Big Changes in the Smartphone Industry: Apple Bows to Google, DouBao Under Ecosystem Siege

On January 12, 2026, Apple and Google jointly announced a multi-year AI cooperation agreement: Apple will use Google's Gemini model to provide underlying support for Siri. According to Bloomberg, Apple plans to pay approximately $1 billion annually for this service.

Over the past two decades, Apple's core competitiveness has been built on the vertical integration of software and hardware—designing its own chips, writing its own operating systems, and controlling its own application ecosystem. But in the AI field, which is considered to define the next decade,Apple has handed over its most crucial brain to the creator of Android.。

Moreover, Apple is not the only smartphone maker to make this choice: Samsung has its self-developed Gauss, Huawei has its self-developed Pangu, but their flagship models are now all using external AI capabilities. Globally speaking,No smartphone manufacturer has produced an AI large model that can be compared with Google Gemini, ChatGPT, ByteDance’s DouBao, or Alibaba’s Qwen.。

This is not a failure of any one company, but a structural dilemma for the entire industry. However, the deeper issue is not about 'who provides the AI,' but rather 'in what form should AI enter the mobile ecosystem.' Just over a month ago, ByteDance’s DouBao attempted to let AI operate across apps directly—helping users compare prices, place orders, and make payments without opening any app. The result was that several major Chinese core applications collectively banned this feature within days.

One was accepted, the other was killed—the difference does not lie in how smart the AI is, but in whose cheese it moved. This is precisely the real issue behind Apple’s $1 billion order:

The most valuable capabilities of AI may well be the ones it is least allowed to exercise.

01 Smartphone manufacturers struggle to build top-tier AI independently

Apple has reached this point through a rather inglorious process.

At WWDC in June 2024, Apple made a high-profile release of Apple Intelligence and the new Siri. But the vision soon collided with reality. In March 2025, according to Bloomberg, Apple’s AI head Robby Walker described the delay as “ugly and embarrassing” at an internal all-hands meeting—internal testing showed that the new Siri could handle only 67% to 80% of user requests correctly, failing once every three to five interactions.

During the process of seeking external solutions, Apple also tested OpenAI and Anthropic. According to multiple media reports, Anthropic performed better in technical evaluations but demanded over $1.5 billion annually. In the end, Apple chose Google —which pays Apple approximately $20 billion per year to maintain its position as the default search engine on Safari.。

The situation is quite similar domestically. Huawei, as the Chinese smartphone manufacturer with the strongest independent technological capability, has integrated its Pangu large model into the Xiaoyi assistant within the HarmonyOS system. However, when DeepSeek gained popularity at the beginning of 2025, Huawei announced on February 5th that Xiaoyi would integrate with DeepSeek-R1. Honor followed up three days later, and OPPO also announced integration around the same time. The enthusiasm for self-developed models among various manufacturers began to wane, and accessing third-party models became the mainstream choice.

Why are smartphone manufacturers collectively absent?The fundamental reason lies in differing strategic priorities.。

Google and Microsoft invest nearly $100 billion annually in AI because it directly affects their core businesses — Google's search advertising and Microsoft’s cloud computing both face the risk of being disrupted by AI, and not investing would mean certain death. OpenAI is projected to lose about $9 billion in 2025 and won't break even until 2029 or 2030, but Microsoft is willing to bear this cost because it is a central growth driver for Azure. Smartphone manufacturers are in a completely different situation: AI is an additive feature for them, not a matter of life and death strategy. No smartphone company's board would approve spending five to ten years and hundreds of billions on a technology direction that is not closely related to selling phones.

Differences in strategic priorities are directly reflected in the scale of resource investment. In 2025, Amazon, Google, Microsoft, and Meta will collectively spend about $370 billion on AI-related capital expenditures; Apple's total capital expenditure for the same period will be only $12.7 billion, less than one-thirtieth of the former. With more than $200 billion in cash reserves, Apple chooses to pay Google $1 billion annually for AI capabilities, which speaks volumes.

