[Summary]

The main battleground in the Chinese AI market has shifted from comparing the performance of large-scale language models to how much AI agents can be incorporated into practice and how much can be converted into cash.

Why will the AI ​​agent market become the next huge market?

The answer is simple. This is because API charging for a single model tends to lead to price competition, but once agents are involved in business flow, EC, advertising, payment, and internal systems, it is easy to develop into recurring billing SaaS and platform fees.

The three companies to watch in 2026 are Tencent, Alibaba, and Huawei.

Tencent has the WeChat economic zone, which makes it easy to keep customer acquisition costs (CAC) low. Alibaba uses Qwen, DingTalk, and EC merchant infrastructure, making it easy to convert agents into SaaS. By combining Ascend, Pangu, and solutions for telecommunications, government, and state-owned enterprises, Huawei can target high-priced industrial agents.

However, it is not the intelligence of the AI ​​that determines the winner.

The key is to increase LTV, reduce CAC, absorb inference costs and capital expenditures, and finally leave cash as free cash flow (FCF).

The model competition is over and the competition for agent monetization has begun

In the Chinese AI market, model names tend to become the first thing talked about, such as Qwen, Hunyuan, Pangu, DeepSeek, ERNIE, and Doubao.

Of course, model performance is necessary as a foundation.

However, from an investor's perspective, the superiority of each model alone is not enough.

The reason is that model APIs are prone to price competition. If you just sell tokens cheaply, you will have to use a large amount of cloud computing resources, but your profit margin will be difficult to increase.

What the market will look at from here on is what to put on top of the model.

AI model
↓
AI agent
↓
Embedded into workflows, e-commerce, advertising, and payments
↓
Recurring SaaS, fees, and better ad efficiency
↓
Free cash flow

Companies that can create this trend can use AI not just as a cost, but as a tool to improve profit margins.

On the other hand, companies with strong models but unable to use agents as customer contact points are likely to see GPU costs and personnel costs always coming first.

Agent market revenue model

When looking at the AI agent market, it is better to separate ``why you will be charged.''

Even though the AI ​​is the same, the profitability is quite different.

Earnings modelProfitabilityKey points for investors
API chargesLowPrice competition is likely to occur, and inference costs reduce profit margin
Agent usage feeMedium to highIf you can charge on a per-job basis, it is more powerful than just token sales
SaaS conversionExpensiveMonthly billing, seat-based billing, and business-specific billing make it easy to generate recurring revenue
PlatformBestYou can even collect fees for payments, advertising, EC, and business systems

The most important thing in this table is not the AI model itself, but the "billing aspect" that sits on top of it.

APIs tend to be sold cheaply.

Agents are a substitute for work, so it is easy to get a good price.

If it can be converted to SaaS, it will be a recurring charge.

Furthermore, if you link advertising, payments, e-commerce, cloud, and internal systems, you will generate platform revenue.

The order that investors should look at is LTV, CAC, gross profit margin, inference cost, and FCF, not model performance.

Agent economic zone star table

The agent monetization capabilities of the three companies are summarized as follows from the editorial department's investors' perspective.

This is not an absolute evaluation. This is a relative evaluation based on public information as of June 1, 2026 and the business structure of each company.

CompanyAgent penetration powerMonetization powerFCF contributionPerspective
Tencent★★★★★★★★★★★★★★★WeChat and existing FCF are strong, making it easy to keep additional CAC low
Alibaba★★★★☆★★★★☆★★★☆☆EC and DingTalk have strong business connections, but cloud investment burden is heavy
Huawei★★★☆☆★★★★☆★★★★☆The range of penetration is limited, but the unit price tends to be high for government and industry

The reason Tencent looks stronger in this comparison is not model performance.

This is an existing contact point.

Qwen alone is not the reason why Alibaba looks strong.

It is a conduit that connects EC and business SaaS.

What makes Huawei look strong isn't consumer flashiness.

This is due to the depth of domestically produced AI infrastructure and industrial projects.

Tencent: WeChat Agent economic zone that can keep CAC low

The biggest part of Tencent's AI agent strategy is that it already has a distribution aspect called WeChat.

