" and promote the interoperability between their own models and external high - quality models, allowing users to use Huawei's Pangu to handle local office tasks and Alibaba Qianwen to handle e - commerce needs on a single phone, adapting technology to user needs rather than the other way around.

Step 2: Ecologically, transform from an "exclusive closed loop" to an "inclusive closed loop"

The closed loops of domestic giants should not be exclusive ones that only allow their own services but inclusive ones that can accommodate others—HarmonyOS can be compatible with the API of Alibaba Qianwen, allowing users to call Qianwen's e - commerce services within the HarmonyOS ecosystem; Pengpai OS can open up the model interface, allowing Doubao's cross - application proxy to run on Xiaomi phones.

The core of the inclusive closed loop is "sovereignty in hand, service sharing": Giants control the system's underlying layer and user data sovereignty while opening up standardized interfaces to allow external high - quality AI services to access. This way, they can safeguard the core ecological interests and improve the user experience at the same time. This is not "self - disruption" but "transformation"—Exchanging partial ecological control for user retention in the AI era is far more valuable than sticking to the closed loop.

Step 3: Commercially, shift from "hardware - based profit" to "service - based profit"

The root cause of domestic giants' reluctance to open up the closed loop is their profit model's dependence on hardware: The more expensive the phone is sold, the higher the profit. The ecological closed loop serves to increase the hardware's premium. The profit logic of AI phones should shift from "selling hardware" to "selling services"—Apple plans to include Gemini's advanced features in the Apple One subscription package, and Google earns huge revenues through model licensing. This "hardware + subscription" model is the mainstream in the AI era.

When domestic giants' profits no longer depend on the hardware closed loop but on AI service subscriptions, model licensing, and ecological revenue sharing, they will naturally have the courage to open up the interface—Huawei can charge for AI service subscriptions in the HarmonyOS ecosystem, Xiaomi can earn money through model licensing of Pengpai OS. Hardware is just a carrier for AI services, not the only source of profit. This is the fundamental solution to breaking the "closed - loop dependence."

05 The Next Step for AI Phones

The cooperation between Apple and Google marks that AI phones have entered the "native reconstruction period" from the "function - addition period"—The next - generation AI phones will no longer be "smartphones with AI features" but "native terminals centered around AI," with three core characteristics:

First, interaction will be interface - less. The "natural language agent" attempted by the Doubao Mobile will become the mainstream interaction method—Users don't need to find apps or click buttons. With just speaking or typing, the AI can understand the intention and complete tasks. The physical interface and app icons of the phone will gradually fade away, and AI will become the only interface between humans and services.

Second, services will be integrated. The "ecological service closed loop" of Alibaba Qianwen will expand into a general cross - ecological service—Users can use a single phone to call high - quality AI services across the Internet to handle all - scenario needs in work, life, and entertainment. Ecological barriers will be broken by technical standards, and the experience consistency will far exceed the current closed - loop ecosystems.

Third, capabilities will be self - adaptive. After the cloud - edge collaboration technology matures, AI phones can automatically allocate computing power according to the scenario—When connected to the network in the office, it can call on the cloud - side large - scale model for complex data analysis; when the network is disconnected outdoors, the terminal - side model provides basic services; in privacy - sensitive scenarios, all processes are handled locally, truly achieving "computing power adapting to needs and privacy being protected at all times."

Going back to the initial question: Which Chinese mobile phone company has the courage to make self - disruptive changes? The answer may not be a single giant but the collective transformation of the entire industry—Huawei can open up the AI interface of HarmonyOS, Xiaomi can make Pengpai OS compatible with external models, and vivo can let the Lanxin large - scale model serve cross - ecological users. Only when each company is willing to cut off the small "closed - loop interests" can they share the big "AI - era" cake.

The handshake between Apple and Google is not the end but the beginning—it proves that in the AI era, open cooperation is more viable than working in isolation. If domestic mobile phone manufacturers can put aside their obsessions and embrace change with an "open closed loop," they may find an innovative path of their own in the competition for the next - generation AI phones, rather than missing the whole era behind the closed - loop walls.

This article is from the WeChat official account "Competition and Cooperation in Artificial Intelligence," author: Competition and Cooperation. Republished by 36Kr with permission.