Money is just the surface; the deeper gap lies in data and infrastructure. Google owns the world's largest search engine, Gmail, and YouTube, while Zuckerberg mentioned during an earnings call that Facebook and Instagram have 'hundreds of billions of publicly shared images and tens of billions of public videos.' What do smartphone manufacturers have? Local photos, contact lists, app usage habits —highly sensitive to privacy and low in text content, making them unsuitable for training general-purpose large models.。

In terms of computing power, Google started deploying its custom TPU chips as early as 2015 and has now iterated to the seventh generation; Microsoft and Amazon's cloud computing businesses have enabled them to accumulate the world's largest-scale data centers. The servers of smartphone manufacturers only need to support App Stores and cloud synchronization — they are not designed to train trillion-parameter models. Even if they were to start building now, there wouldn’t be enough time.

This doesn't even account for top talent. According to foreign media reports, there are fewer than 1,000 AI scientists globally with the capability to build top-tier large models. In July 2025, Ruoming Pang, head of Apple's foundational model team, was poached by Meta with a compensation package exceeding $200 million.Apple didn’t even attempt to match it—a figure far surpassing the compensation of any Apple executive other than Cook. When the bidding war for talent exceeds the salary structure of smartphone companies, this competition isn’t even on the same playing field from the start.

Talent, capital, data, computing power, strategy—falling behind in all five dimensions explains why no smartphone maker has created a top-tier AI large model. Since they can't develop it themselves, leveraging third parties becomes the only rational choice.

But here’s the question: What role should external AI play in entering the smartphone ecosystem?

02 How should AI divide the pie?

To understand AI's position in the smartphone ecosystem, one must first grasp the existing business model of smartphone manufacturers.

Take Xiaomi’s Q3 2025 earnings report as an example: the gross profit margin for its smartphone business was only 11.1%, while the gross profit margin for internet services reached 76.9%. The latter’s revenue sources include app store commissions, pre-installed app fees, ad distribution, and search promotion placements—In essence, smartphone manufacturers aren’t selling hardware profits but rather the entry point for users to access apps.The premise of this entire system is that users must actively 'open' apps to complete tasks, and the gateway to this action lies in the hands of smartphone manufacturers.

The AI partnership between Apple and Google has not shaken this system. According to Bloomberg, Apple pays Google approximately $1 billion annually for Gemini to provide underlying capabilities to Siri. In exchange, Gemini operates in a white-label form on Apple's private cloud servers, the user interface belongs to Apple, the data remains in Apple's hands, and users are unaware of Google’s presence.

The reason Google is willing to offer top-tier AI capabilities at a relatively low price is that it’s not buying revenue, but strategic security — keeping OpenAI and Anthropic out of Apple’s ecosystem while maintaining the approximately $20 billion annual default search agreement with Safari.

A key prerequisite for this arrangement to work is:The most advanced AI models in the US are closed-source.OpenAI’s GPT, Anthropic’s Claude, and Google’s Gemini do not disclose model weights; mobile manufacturers must pay, negotiate, and sign agreements to use them, making AI capabilities a scarce resource that can be locked down through exclusive partnerships.

The situation in China is completely different. DeepSeek’s R1 is open under the MIT license, Alibaba’s Qwen series has over 300 million global downloads and more than 100,000 derivative models, and Baidu will open-source its Wenxin large model in June 2025 — almost all of China’s most advanced large models are freely available. This means any manufacturer can access AI capabilities at the same level, and when AI is no longer scarce, it cannot serve as a differentiated moat.The focus of competition shifts from 'who can get AI' to 'who can use AI better.'。

But a question arises: If AI capabilities are accessible to everyone, who maintains the existing distribution of interests?

In December 2025, DouBao AI Phone provided an answer — not government regulation, not industry agreements, but the immune response of the ecosystem itself. The phone, co-developed by ByteDance and Nubia, features cross-application task automation. When a user says, “Order the cheapest bubble tea for me,” the AI can compare prices across multiple food delivery platforms, place orders, and make payments. The first batch of devices sold out within 24 hours, with second-hand prices soaring to ten times the original price, but the frenzy lasted only a day. WeChat, Taobao, Meituan, and Alipay collectively banned the AI's cross-application operations.

The logic behind the ban is clear: if users no longer open apps, no one views splash screen ads, information feed recommendations have no chance to display, and the carefully designed user pathways are completely bypassed; AI-driven cross-platform price comparisons directly compress profit margins that platforms rely on due to information asymmetry.

