A few days ago, Huawei rolled out a comprehensive update for its Xiaoyi Claw model. As I placed my phone on the table and glanced at the system notification, a memory from three months prior suddenly resurfaced.

Back in March, an open-source AI project dubbed "OpenClaw" took GitHub by storm within a mere two weeks. However, by April, the excitement had waned—uninstallations surged, group chats quieted down, and the frenzy faded as quickly as it had begun, lasting less than a month.

What about mobile phone manufacturers? Huawei, the swiftest among them, had to wait over three months before finally rolling out "Xiaoyi Claw" to all users on June 25th. The slower ones lagged even further behind: vivo's X Fold 6 just introduced the "Atomic Workbench," where Atomic Notes assist in continuous writing, text refinement, and mind mapping. While it appeared vibrant, this launch came in summer, long after the "OpenClaw" craze had cooled. Honor was even later, announcing that its next-generation Agentic OS architecture wouldn't be released until July.

Everyone seems to be rushing to submit their summer homework before the deadline, creating a lively yet frenetic atmosphere.

If you were to plot these timelines on paper, it becomes evident that while vivo isn't the slowest, it is undoubtedly the most "consistent"—so consistent that every move lags half a beat behind others, to the point where the concept of "speed" seems irrelevant.

This is Shen Wei's deliberate choice. He refers to it as "choosing what not to pursue."

However, this time, the cost of "not pursuing" may be greater than anticipated.

"Daring to Be Last" Falls Short This Time

Let's begin with a series of contradictory decisions.

In early 2026, Duan Yongping posted on Snowball: "In the new year, I need to seriously learn how to use AI." This early-retired investor, who rarely discusses trends, made a rare statement about AI. Six months earlier, he had predicted that among those developing large models, probably no more than 10% would survive in ten years. "I would bet on Gemini because Google is very powerful and has woken up."

Judgment is one thing, but actions speak louder. By the end of the first quarter this year, the total market value of Duan Yongping's U.S. stock holdings exceeded $20 billion. His portfolio adjustments were even more aggressive than his increase in NVIDIA holdings a year ago—he nearly overhauled his entire portfolio.

NVIDIA is his biggest AI bet. In Q4 last year, he increased his holdings by more than elevenfold in a single quarter. In the first quarter of this year, he nearly doubled his holdings again, reaching 13.84 million shares worth $2.4 billion, becoming the third-largest holding in his portfolio. But this is just one facet of his AI strategy—he also newly acquired 3.4 million shares of Tesla, doubled his holdings in Google, and simultaneously invested in a series of AI supply chain companies such as Palantir, CrowdStrike, and Snowflake. His Apple holdings were reduced from 50% to 37%, Alibaba was completely liquidated, and TSMC was left with only a small position.

This is no longer just "buying a chip company" but a shift from an "Apple-centric" to an "Apple plus AI ecosystem matrix." Duan Yongping has reallocated $20 billion worth of assets for the AI era—he has not only bought shovels but also mines, the roads to transport the mines, and the miners.

Shen Wei took a different path.

His vivo did not invest its maximum resources in chasing large models. Instead, he leveraged his most valuable assets—user data access points, channels, and brand—to connect with others' "brains." The intelligent assistant in his phones is linked to ByteDance's Doubao, and the overseas version is connected to Google's Gemini. At the Boao Forum in March this year, he officially announced the establishment of the vivo Robotics Lab, offering high salaries to attract talent, with technical planning positions offering annual salaries of over 1.2 million yuan, marking vivo's formal entry into the home robotics market.

Both master and apprentice understand the same principle: At the model level, the giants will ultimately dominate. But their solutions are diametrically opposed—Duan Yongping's approach is to buy these winners and become their shareholder; Shen Wei's approach is to connect with these winners and become their customer.

Duan Yongping operates from behind the scenes, while Shen Wei sits at the table and plays.

This is not just Shen Wei's choice. It reflects the broader reduction the entire mobile phone industry has been forced to make in the face of AI. But Shen Wei has embraced it most thoroughly—while others are still hesitating about whether to develop their own "brain," he has already written "not pursuing" into the company's annual keywords.

To understand Shen Wei's move, one must first grasp the mindset he shares with Duan Yongping.