The total MAU of Weixin/WeChat is 1.432 billion as of 1Q of 2026. Tencent doesn't need to distribute new AI apps from scratch.

This is the decisive difference between us and companies specializing in AI.

If AI agents can be naturally integrated into WeChat, mini-programs, advertising, payments, and enterprise services, customer acquisition costs will be significantly lower. Of course it's not completely zero. Development costs, computing resources, promotion, and operating costs are required.

Still, companies that already have daily contact are strong.

In 1Q 2026, Tencent had sales of 196.458 billion yuan, non-IFRS net profit of 69.8 billion yuan, and FCF of 56.7 billion yuan. FCF increased by 20% compared to the same period last year. Additionally, the company explains that productivity AI agents such as WorkBuddy are gaining initial traction.

The next thing to see is what WeChat Agent will push.

Monetization routePoints to see
AdvertisingWill AI increase ad conversions and bids
Mini shopIs it possible to automate store operations, customer service, and purchase flow
PaymentShould we increase WeChat Pay transaction frequency and amount
Agent for companiesCan WorkBuddy etc. be used continuously and reduce the cancellation rate

Tencent is inferior to Alibaba in terms of the purity of its AI cloud sales.

However, the path to turning AI into cash is quite thick.

When looking at Tencent, investors should look at the growth in Marketing Services, FinTech and Business Services, Business Services, and FCF rather than the model name.

Alibaba: Can Qwen Connect Agents, SaaS, And FCF?

Qwen and Alibaba Cloud are not the only strengths of Alibaba.

There are business contacts that connect e-commerce merchants, DingTalk, the cloud, payments, logistics, and advertising.

Agents are a good match for this contact point.

For e-commerce stores, AI agents are more than just a chat function. If you can handle everything from product descriptions, image generation, inquiry response, inventory checks, ad copy writing, customer segment analysis, and campaign operations, it can replace labor and outsourcing costs.

At this point, it approaches SaaS billing rather than API billing.

In Q4 of the fiscal year ending March 2026, Alibaba's Cloud Intelligence Group sales were 41.626 billion yuan, an increase of 38% year-on-year. AI-related product sales were 8.971 billion yuan, representing 11 consecutive quarters of triple-digit growth year-on-year.

Agentic AI platforms like Wukong are also reported to be moving into DingTalk's internal business flow.

This trend is strong.

However, for investors, just being "strong" is not enough.

Alibaba's adjusted EBITA for the entire group was 5.102 billion yuan, down 84% year-on-year, with a FCF outflow of 17.3 billion yuan. The company attributes the decline in FCF to Quick Commerce, Qwen app user acquisition, and cloud infrastructure spending.

So Alibaba's problem isn't growth.

It is a return on investment.

Items to check on AlibabaReasons to watch
Cloud Intelligence Group's EBITA marginIs AI cloud growing with profit margin
AI-related product ratioAre you transitioning from server rental to high value-added services
Charging DingTalk/EC AgentIs Agent turning into SaaS revenue?
FCFAre you able to absorb capital investment and user acquisition costs?

Alibaba may have the biggest upside in the agent market.

However, until FCF returns, the market will continue to view AI as growing but cash being heavy.

Huawei: Seize the high-price market for government and industrial agents

Huawei has a different personality from Tencent and Alibaba.

It's not a company that expands its business all at once with consumer apps. The company is deeply involved in industries such as government, communications, finance, manufacturing, electric power, and urban infrastructure.

Huawei is not a listed company, so ordinary investors cannot buy its shares directly.

Still, it cannot be ignored when considering the Chinese AI agent market. The reason is the vertical integration that combines the Ascend chip, CANN, Pangu, Huawei Cloud, and communication infrastructure.

As the United States continues to restrict the export of high-performance GPUs, demand for domestically produced AI infrastructure is likely to increase in China's government, state-owned enterprises, and public infrastructure. Rather than being a flashy consumer app, Huawei's Agent is more likely to be used for operational monitoring, troubleshooting, manufacturing, city management, or autonomous communications networks.

