The reigning champion's sprint to the finish line is imminent.

Author | Li Yujia

Editor | Gu Nian

In July, an IDC report positioned Alibaba at the pinnacle of the AI Coding arena.

According to the report, Qoder, an Alibaba subsidiary, leads the pack with a 47.6% market share. Zhipu CodeGeeX follows with 11.5%, SenseTime Raccoon with 10.5%, Tencent CodeBuddy with 6.9%, and Baidu Comate with 6.0%. The cumulative share of the second to fifth-ranked companies still falls short of Alibaba's individual standing.

Claims such as "Alibaba commands nearly half of China's AI programming market" and "Qoder is the undisputed leader" swiftly became headline tags for numerous technology and finance media outlets.

Focusing solely on this ranking, Tencent and Baidu do not seem to be performing as well.

With market shares of 6.9% and 6.0%, respectively, they appear somewhat weak as "challengers" when compared to Alibaba's commanding 47.6%. However, Alibaba's decision to claim the 2025 title in July 2026 underscores a certain perseverance—the belief that patience yields rewards.

According to tech influencer MacTalk, based on revenue, Alibaba remains the frontrunner as of July 16 this year. This implies that, starting from July 16, Alibaba needs just 167 more days to defend its AI Coding championship.

However, AI programming is not akin to liquor or home appliances. Liquor prices can be influenced by channel inventories from years past, and appliance sales can be linked to previous real estate cycles. AI Coding, on the other hand, operates differently. The product landscape in this sector evolves too rapidly.

Last year, the emphasis was on code completion; this year, it has shifted to the comprehensiveness of projects, task files, testing, and other full-scenario agent performance metrics.

Alibaba has presented its achievements from an old battlefield as a coronation photo for a new one. Yet, just as it declares victory in one round, the cards for the next round have already been reshuffled.

How Significant Is Alibaba's Leading Position in AI Coding?

IDC's ranking methodology is quite "narrow," focusing solely on subscription fees paid specifically for AI programming tools. It excludes tools bundled with cloud servers, internally developed and utilized tools within companies, and those accessed by developers via API interfaces.

The degree of market segmentation under this methodology can be inferred from the market size. According to the IDC report, the market was valued at 399 million yuan in 2025.

What does a market size of less than 400 million yuan signify? Globally, according to Gartner, as of April 2026, the annualized size of the global enterprise-level AI programming market is estimated to have reached $9.8 billion to $11 billion.

If we calculate based on Alibaba's 47.6% share, Qoder's corresponding revenue would be approximately 190 million yuan, which is only about 1/360 to 1/410 of the global market.

For Alibaba as a whole, this is negligible.

In the first five months of 2026, the Token revenue of Alibaba Cloud's MaaS business surged 15-fold, reaching hundreds of millions of yuan per month. Alibaba CEO Wu Yongming anticipates that, in the second quarter, Alibaba's annualized recurring revenue (ARR) from AI models and application services will exceed 10 billion yuan and surpass 30 billion yuan by the end of the year.

This means that even if Qoder captures nearly half of the domestic AI Coding market, it accounts for less than 2% of Alibaba's AI model and application services ARR, which is expected to exceed 10 billion yuan. Compared to the anticipated 30 billion yuan ARR by the year's end, Qoder's revenue share would be diluted to less than 1%.

Of course, comparing last year's revenue with this year's estimates is not entirely rigorous. However, promoting last year's first-place finish this year may lack timeliness.

Qoder's performance last year only proves that Alibaba was swift in the previous stage. It does not yet confirm that Alibaba has secured the AI Coding business.

After all, the AI programming market in 2025 was in its infancy, with companies using "free" offerings to drive user growth and commercialization yet to commence. It was not until the first half of this year that the industry reluctantly reached a commercialization inflection point.

In February, Zhipu raised API pricing, and ByteDance's Trae switched to a paid model. In May, Alibaba's Qoder ended its long-standing free offerings, and Tencent raised subscription fees for its enterprise version. In June, ByteDance's Doubao launched professional paid packages. In July, Tencent introduced a personal premium version, Kimi reduced its free quota and adjusted pricing.

As companies officially transitioned to paid models this year, how valuable is Alibaba's claim of a "dominant lead" based on 2025 revenue data?

From a user base perspective, Qoder's dominant lead may not be as significant.