Duan Yongping has long advocated, and OV has followed for thirty years: "All masters dare to be last in the world, but they just do it better than others." The idea is not to be a pioneer. Blazing a trail in a new category requires significant effort to cultivate the market and educate consumers, a process that is extremely slow and costly. It's better to wait for others to figure out the demand and then enter, making the product the best. Be a little slower, be a little more stable, and strive to be the best among the latecomers.

This strategy has been masterfully executed by vivo. After Apple validated touchscreen phones as a necessity, vivo entered the market and became a leader in China with its imaging and audio capabilities. It believes that time is on the side of those who do better.

Shen Wei himself is a staunch believer in this mindset. He rarely accepts interviews, appearing in public no more than two or three times a year, but each time he reiterates the same message—focus on planting causes, and the fruits will come naturally. At last year's annual meeting, he inscribed the four words "choosing what not to pursue" into the annual strategy, stating, "Choosing what not to pursue is determination and wisdom; it determines the bottom line of a company's existence." He also said, "Bet less, but bet heavily."

At the same event, he rarely admitted to anxiety, saying that the mobile phone market "is expected to shrink further." But he immediately added, "We don't compete for short-term speed; we compete for long-term depth."

These words from Shen Wei serve as both a boost to himself and an answer to all those who ask him, "Why aren't you in a hurry?" At the annual meeting, he even clarified the pace: After mobile phones, MR will take three to four years, and robots will take more than five years. Let others fight the current battles; vivo will first improve agents like imaging and photo albums that do not require real-time response. What he wants is not to be the first to cross the finish line but to be the last one still running.

The words are restrained, but coming from Shen Wei, who is usually quiet, this is already his most candid statement to the outside world.

But AI does not recognize this sense of time.

Going back to the three-month time difference—this is not just vivo being lazy; it is an industry-wide issue. A mobile phone's operating system only releases a major version once a year. The on-device model installed inside is frozen in the system on the day of manufacture and can only be patched up afterward. The really complex tasks must be offloaded to the cloud (according to InfoQ's breakdown of on-device-cloud collaboration mechanisms). Meanwhile, large models in the cloud are almost being upgraded every month. Capabilities running on GPT-5 in March may already be two generations behind by June.

Google is relatively more flexible, but it is pursuing two paths—one for its own Gemini and one for everyone else. Google has given Gemini system-level access to core apps like Phone, Messages, and WhatsApp. Without root access, Gemini can read your text messages, make calls, and reply to WeChat for you. But what about third-party AI agents? They are still stuck outside the sandbox. The barriers for an AI to read the screen and operate apps on behalf of users have not been lowered at all. In March this year, the National Internet Emergency Response Center specifically issued a risk warning about prompt injection and malicious plugin attacks on OpenClaw, and the Ministry of Industry and Information Technology followed up with a notice. Several universities even required students to uninstall it. Technologically slow, platform providers are also getting involved—compliance and competition, two barriers.

In Duan Yongping's "daring to be last in the world," there is a hidden premise: Pioneers are slow and expensive, so being a latecomer is more stable. This premise has collapsed in the AI era. Pioneers can complete the entire process in two weeks, and latecomers can no longer catch up.

Shen Wei is not unaware of this. He just hasn't found another way to survive yet.

Nine Factories, One Brain

Shen Wei did not chase blindly. He took "not pursuing" to the extreme.

vivo's self-developed Blue Heart large model is not bad. Blue Heart Xiaowei can also write poems, polish text, and create summaries. But Shen Wei clearly has no intention of making it the "smartest." His biggest bet is placed three to five years in the future—the Robotics Lab, betting that mobile phones will accumulate data, MR headsets will train spatial perception, and serve as the eyes and brains for future home robots. As for the connected brain needed in mobile phones right now, he handed it over cleanly: Doubao in China, Gemini overseas.

On the same question, his fellow apprentices gave opposite answers.

OPPO's Chen Mingyong, who also studied under Duan Yongping, is much more impatient. More than two years ago, he sent an internal letter to all employees, declaring "the first shot of AI phones," saying that AI phones represent the third stage after feature phones and smartphones. He immediately established an AI center, allocating resources to AI without limit, going all in on AI (according to reports from IT Home and QbitAI). But to this day, in ColorOS, the areas that truly rely on models still follow the on-device-cloud collaboration approach. The global version even directly incorporates Google's Gemini. One is high-profile and rushing ahead, while the other is quietly "not pursuing," ultimately converging on the same thing in the end—neither has truly seized control of that "brain." Shen Wei did not join the hype, saving hundreds of millions in R&D expenses.

