In every tech wave, those who invent or refine technologies are remembered first, but those who monetize them most effectively are the ones who ultimately profit.

In the second half of the AI race, Chinese internet companies are rethinking one key question.

Over the past three years, the AI narrative has revolved around parameters, leaderboards, and capital expenditures, with Alibaba, Tencent, and ByteDance investing tens of billions in computing power and models. The race for model infrastructure has become increasingly clear—only a few players can afford to burn cash at this scale.

How can other internet companies compete in the AI era without being overshadowed by the giants?

Recently, Baidu, Kuaishou, and Meituan have each shared their strategies.

On September 19, Kuaishou revealed at a tech salon that its Data Agent had over 10,600 weekly active users, with user scale, frequency, and outcomes drawing attention. The next day, Meituan announced that its AI tool, "Smart Seller," now serves approximately 1.3 million Catering (catering) merchants. Baidu Smart Cloud reported a 9x quarter-over-quarter growth in user scale for Baidu Partner.

Whether in data services, on-demand retail, content creation, or enterprise productivity, these three internet companies are focusing on "how many people are using" their AI tools—not on model parameters.

Indeed, for internet companies with established scenarios, merchants, and hundreds of millions of users, the next meaningful step is to ask: How can AI penetrate our core businesses, rather than endlessly competing with giants on infrastructure?

After growth peaks, AI commercialization becomes a "renovation plan for old houses."

The golden age of mobile internet has ended—a sentiment felt more acutely this year.

Take Meituan’s financials: Last year, its core local commerce business swung from profit to loss, with nearly 60 billion yuan in profits burned on subsidies in the food delivery war. With traffic (traffic) dividends exhausted and inventory competition intensifying, change is inevitable.

Baidu, Kuaishou, and Meituan represent three types of internet companies urgently seeking transformation.

First, local service platforms like Meituan face the challenge of retaining merchants and users efficiently and cost-effectively after competition becomes the norm.

A Goldman Sachs report in April estimated Meituan’s food delivery market share had dropped from 75–80% pre-war to a long-term stable range of 50–55%. Management’s long-term guidance to investors shifted from earning 2 yuan per order pre-war to just 1 yuan.

In Q2 2026, Meituan’s operating profit rebounded to 2.69 billion yuan quarter-over-quarter, but the first half still posted a 4.67 billion yuan loss. While half of its market share was defended, profit margins never recovered.

Behind the scenes, Meituan has made many changes, including using AI to protect its supply side. "Smart Seller" covers operational analysis, reputation management, marketing, and food safety alerts, helping 1.3 million Catering (catering) merchants and 130,000 branded stores resolve approximately 8.6 million operational issues.

After the food delivery wars, omnichannel merchant operations are now standard, but Meituan hopes to do more to retain them.

Second, content platforms like Kuaishou face the challenge of revaluing their platforms with AI after traffic growth stalls and the content industry remains undervalued.

Kuaishou’s Q2 revenue reached 35.5 billion yuan, but growth slowed to 1.4%. Advertising revenue, its largest segment, grew just 4.4%, while live-streaming revenue dropped 13.5%. Adjusted net profit fell 30% year-over-year, marking a second consecutive quarterly decline. With 412 million daily active users, growth slowed to just 0.8% year-over-year. The "traffic-for-ads" model no longer works when user growth stalls.

All content formats—long, mid, and short-form videos—face this issue. Kuaishou’s advantage lies in its advanced AI Layout (deployment). Kling AI generated 850 million yuan in Q2 revenue, up over 200% year-over-year, with a global user base exceeding 100 million. In July, Kling AI raised nearly $3 billion at a post-money valuation of approximately $18 billion—nearly matching its parent company’s entire market cap.

Now, the key is to showcase how AI is transforming Kuaishou itself. Data Agent first improved internal efficiency to prove its value. According to Securities Times, by June 2026, over 92% of Kuaishou employees used its in-house Agent products, with AI contributing 60% of code from R&D staff.

Finally, complex tech conglomerates like Baidu—spanning advertising, smart cloud, autonomous driving, and even AI chips—must convince markets of their growth narrative shift.

Baidu’s AI penetration is already high. Traditional search advertising remains under pressure, but AI-driven businesses contributed 12.5 billion yuan to its 31.3 billion yuan Q2 revenue, accounting for over half for the second consecutive quarter. AI cloud infrastructure, Apollo Go, and Kunlunxin chips all performed strongly, with Morningstar and JPMorgan now valuing Kunlunxin at tens of billions independently.

But Baidu needs to attract more attention to its applications, hence the disclosure of a 9x quarter-over-quarter growth in Baidu Partner users.

Meituan, Kuaishou, and Baidu are all emphasizing AI application narratives.

In *Technological Revolutions and Financial Capital*, Carlota Perez divides tech revolutions into two phases: the installation phase, driven by financial capital chasing infrastructure, and the deployment phase, driven by production capital integrating technology into industries.

These three companies are moving into AI’s deployment phase.

From Parameters to Penetration: An Alternative Path in Big Tech’s Arms Race

The shift from competing on parameters to emphasizing penetration reflects an inevitable trend in the AI industry’s evolution.

One driver is the convergence of model capabilities.

