AI Coding: Dominance of Big Tech Companies

Recently, significant news emerged from Silicon Valley.

AI programming startup Cursor is reportedly negotiating a funding round of $2 billion, with a valuation expected to exceed $50 billion. This comes just six months after its previous valuation of $29.3 billion.

Around the same time, another AI programming startup, Factory, announced it had completed a $150 million Series C funding round, reaching a valuation of $1.5 billion.

However, this is not a narrative of a single startup leading the charge.

Concurrently, Anthropic’s Claude Code is rapidly gaining market share, becoming the most favored AI programming tool among developers, with a preference rate of 46%, far surpassing Cursor’s 19%.

OpenAI’s Codex is also rising quickly; despite entering the market later than Cursor, it already commands 60% of Cursor’s usage.

In other words, the substantial funding for Cursor and Factory occurs against the backdrop of fierce competition from giants like Anthropic and OpenAI.

In contrast, the situation in China is markedly different.

ByteDance’s Trae, Alibaba’s Tongyi Lingma, Tencent’s CodeBuddy, Baidu’s Wenxin Kuai Ma, and Huawei’s CodeArts have all entered the fray. Yet, there is a noticeable absence of startups, with no unicorns emerging like Cursor.

Why is it that the vibrant AI programming field in China has not produced a company like Cursor?

In the face of intense competition from global giants, who will emerge victorious first?

U.S. AI Programming Competitive Landscape: Clear Ecosystem

The AI programming market is experiencing unprecedented growth. According to Research and Markets, the global AI programming tools market is projected to reach $29.57 billion by 2025 and $64.68 billion by 2030, with a compound annual growth rate of approximately 17.1%.

This explosive growth is driven by a collective awakening among developers. A 2025 Stack Overflow developer survey indicates that 84% of developers are using AI tools at work, with 51% using them daily.

Currently, the U.S. AI programming landscape exhibits a tiered competitive structure.

First Category: Independent Model Vendors

Anthropic and OpenAI represent the native model path in AI programming. They redefine programming itself through model capabilities rather than relying on IDE ecosystems. Anthropic’s Claude Code is at the forefront of this revolution, achieving an annual revenue of over $1 billion just six months after its launch in early 2025, and reaching a staggering run-rate of $2.5 billion by early 2026.

According to Menlo Ventures, Anthropic holds a 54% market share in programming scenarios.

Claude Code’s advantage lies in its terminal-native architecture, allowing it to directly manipulate file systems, execute bash commands, and maintain state across sessions, functioning like a junior developer rather than a simple autocomplete tool.

Second Category: Startup Unicorns

Cursor, while not a model company, has grown into an AI programming unicorn through exceptional product experience and developer reputation. Founded in 2022, it achieved the fastest growth in B2B SaaS history, scaling from $1 million ARR to $100 million ARR in just 24 months.

By May 2025, Cursor’s annual revenue reached $500 million; by October 2025, it doubled to $1 billion; and by early 2026, it surpassed $2 billion.

Cursor’s success stems from its positioning as an AI-native IDE, offering a complete development environment that integrates multiple top models for a seamless “AI-first code editor” experience. This has made Cursor a favored entry point for developers.

Third Category: Enterprise-Level Startups

Factory represents another path, focusing on a model-agnostic enterprise-level AI programming platform. Founded just three years ago, its clients include financial and cybersecurity giants like Morgan Stanley, EY, and Palo Alto Networks.

Unlike Cursor and Claude Code, Factory’s autonomous AI programming agent operates across various environments, including IDEs, terminals, Slack, and browsers, dynamically switching between different large models to achieve true model neutrality. It specializes in high-margin vertical scenarios that large companies are reluctant to tackle, such as legacy system refactoring and compliance upgrades in financial frameworks.

Its proximity to enterprise needs and organizational governance gives it a competitive edge.

Fourth Category: Platform Giants

Microsoft and Google embody the hegemonic path preferred by platform giants. They do not pursue singular product leadership but leverage the power of ecosystems to dominate competitors. For instance, Microsoft is recognized as the global AI programming stack leader, controlling nearly every aspect of the developer workflow: VS Code (the leading IDE), GitHub (the largest code hosting platform), Azure (one of the top three cloud platforms), and enterprise account systems.

Currently, Microsoft’s GitHub Copilot is the most widely used AI programming tool globally, with over 4.7 million paid users and more than 20 million total users by early 2026, leading in revenue.

Overall, the U.S. market has formed a clear tiered ecosystem, with each player occupying its own niche: independent model vendors like Anthropic provide foundational engines; Cursor demonstrates the product value of AI-native IDEs; Factory captures the gaps in “model neutrality” and “enterprise-level deep customization”; and tech giants like Microsoft leverage distribution channels and cloud ecosystem advantages.

Chinese AI Programming Competitive Landscape: Dominance of Big Tech

In China, a similar market explosion has unfolded, but with a different narrative. According to iiMedia Research, the AI code generation market in China reached 6.5 billion yuan in 2023, expected to grow to 33 billion yuan by 2028, with a compound annual growth rate of 38.4%.

In the Chinese market, the AI programming field is nearly entirely dominated by five tech giants: ByteDance, Alibaba, Tencent, Baidu, and Huawei.