According to Analysys' "Q2 2026 China Office Intelligent Agent Platform Market Insights" released in July, Tencent's WorkBuddy officially launched its desktop version between March and June this year, reaching 20.97 million monthly visits in June, surpassing the combined total of ByteDance's TRAE IDE (domestic version) and Alibaba's QoderWork.

Of course, the 13.09 million additional monthly visits for WorkBuddy compared to Alibaba's QoderWork are not an insurmountable gap.

According to Alibaba's latest annual report, as of March 31, 2026, the group had 131,462 full-time employees. At this rate, if each employee simply visited QoderWork 3.2 more times per month, Alibaba could surpass WorkBuddy.

For example, on July 3, Alibaba issued an internal notice requiring all employees to stop using Anthropic's products, including Claude Code, in office environments starting July 10 and to use its self-developed AgenticCoding platform, Qoder, instead. Shortly after, Ant Group issued a similar notice.

While the intention was to gradually reduce reliance on overseas models, it is evident that in the rapidly evolving AI Coding sector, internal adoption by a large company's vast programmer workforce can provide significant support.

However, Tencent's CodeBuddy, which ranked fourth last year with only a 6.9% share, iterated into WorkBuddy this year and achieved user growth in just three months, overtaking competitors. This demonstrates that last year's product logic no longer applies to this year's new battlefield.

Why Is It More Challenging to Defend the Lead This Year?

A key variable this year is the rapid convergence of capability gaps among leading models.

According to data from third-party evaluation agencies like SuperCLUE, the comprehensive scores of first-tier domestic large models now differ by single digits. Relying solely on model parameters and benchmark scores makes it difficult to create a significant gap in product experience.

As a result, the focus of competition has naturally shifted from technical parameters to product experience, organizational efficiency, and ecosystem depth. Only those who can better integrate model capabilities into real-world scenarios and seamlessly connect with existing enterprise workflows can build true competitive barriers.

Unlike many AI application scenarios, AI-powered office work, including programming tasks, is highly verifiable. Whether code runs, tests pass, and office efficiency improves can all be measured objectively. Every advancement in model capabilities quickly translates into improved efficiency.

Therefore, major companies have come to realize that AI office work is not only a "litmus test" for model reasoning and agent capabilities but also one of the clearest commercialization paths currently available.

After all, the growth ceiling for a single programming tool is limited, while the market for full-scenario productivity platforms is much larger.

As a result, major companies have begun consolidating their business lines.

Alibaba was the first to demonstrate the limitations of its old layout through actions.

This year, it initiated a comprehensive integration of three internal Agent product lines: QoderWork, a desktop programming tool under Alibaba Cloud; Wukong, an enterprise collaboration tool within the DingTalk ecosystem; and MuleRun, an internally incubated product targeting overseas markets. All three product lines were consolidated into the final 「Qianwen Office,」 led by Chen Yusen, the newly appointed DingTalk CEO in June.

On July 27, Qianwen Office was launched and began limited testing, shifting its focus from a single programming tool to a full-scenario enterprise productivity platform.

If a single product line's layout was sufficient to maintain a leading position last year, Alibaba would not have spent six months making deep organizational and product adjustments.

Baidu also completed three rounds of R&D line consolidation in the first half of the year: merging the Comate product R&D teams under the cloud and middleware lines in January-February; integrating government, enterprise, and financial industry code solution teams into the mainline in March; and establishing a group model committee in May to connect underlying models with upper-layer programming tools, upgrading to the Comate full-process R&D platform.

On July 20, Tencent announced an internal organizational adjustment, further transferring QClaw product center-related businesses and some teams to Cloud Product Division VI. After the adjustment, QClaw and WorkBuddy fell under the same management system.

On July 30, ByteDance issued an internal letter integrating Doubao, Feishu, and Volcano Engine. The Feishu and Doubao product teams merged to form a new Doubao product team, led by Doubao head Zhao Qi. Feishu CEO Xie Xin and the Feishu product team now report to Zhao Qi.

The decision-making logic for major companies is simple: when industry directions are unclear, decentralized competition allows for multiple experiments to avoid missing opportunities. However, now that product forms and market directions are clear, continuing to disperse resources only exacerbates internal friction. Concentrating resources to build a unified platform-level product is the only way to seize opportunities in a larger market.

However, compared to Tencent and ByteDance's "first-mover advantage" in AI office work, Alibaba no longer leads in terms of product implementation or user mindset.

