On February 6th, within just nine hours of launching the Qwen APP’s “3 Billion Yuan Spring Festival Free Orders Event,” AI-driven orders surged past 10 million. Apart from me, who was on a strict pre-holiday diet, my entire social circle was buzzing about the milk tea they had ordered through the Qwen APP. A friend working at a leading internet company shared, “You won’t believe the craze! The milk tea shops near our office were inundated with Qwen orders and had to temporarily halt new orders!” This likely explains why some users experienced delays in their milk tea deliveries—merchants simply weren’t prepared for such an overwhelming demand.

The burning question on everyone’s mind: “Is Qwen’s surge sustainable?” In my view, regardless of short-term trends, Alibaba has made a strategic masterstroke by placing AI at the heart of its operations and integrating it seamlessly with its core e-commerce business. Its long-term vision is to position AI as the next-generation interactive gateway for the consumer internet. The “3 Billion Yuan Spring Festival Free Orders Event” wasn’t just a short-lived traffic spike; it was a “stress test of AI lifestyle” for consumers. For the first time on a large scale, AI’s interactive capabilities and practical value as a lifestyle assistant were validated in real, high-frequency, multi-scenario daily consumption. Overall, the vast majority of users embraced it and were satisfied.

Since the Qwen APP’s upgrade, Alibaba’s AI strategy has become crystal clear: focusing on both large-scale models and AI applications, rapidly transitioning AI from a “tech marvel” to everyday commercial use, spanning information, entertainment, and daily essentials. I must stress that during festivals like the Spring Festival, Alibaba’s diverse consumption ecosystem—including traditional e-commerce, instant retail, travel, transportation, and entertainment—holds immense appeal for users. Qwen has been seamlessly integrated with Taobao flash sales, Hema, and Tmall Supermarket, with plans to incorporate more transactional functions in the future. I believe that during this Spring Festival, a significant number of consumers ordered meals, fresh produce, and New Year goods through the Qwen APP. This isn’t just Qwen’s “breakout moment”; it’s also the first time hundreds of millions of users have experienced “AI transforming daily life.”

Of course, there are always skeptics in the market. Recently, a narrative has emerged: “American AI builds models, while Chinese AI makes milk tea,” attempting to portray Chinese tech giants as “backward” compared to their U.S. counterparts. As a seasoned AI user who frequently experiments with various large models, I must retort—don’t these critics know that Google Gemini launched an AI shopping guide feature last month, partnering with Walmart to enable direct orders through Walmart accounts? To enhance fulfillment, Walmart has also joined Google’s “Drone Delivery Program,” planning real-time delivery in 270 U.S. towns. By the way, Walmart and OpenAI have also announced similar cooperation intentions, though they haven’t officially launched yet.

As for Amazon, North America’s e-commerce behemoth, it has been actively testing AI shopping guides and plans to promote them through its Alexa smart hardware ecosystem. Unfortunately, Amazon’s own large model falls short, with few users. I suspect its most logical strategy would be to collaborate with Claude. From this perspective, the task of “AI making milk tea” is no easy feat for ordinary tech giants—it requires not only consumption scenarios but also a robust foundational large model, making it an extremely challenging composite task. Google has the latter, Amazon has the former, while Alibaba is a rare global case that “has both.” This is clearly a sign of progress, not backwardness.

I’d also like to point out that enhancing the consumer-facing (C-end) commercialization capabilities of AI applications is a groundbreaking proposition globally.

Currently, mainstream large model vendors abroad still rely on tiered monthly fees, with limited growth potential. OpenAI has long considered integrating ads into conversations but fears it would compromise user experience, even facing public ridicule from Anthropic recently. In reality, helping users execute e-commerce and other consumption transactions based on their active instructions and forming a closed transaction loop is a commercialization model that is not only more efficient but also enhances—rather than degrades—user experience. If you’ve ever made a purchase through the Qwen APP, you’d likely agree with my viewpoint, right?

By the way, for Alibaba, there’s no need to choose between “building models or making milk tea”—both can be pursued simultaneously. This isn’t just pioneering; it’s something many U.S. counterparts aspire to but struggle to achieve in an ecological closed loop. The recently released Qwen3-Max-Thinking large model is one of China’s most advanced deep reasoning large models, comparable to the best globally. Market rumors suggest Qwen3.5 will be released around the Spring Festival, which I believe is highly plausible. In the open-source model arena, since Meta’s LLaMA fell behind, Qwen has become the most widely used and derivative model-rich series globally, with only xAI’s Grok able to compete to some extent. In this track, even Google and OpenAI have to concede, as their technical development efforts focus on closed-source models, with minimal influence on the open-source ecosystem. (Of course, Qwen’s closed-source version is also exceptionally strong, which goes without saying.)

Some may ask, “If Qwen successfully integrates AI into consumption scenarios, won’t competitors simply imitate it, eroding Alibaba’s first-mover advantage?” I believe this is a misunderstanding—for the “AI shopping” business model, first-mover advantage isn’t the key. What matters first is a complete consumption ecosystem, an area where few competitors can match Alibaba. Second is full-stack technological capabilities. Imagine if hundreds of millions of Chinese consumers develop the habit of ordering goods through AI—how much computing power would that consume? What a challenge it would pose for cloud service platforms? Would AI applications need to become fully agentized to automatically execute complex tasks?... This isn’t just a test of technological infrastructure but also of application development capabilities.

As for logistics and fulfillment capabilities, there’s no need for further elaboration. Even Walmart, with its vast logistics network in North America, can only fulfill demands from Gemini in limited regions and relies on Google’s drones. In China, there’s probably no other platform besides Alibaba that possesses the “trinity” fulfillment capabilities of traditional e-commerce, instant retail, and fresh produce e-commerce. With the collective support of these resources, Qwen can efficiently respond to user needs, seamlessly access real supply, and complete closed-loop fulfillment. From the user’s perspective, the entire process takes just a few minutes or even seconds, but behind it lies decades of accumulation.

In the foreseeable future, the commercialization model of AI applications may break through on the C-end through consumption closures, while on the B-end, the most promising area is likely software development. In the past week or so, just after Anthropic released the latest version of Claude Code, OpenAI hurriedly released GPT-5.3-Codex, both featuring powerful agentized programming capabilities capable of closed-loop development of complex software applications. For enterprise clients, this is a tremendous boon, saving significant development costs. The upcoming new version of DeepSeek, rumored to be released soon, is expected to focus on programming improvements as well. We’re inching closer to the day when “AI handles daily software development.”

The Qwen large model is poised to make significant strides in this field as well. Considering Alibaba Cloud’s comprehensive development environment and its accumulation of enterprise-level users, it’s almost certain that Qwen will next focus on programming as a primary direction. If Qwen3.5 (or subsequent versions) significantly improves its programming capabilities and possesses certain agentized application development abilities, I wouldn’t be surprised at all. At that point, with C-end integration focused on consumption scenarios and B-end integration focused on application development scenarios, the commercialization ceiling for the entire AI business will be completely opened.

In any case, at this moment, we can be certain: Qwen’s surge represents one of Alibaba’s most successful strategic offensives in recent years. Moreover, it has made the capital market aware of the significance of Alibaba’s heavy investment in instant retail over the past year—if anyone still believes that the food delivery wars are “low-level redundant competition,” then the fact that millions of users ordered food delivery through Qwen should be enough to change their minds. The “lofty” AI and the “down-to-earth” food delivery have been organically combined here. Leaving aside the idea of leveraging a cup of milk tea to tap into a trillion-dollar lifestyle consumption market—isn’t the meaning of technological progress to solve real problems amidst the hustle and bustle of daily life?