Currently, ByteDance’s Trae is the most aggressive player in the Chinese AI programming market. Launched internationally in January 2025 and in China in March, it amassed over 6 million registered users within a year, with 1.6 million monthly active users across nearly 200 countries and regions. According to IDC, Trae holds a 41.2% market share, ranking first in China.

Trae’s strategy is straightforward: it is completely free. While Cursor charges $20/month and GitHub Copilot charges $10/month, Trae is entirely free for individual developers and integrates multiple top models, including Claude 3.5-Sonnet, GPT-4o, and DeepSeek.

This loss-leader strategy essentially allows ByteDance to use capital to gain market share and data through traffic.

Alibaba has taken a different approach. Its AI programming product, Tongyi Lingma, does not pursue explosive growth among individual users but focuses on the enterprise market. Official data indicates that over 10,000 enterprises, including leading clients like FAW, NIO, and China Pacific Insurance, are using Tongyi Lingma.

Tongyi Lingma’s advantage lies in its deep integration with Alibaba Cloud’s ecosystem, supporting private deployment for enterprises, multi-model configurations, and seamless integration with Alibaba Cloud’s DevOps toolchain.

Other major players like Tencent, Baidu, and Huawei have also launched or integrated their AI coding capabilities, relying on their respective cloud platforms, security compliance advantages, and vast developer ecosystems to create a highly concentrated yet multi-faceted landscape.

The core advantage of these giants lies in their full-stack capabilities. They possess top models, cloud infrastructure, IDE/plugins/agents in all forms, enterprise-level security and compliance capabilities, along with local ecosystem adaptation and free market strategies, effectively meeting nearly all developer needs in a one-stop manner.

In this context, there is virtually no room for startups to survive in the Chinese AI programming field. Even independent large model vendors like Zhipu and Kimi face direct competition from these giants. However, they lack the ecological advantages that the big firms possess.

The few remaining startups, such as AutoCoder and New Words, have completely shifted towards the “vibe coding” direction, avoiding direct competition in enterprise-level programming.

For example, Silicon Heart Technology focuses on domestic adaptation, while Yanchuang Wanzhi bets on Agentic Coding, both areas that the big firms have not prioritized.

In the past year, only a handful of top projects in the domestic AI programming field have managed to secure funding.

Comparing the U.S. and Chinese markets reveals a harsh reality: Chinese tech giants have monopolized all the ecological niches that exist abroad. They do not need to collaborate with startups, as every segment of the industry chain can be self-sufficient.

This explains why a company like Cursor cannot emerge in China—not due to a lack of technical capability but because there is no ecological niche left for startups.

When big firms dominate the entire stack, startups have no space to survive.

Diverging Paths of “Strangulation”

The AI programming landscape in the U.S. and China is heading down two distinctly different paths: the U.S. retains some window for entrepreneurial innovation, while China is dominated by big tech.

However, the ultimate outcome may converge.

In the professional developer market, time is running out for startups.

By 2026, the AI programming field is transitioning from tool assistance to paradigm reconstruction, with Agentic development becoming the main theme. Claude Code and Codex are evolving towards more autonomous cloud agents.

Foreign giants like Microsoft are gradually replacing parts of the IDE workflow with agents, leveraging their powerful models and ecosystems.

This trend towards agent-native pathways suggests that future programming workflows may no longer revolve around IDEs, with true programming occurring through the autonomous execution of agents.

This poses a direct threat to Cursor.

Cursor’s moat lies in its product experience as an AI-native IDE, but if developers no longer rely on IDEs in the future, the foundation of Cursor will be shaken.

This essentially constitutes a dimensionality reduction attack.

Whether in China or abroad, the trend of big firms dominating the market is irreversible.

However, this does not mean there are no gaps for new opportunities.

First, IDEs are unlikely to disappear entirely in the short term; rather, they will coexist with agents. In this scenario, if Cursor can maintain its positioning as an “AI-native IDE” and differentiate itself from Claude Code and Codex, it may have a chance to retain its position in the second tier, but returning to its peak will be challenging.

Second, enterprise clients are increasingly focused on “sovereign AI,” creating opportunities for neutral platforms that support mixed multi-model deployments. Due to concerns about being locked into a single large firm, data security, and loss of bargaining power, AI programming products are likely to evolve towards a direction of “multi-model mixed deployment + intelligent invocation” neutral platforms.

American enterprises have mature procurement processes and vendor evaluation systems and are willing to pay for the best combination of tools.

In China, a combination of closed/open-source models and system/industry integrators’ service capabilities can form a relatively neutral solution.

Lastly, products with extremely high technical barriers and strong customization demands have a 12-24 month window period. The survival space similar to Factory presents potential opportunities for domestic startups: while the market size may be small, the demand for customization is strong.

Big firms are often unwilling to deeply optimize for such demands due to low ROI.

If a startup can establish a sufficiently deep enterprise-level moat during this window period, such as a robust industry-specific knowledge base and strong customer loyalty, it may find a way to survive in the shadow of the big firms.

However, such opportunities are destined to be small and beautiful businesses, rather than the next Cursor.

Conclusion

The ultimate fate of the AI programming field is already inscribed in the genes of the U.S. and Chinese markets.

In the U.S., the celebration of startups may continue for a few more rounds, but in China, from day one, the giants have locked down every inch of entrepreneurial soil with free offerings and closed ecosystems.

Regardless of which side of the Pacific, AI programming will ultimately be dominated by big firms, leaving the remaining players to survive in vertical niches.