Tencent's WorkBuddy went live in March this year and, after several rounds of iteration, now covers full office scenarios, including document creation, data processing, code development, and meeting collaboration. ByteDance's Feishu Intelligent Partner was embedded deep within the Feishu suite earlier, achieving widespread adoption among enterprise clients.

In contrast, Alibaba's Qianwen Office only began limited testing on July 27, with the web and DingTalk built-in versions still pending. The user rights, data, and teams of existing products also require further integration. The urgency to launch before full preparation is complete reflects the pressure Alibaba faces.

The gap in user mindset and ecosystem layout has widened.

In developer communities like Juejin and CSDN, discussions about AI office tools mention Tencent's WorkBuddy and ByteDance's Trae more frequently than Alibaba's products. WeChat Index shows that, on July 30, WorkBuddy's popularity was an order of magnitude higher than Qianwen Office.

Alibaba has missed the first golden window for rapid AI office penetration in the first half of the year.

Currently, Tencent and ByteDance's agent development platforms are open to third-party developers, with the number of office-related agents and plugins growing rapidly.

While Alibaba's Qianwen App announced in June 2026 that it was open to third-party agents and skills, Qianwen Office, which only began internal testing on July 27, has unclear developer interface status, and its ecosystem is still in its infancy.

Alibaba's Obsession with Being "First"

As a seasoned player, Alibaba should understand better than anyone the weight of its 2025 "revenue-dominant first-place" finish in today's new competitive landscape.

Its continued promotion of this achievement is not just for pride but also to generate buzz and amplify its remaining trump card: "B-end Existing customers."

Alibaba Cloud and DingTalk's accumulated enterprise client base is Alibaba's most valuable asset in the AI office battlefield.

Public data shows that, as of the first half of 2026, DingTalk serves over 26 million enterprise organizations nationwide, deeply covering government, manufacturing, finance, and other process-heavy industries. Among Alibaba Cloud's millions of enterprise clients, government and large enterprises account for over 40%.

These organizations' accumulated structures, approval processes, and business data over more than a decade form barriers that pure C-end products cannot quickly overcome.

ByteDance's internal letter also states, "AI will profoundly impact B-end productivity and everyone's work style. This integration aims to better build high-quality products for office and productivity scenarios and better serve B-end clients."

As mentioned earlier, AI office work is still in the early stages of commercialization, and retaining some "free offerings" to drive user growth remains the most effective strategy. Thus, the revenue burden naturally falls on B-end enterprises.

DingTalk’s AI-driven native work platform, “Wukong,” rapidly gained traction among its existing customer base following its launch, securing significant benchmark orders from companies such as SF Express, which boasts a workforce of hundreds of thousands. Leveraging cross-selling opportunities and feature enhancements among its current clients stands as Alibaba’s primary growth strategy for AI-powered business-to-business (B2B) solutions.

For B2B enterprises, market share certifications issued by reputable institutions serve as a valuable asset in government and corporate bidding and evaluation processes. This is particularly evident in highly regulated sectors like finance and government services, where market rankings from leading consulting firms such as IDC often act as a decisive factor in securing multi-million or even multi-billion-dollar projects.

For Alibaba, leveraging a ranking to solidify its leading position in the AI programming sector not only helps cement its foothold in the advantageous category of programming tools but also provides an entry point to facilitate subsequent integrated procurement of full-scenario AI office solutions, computing power, and cloud services by enterprise clients. This represents the most direct incentive for Alibaba to concentrate its resources on promoting this IDC ranking.

The capital market also acknowledges the significance of rankings. The valuation of large-scale model companies is closely linked to their industry standing, with even a single shift in ranking capable of triggering substantial fluctuations in stock prices.

Within major corporations, achieving a top market ranking can secure increased budgets, larger teams, and higher strategic priority. The competition for rankings essentially becomes a battle for resource allocation within the organization.

The same principle applies to the recruitment of top-tier talent. In the AI field, where elite professionals are in short supply, a leading brand reputation often carries more weight than merely offering high salaries.

However, all current rankings and claims of being “number one” are merely periodic snapshots and tactics aimed at capturing user attention. The true outcome of industry competition will only be determined when products are deeply integrated into enterprise R&D workflows, and migration costs become prohibitively high for clients to consider switching.

Alibaba’s 47.6% market share last year resembles a carefully crafted industry calling card—sufficient for a striking brand campaign but insufficient to predict the final outcome of the competitive landscape.