The most competitive-looking is Xiaomi. It is the only mobile phone manufacturer that wants to bring the system, chips, and large models under one roof—self-developing HyperOS, self-developing Surge O1 chips, and in March this year, rolling out a self-developed model with one trillion parameters, MiMo-V2. Lei Jun declared that Xiaomi's annual R&D investment in AI this year would exceed 16 billion yuan. The direction is right, and the time to catch up is tight. But in the year or so when its self-developed model was not yet mature, its Super Xiaomi AI first connected to DeepSeek and Doubao. The dream of a full stack is complete, but it still lacks a usable self-developed brain in the dream. Shen Wei did not make chips or invest 16 billion yuan, but his Atomic Workbench, developed in cooperation with ByteDance, has already been delivered on the X Fold 6.

Even more eager to make a statement is Honor. Last year, it announced the "Alpha Strategy," pledging to invest $10 billion over five years to transform into an AI terminal ecosystem company. In June, it "defined for the first time" AgenticOS, with the product line president, Fang Fei, describing it as "equivalent to redoing the hardware with AI." It all sounds impressive, but the complete architecture of this system will only be released this month, and its capabilities will depend on subsequent versions to gradually materialize. Yet, in the past year of dense press conferences, Honor's domestic market share has dropped from first place in the first quarter of last year to sixth (according to Counterpoint data). The press conferences have become louder, but the cards have become thinner. Shen Wei did not hold many press conferences, keeping his words low-key at the Boao announcement.

Even Apple has to admit defeat. Its new Siri, which has been talked about for two years, was delayed and settled for a $250 million compensation. When it was finally unveiled at WWDC in June, it switched to using Google's Gemini to drive it—the company that has achieved the ultimate in software-hardware integration has concluded that it needs to rent someone else's brain. The Chinese version is even more awkward, relying on Alibaba's Tongyi Qianwen as a backup, but it is still stuck in regulatory approval and was accidentally launched for three hours at the end of March before being withdrawn. Shen Wei made the same choice a long time ago—Doubao in China, Gemini overseas—but he did not make it into news.

Putting all these pieces together, you will see a striking fact: By the middle of last year, nine of the top ten global mobile phone manufacturers had deeply bound themselves to ByteDance's Volcano Engine, with the Doubao large model covering over 400 million devices. ByteDance has made it clear that it does not make mobile phones but only cooperates with manufacturers at the "operating system level." Later, even WeChat joined in—on June 4th, it announced that it would gradually connect its intelligent agent with Huawei, Xiaomi, OPPO, and vivo—more and more key access points in mobile phone manufacturers' systems belong to others.

Nine mobile phone manufacturers share one brain. Shen Wei's "not pursuing" has placed him within the coverage of this brain as well.

The sole exception is Huawei. With its proprietary HarmonyOS, Pangu large model, and Kirin chips, it stands alone in retaining full control over its technological core and has successfully delivered this comprehensive solution to users. Holding the reins of the system, model, and chips simultaneously represents the only viable path forward. Xiaomi, OPPO, Vivo, Honor, and even China’s counterpart to Apple are not destined for this trajectory.

Where Has All the Profit Gone?

Expanding our perspective reveals that this is not merely a question of strategic choice but also one of profit distribution.

Over the past two years, mobile phone manufacturers have faced profit squeezes from both ends. Firstly, at the upstream level, storage prices have skyrocketed. According to Counterpoint, storage prices surged by 40-50% in the fourth quarter of last year and another 40-50% in the first quarter of this year. Memory now accounts for over 20% of the material costs of a mid-range Android phone, up from around 10% previously. Consequently, Xiaomi, OPPO, and Vivo have all reduced their orders for complete devices this year. Vivo even implemented a blanket price increase of 100 to 700 yuan in March, but this move only resulted in fewer consumers upgrading their phones. The mobile phone, this "pile of iron," is becoming increasingly costly.