In the official announcement, Huawei has emphasized infrastructure and operational automation for the AI ​​agent era, such as Atlas 950 SuperPoD, AI-Native Intelligent Operations, and Agentic Core, at MWC 2026.

Huawei's strength lies in its ability to easily take on high-priced projects.

In the government, state-owned enterprises, telecommunications, and finance, it takes time to introduce systems, but once installed, the cost of switching systems is high. This structure tends to result in a long LTV.

However, its weaknesses are also clear.

With the spread of private startups and the global developer community, it is hard to say that it has the natural diffusion power that Alibaba's Qwen does. Huawei itself is promoting the openness of CANN and supporting developers, but compared to global developer standards such as CUDA, the market is still being determined.

Huawei is an agent that penetrates narrowly and deeply, rather than an agent that spreads widely and thinly.

What investors should be looking at is not Huawei's own stock price, but how the company's domestically produced AI infrastructure will change the competitive environment for China's cloud, semiconductors, communications equipment, servers, and state-owned enterprise DX.

Comparison of 3 companies: Who is structurally more profitable?

When comparing the three companies in terms of their earnings structures, the differences are quite clear.

Comparison itemsTencentAlibabaHuawei
Biggest entranceWeChatQwen, DingTalk, ECAscend, Pangu, industrial infrastructure
CACEasy to keep lowMerchant base is strong but competitive investment is requiredProject acquisition type with long sales cycle
LTVLong in advertising, payments, and SaaSLong if EC/business can be converted to SaaSGovernment/industrial projects tend to take a long time
Inference cost burdenEasy to absorb with existing revenueCloud investment tends to seem heavyWe want to reduce it through vertical integration, but disclosure comparison is difficult
How FCF looksAlready a strong positiveRecent outflow due to investmentHard to compare in detail as it is unlisted
Biggest riskAI purity is low and easy to seeCAPEX and user acquisition costsClosed nature of private ecosystem

Looking at this table, Tencent is the strongest in terms of short-term cash generation.

Alibaba is large in terms of commercial scale and upside.

Huawei is strong in terms of depth in security and industrial projects.

However, none of them are perfect.

Tencent is not a stock to buy for the dream of an AI specialist. Alibaba needs to swallow the investment burden. Huawei is unlisted, and challenges remain for the expansion of private developers.

Bottom Line: LTV ÷ CAC And FCF Determine The Winner

What will determine the winners in the AI agent market is not model performance itself.

The key is to keep customer acquisition costs low, increase customer lifetime value, reduce inference costs, and ultimately maintain FCF.

If you're looking for a structural financial winner, Tencent is the most solid. We have a huge contact point called WeChat, and our existing businesses have strong FCF.

The likely winner on a commercial scale is Alibaba. If we can connect EC, DingTalk, Qwen, and Alibaba Cloud, we can see a path from Agent to SaaS and from SaaS to FCF. However, the investment trough is still deep.

The winner for security and industry is Huawei. Although it is not a direct investment target, it cannot be ignored as it will change the competitive map of China's AI infrastructure.

The numbers that investors should look at from here are not flashy AI announcements.

Advertising, Business Services, and FCF for Tencent.

For Alibaba, Cloud Intelligence Group's EBITA margin, AI-related product ratio, and FCF.

For Huawei, adoption of AI infrastructure for government and industry, Ascend/CANN developer ecosystem, and expansion into the private market.

The AI ​​agent market may seem to be a battle of technology, but it is actually a battle of profit models.

In the end, the battle for profit models always comes back to cash.

This article is intended to summarize the thinking behind investment decisions, and is not intended to recommend buying or selling specific stocks. Huawei is a private company, and general investors cannot directly buy or sell its shares. Chinese stocks, Hong Kong stocks, and US-listed ADRs are subject to risks associated with price fluctuation risk, exchange rate risk, liquidity risk, regulatory risk, geopolitical risk, and differences in accounting and disclosure systems.

Related articles

Source

This article is for educational and informational purposes only, based on public information. It is not a recommendation or solicitation to buy or sell any specific security or financial product. Although care is taken with accuracy, the content and future investment outcomes are not guaranteed. Final investment decisions should be made at your own judgment and responsibility.