Domestic Tech Giants Accelerate AI Coding Push; Alibaba Qoder at 47.6% Revenue Share, Global Market $98-110B

Competition among major tech companies in the AI coding field is intensifying. Tan Dai, president of ByteDance's Volcano Engine, said the company has always regarded coding as a core direction. Liu Weiguang, president of Alibaba Cloud Public Cloud, stated in May 2026 that coding is the most important direction for Alibaba Cloud. According to IDC data, Alibaba's intelligent coding platform Qoder, which has been online for less than a year, achieved a revenue share of 47.6%. Currently, products such as ByteDance's Trae, Alibaba's Qoder, Tencent's CodeBuddy, Zhipu AI's CodeGeeX, SenseTime's Raccoon, Baidu's Comate, Huawei's CodeArts, and Moonshot AI's Kimi Code have all entered the market.

AI coding has become one of the main consumption scenarios for model inference compute. Anthropic's Claude Code saw its annualized revenue grow from $17 million in April 2025 to $2.5 billion in February 2026, surpassing $1 billion just six months after launch. OpenAI's Codex product also exceeded $1 billion in annualized revenue at the end of January 2026, becoming a core growth driver for the company. In February 2025, Anthropic released Claude Code, first moving the industry into the Agentic Coding phase, shifting the competitive focus from "Can AI write code?" to "Can AI independently complete software engineering?" In September 2025, Codex usage was only 5% of Claude Code's, but by January 2026 it had approached 40%.

Domestic tech giants started later in the AI coding field but are accelerating their catch-up. Ant Group subsequently issued a similar notice. Tencent and ByteDance have not explicitly banned Claude Code, but both are tilting toward their own models. Tencent's CodeBuddy supports multiple models but continues to strengthen the integration of its self-developed Hunyuan model with agent capabilities. In early July, after the Ministry of Industry and Information Technology's Network Security Threat and Vulnerability Information Sharing Platform issued a risk warning about a security backdoor in Claude Code, ByteDance internally treated it as a "red line."

ByteDance entered the AI coding space relatively late, with a path of first building an integrated development environment and then supplementing the model. Trae initially integrated multiple models rather than relying entirely on its own model. In November 2025, ByteDance released Seed-Code, but its performance still lagged behind Claude Sonnet 4.5. Since 2026, multiple internal application departments at ByteDance have been forced to use the Seed model. After the release of Seed 2.1 on June 23, Tan Dai said ByteDance had truly earned a seat at the table in the coding field. ByteDance officially stated that Seed 2.1 Pro's coding capabilities are on par with Claude Opus 4.6, with comprehensive usage costs reduced by nearly 80% compared to the Claude Opus 4.6 to 4.8 series.

AI coding is becoming an important driver of cloud service growth. OpenAI's first-quarter 2026 revenue was approximately $5.7 billion, with enterprise business revenue accounting for 40% of total revenue, and Codex was one of the core growth drivers. Anthropic generates 70% to 80% of its revenue from enterprise customers and API business, with gross margin jumping from 38% a year ago to over 70%. In April 2026, after ByteDance senior management visited Anthropic, the company internally adjusted its AI resource allocation, shifting focus toward serving enterprise products. Liu Weiguang, president of Alibaba Cloud Public Cloud, said that in the cloud computing era, internal software development and human outsourcing budgets are untouchable, but AI coding can hit 100% of those budgets.

Microsoft has integrated GitHub Copilot into GitHub and Azure, Google has incorporated Gemini Code Assist into Google Cloud's development system, Alibaba is strengthening its Alibaba Cloud development ecosystem around Qoder, Tencent is promoting synergy between CodeBuddy and Tencent Cloud, and ByteDance is enhancing the linkage between Trae, Volcano Engine, and the Seed model. Internal ByteDance data shows that over 90% of the code from the Trae team is written by AI, and per-person demand throughput rate has increased by 60%.

AI coding is one of the most fiercely competitive tracks at the model level. Token consumption for complex programming agent tasks can be 1,000 times that of traditional code Q&A tasks.

Alibaba Cloud's MaaS business saw token revenue grow 15 times in the first five months of 2026, with monthly token revenue reaching several hundred million yuan. Zhipu AI cumulatively raised API call pricing by 83% in the first quarter of 2026, yet call volume still grew 400%, and as of March, annualized revenue exceeded $250 million. After the price adjustment, Zhipu GLM-5.1's cache hit token price in coding scenarios approached Claude Sonnet 4.6 levels.

The high-frequency calls of AI coding also bring cost increases. Uber internally found that after employees widely used AI development tools, they consumed most of the annual AI budget within a few months, and therefore began restricting the use of some high-cost tools. Accenture restricted non-essential use of AI tools for some non-R&D positions to control overall enterprise AI spending. These cases reflect that as agents continue to run, AI coding is becoming one of the most expensive large model applications for enterprises. Reducing inference cost is the core of the next phase of competition.