Now, let’s examine how much revenue this "pile of iron" can generate. Counterpoint’s Q1 2026 data reveals that Apple alone captured 42% of global smartphone revenue, maintaining stable pricing and even increasing its average selling price by 11% year-on-year. Half of the industry’s profits are concentrated in one company, leaving all Android manufacturers to vie for the remaining half. To put it colloquially, in this year’s smartphone battle, Huawei excelled in sales, Apple secured the profits, while Xiaomi, OPPO, and Vivo were squeezed at both ends.

This scenario mirrors exactly what transpired in the PC industry over two decades ago. Back then, Microsoft and Intel, with their Windows and CPUs, reaped over 90% of the industry’s profits, while manufacturers like Lenovo, Dell, and HP shared less than 10% and often hovered around the break-even point.

So, where does the money ultimately end up?

The first beneficiary that may not immediately come to mind for many is Google. On one hand, it integrates Gemini into the Android foundation; on the other, it secures a deal with Apple—the new Siri is powered by Gemini, with Apple reportedly paying Google hundreds of millions of dollars annually. The iOS and Android camps, once locked in fierce rivalry, now have their smartest components driven by the same company’s model. Sensor Tower data indicates that over the past year, ChatGPT’s share in generative AI has dropped from around 70%, while Gemini’s share has surged from single digits to 27.7%, and Alphabet’s stock price has more than doubled. Despite not manufacturing a single phone, Google stands at the forefront of this paradigm shift.

In China, it’s ByteDance that captures this value—the "intelligence" of nine phone manufacturers ultimately rests with its Doubao. It even partnered with ZTE to launch a Doubao phone, directly integrating large models into the system and bypassing the mainstream ecosystems of Huawei, Xiaomi, OPPO, and Vivo.

Among phone manufacturers, only Huawei has secured its own share. Shen Wei has not; his Vivo is still reliant on others’ technological brains.

Further afield stands a disruptor: OpenAI. Its AI device, developed in collaboration with Jony Ive, is planned to debut in the second half of this year, with rumors of a partnership with Qualcomm and MediaTek to manufacture its own phones by 2028, targeting 300-400 million units annually. Even Altman admits it’s a new category, not meant to replace phones—the moment to truly overturn smartphones hasn’t arrived yet.

In Closing

Returning to the phone manufacturers rushing to release new products this summer.

What they’ve unveiled looks impressive—Huawei’s Xiaoyi Claw, Vivo’s Atomic Workbench, Honor’s Agentic OS set to launch in July. But beneath the surface, most phone manufacturers truly hold only hardware arrangements. Yet, systems will inevitably be replaced. Even in terms of app activity within phones, these fragmented work demands will be supplanted by co-work products like Work Buddy and Marvis, quietly eroding phone manufacturers’ software pricing power.

Thirty years ago, Duan Yongping imparted to Shen Wei the mantra "Dare to be last, silently cultivate causes"—the most effective mindset for the smartphone era. Slowing down was acceptable; time would favor those who excelled. Back then, Vivo could afford to wait—waiting for Apple to validate touchscreens as a necessity, for market education to complete, and then reaping the rewards. Those words likely still adorn the wall of Shen Wei’s minimally lit office.

The imperative to "seriously learn how to use AI" feels less like a sentiment and more like a final confirmation before placing an order.

The master has acquired the entire map. The disciple is still selectively opting out.

But to be fair.

Shen Wei’s heavy bet on robots and spatial awareness may be a gamble that the next gateway lies not in phone-based models but in bodies capable of navigating the physical world. Phone cameras as eyes, MR headsets for spatial computing, on-device chips running perception algorithms—these fragments point not to a cloud-based brain but to a body that can move around your home, serve tea, and care for the elderly. If that’s where the future lies, his current "silent cultivation" might just be planted in the right, albeit farther, field.

However, the Robot Lab was only established in March this year. Talent is still being recruited, technical routes explored, and products are still just shadows.

Meanwhile, the battle has already reached mid-mountain. Doubao covers 400 million devices, WeChat’s intelligent agents are integrating with phone manufacturers, and Gemini’s global share has surged to 27.7%. Whether Shen Wei’s shield of "opting out" can withstand the tides of the AI era, not just the artillery of the smartphone era, remains uncertain.

The track remains the same, but after the starting gun fires, no one is waiting.

Phone manufacturers can no longer view each other as mere competitors; in the AI era, their rivals are everyone.