Open-source models are catching up faster, and domestic computing power is maturing. "Whose model is stronger" is no longer a sustainable moat—today’s leaderboard top spot may become industry average within six months.

Leading large model companies report high revenue growth but sustained losses. Alibaba’s quarterly capital expenditures reached 67.7 billion yuan, up 75% year-over-year. The model layer’s competition has become a capital-intensive slog with uncertain returns.

Another driver lies in financial realities.

Companies like Meituan, Kuaishou, and Baidu occupy an awkward position.

By scale, they’re not cash-strapped: Meituan spent 7.67 billion yuan on R&D in a single quarter, while Kuaishou’s R&D spending grew 34.7% year-over-year. Both continue to increase investment.

But they lack the war chests of Alibaba, ByteDance, or Tencent, which treat multi-billion-dollar quarterly capital expenditures as routine.

For these companies, the only path is to accelerate AI penetration into their businesses, forming a virtuous cycle: penetration drives revenue and efficiency, which funds further R&D, deepening penetration.

The battle for AI penetration favors second-tier platforms.

Meituan, Kuaishou, and Baidu all possess three key assets: high-frequency real-world scenarios, granular operational data, and transaction loops for instant validation.

Every Meituan food delivery order, Kuaishou livestream, and Baidu search is an AI training ground.

With scenarios, merchants, and hundreds of millions of users, these companies don’t need to search for scenarios like AI startups—they already have them waiting for affordable technology.

Of course, Kuaishou’s 10,000 weekly active users are internal employees, and Meituan’s 1.3 million merchants use free tools. Today’s AI discussions focus on user counts, but profit-and-loss statements remain unsupported. Internal empowerment is just the early stage of AI applications—future competition will focus on external ecosystems.

Take Data Agent as an example: IDC evaluates 18 Chinese Data Agent vendors, but only 4 qualify as leaders, with selection criteria shifting from capability demos to scalable deployment.

Against this backdrop, capital markets must reevaluate internet tech. Compared to giants’ AI infrastructure revaluations, these companies better reflect application-driven changes—shifting from user engagement and monetization rates to AI penetration.

Valuation Anchors Shift: Three Companies Sit at the "Gateway"

This isn’t the first time valuation logic has changed for Chinese internet companies.

Over the past two decades, capital markets have twice recalibrated their pricing anchors.

The first shift was to user scale.

During the traffic dividend era, internet companies pitched "attention economics": user counts, time spent, and click-through rates.

The second shift was to profitability.

After 2021, regulatory cycles and stalled traffic growth triggered a collective pullback in Chinese tech stocks. Markets prioritized profits over growth and cash flow over narratives. Cost-cutting became industry-wide, with Tencent and Alibaba launching multi-billion-dollar buybacks.

Each anchor shift revalued companies. Those that failed to adapt were left with outdated valuations. Meituan, Baidu, and Kuaishou were all victims of the second shift.

Now, the third shift is underway: capital markets are starting to price AI assets separately.

Kling AI’s $18 billion standalone valuation, Kunlunxin’s tens-of-billions valuation reference, and penetration rates becoming standard disclosures signal a revaluation window has reopened. Currently, Meituan, Kuaishou, and Baidu’s market caps reflect their legacy businesses, with AI’s monetization potential barely priced in.

In the future, profit distribution across the AI supply chain will inevitably rewrite, and these companies are likely beneficiaries.

Nearly all AI profits in this cycle have accrued upstream. Dividing the chain into upstream devices, midstream cloud/model companies, and downstream applications, money now flows to "shovel sellers." But history shows this is an early-stage dynamic.

Cloud vendors’ massive capital expenditures today are effectively prepaying future profits to upstream hardware companies. Once capital expenditure growth slows and AI revenue scales, profits will shift downstream to clouds and applications.

In mature internet industries, profit distribution follows a pattern: upstream devices/tech take 20–30%, midstream networks/clouds take 10–20%, and downstream platforms/applications take 50–70%.

Proximity to user entry points correlates with higher shares. Meituan, Kuaishou, and Baidu sit precisely at these application gateways.

A similar trend is emerging in the U.S.: Meta, Google, and AppLovin’s advertising AI lead, followed by e-commerce and other traditional internet scenarios. These areas share strong ROI quantifiability and client willingness to pay.

Now consider the three companies’ positions: Meituan controls local commerce marketing and transactions, Kuaishou dominates advertising, e-commerce, and content subscriptions, and Baidu leads in cloud and search advertising—all scenarios with easily measurable ROI.

In China’s AI development roadmap, the application layer will be the most "narrative-rich" sector.

China offers a unified 1.4 billion-person market, the world’s densest mobile payment and instant delivery networks, and policy support for domestic demand stimulation.

For Meituan, Kuaishou, and Baidu, this is an opportunity to redefine their "tech" credentials.

Capital markets once focused on revenue growth and profit margins. Now, they must examine AI call volume, task execution rates, automation penetration, and whether pricing architectures have evolved.

Today’s numbers—1.3 million merchants, 10,000 weekly active users, 9x growth—may seem like utilization metrics, but in hindsight, they could mark the start of three new profit statements.

Supply chain profits always flow closest to users, a pattern seen once in the internet era and likely to repeat in AI.

Source: HK Stock Research Society