For mobile phone manufacturers, the threat is equally present — pre-installation fees, app store promotional spots, and negative-one-screen ads are essentially payments made by apps to be seen by users. If AI bypasses these entry points to directly access services, the value of mobile phone manufacturers as traffic intermediaries will be undermined. This also explains why ByteDance ultimately chose to collaborate with fringe manufacturers like Nubia.

The fates of the two models reveal the real boundaries of AI within the mobile ecosystem. The Apple-Google model has been accepted because it doesn’t alter the existing value distribution: AI makes Siri smarter, but it doesn’t make choices for users, bypass the app store, or break information asymmetry. The DouBao model was suppressed because it attempted to use AI to redefine the entry point itself. Mainstream Chinese manufacturers have recognized this — Huawei, Xiaomi, OPPO, Vivo, and Honor have all integrated third-party large models,but only for safe scenarios like Q&A, translation, and content generation; none dare replicate DouBao’s cross-application mode.。

Currently, a state of equilibrium in the mobile ecosystem has taken initial shape: AI can get smarter, but it cannot seize the entry point. In the U.S., this equilibrium is locked through exclusive agreements with closed-source models; in China, it is maintained through collective action within the app ecosystem.

The forms differ, but the logic is consistent — the value of AI is strictly confined to 'making existing experiences better' rather than 'redefining the experience itself.'

03 How long can this equilibrium last?

The current equilibrium — AI provides capabilities without seizing entry points — appears stable, but it rests on one premise: AI isn’t powerful enough yet. Once this premise changes, the equilibrium will be shaken.

Who might become the variable? The possibility lies very low with mobile phone manufacturers. As beneficiaries of the existing interest structure, they’re unlikely to achieve meaningful breakthroughs in self-developed large models. Their internet service revenue, which boasts a 76.9% gross profit margin, relies on the premise that “users must open apps.” Actively overturning this would be tantamount to cutting their own throats. The experiences of Nubia and DouBao have already shown that risky moves by fringe players only invite collective backlash from the ecosystem.

The more likely variables come from two directions:

One is AI agents within super apps. WeChat is already a self-contained ecosystem — mini-programs cover food delivery, ride-hailing, shopping, and government services, while payment loops are completed within the system. If WeChat deploys an AI agent within this ecosystem, when a user says, “Book me a restaurant for tonight,” the AI can compare prices, make reservations, and complete payments among the mini-programs, all without leaving WeChat. This model does not infringe on the interests of external apps since it never intended for users to leave; instead, it further strengthens the moat of the super app — the more users depend on the AI agent, the less likely they are to leave this ecosystem.

Second is the gradual migration of user habits. While DouBao was banned, what was blocked was cross-app operations, but it couldn’t stop users from doing more and more within AI apps—checking flight tickets, comparing prices, writing weekly reports, summarizing documents. This doesn’t trigger any blocking mechanisms but is slowly eroding the usage time of traditional apps. Once user habits are formed, traditional apps either proactively integrate AI or watch new players who cooperate with AI take over their market share. When incremental changes accumulate to a tipping point, apps will find themselves bypassed—not because AI seized the entry point, but because users no longer need that entry.

Apple spent a decade and invested over $10 billion in developing the Apple Car, only to announce its abandonment in 2024. One reason is that if it could only achieve L2-level assisted driving rather than full autonomy, the Apple car would simply be 'a better car,' failing to justify why Apple should be in the car business.

Smartphone AI faces a similar dilemma—If AI can only handle tasks like Q&A, translation, and photo editing—safe functions—it remains just a smarter assistant, one that won’t change anyone’s fate.The real differentiation lies in whether an AI agent can redefine how people interact with services, but this is precisely what the current ecosystem least allows to happen.

The current equilibrium essentially exchanges technical capabilities for ecosystem peace. How long this peace lasts depends on when AI becomes powerful enough that users no longer need the action of 'opening an app.' At that point, the definition of an entry point will be rewritten, rendering all of today’s arrangements obsolete. $Apple (AAPL.US)$$Alphabet-C (GOOG.US)$$Alibaba (BABA.US)$$XIAOMI-W (01810.HK)$$TENCENT (00700.HK)$

Disclaimer: This article is intended solely for learning and exchange purposes and does not constitute investment advice.

575K Views