Microsoft is preparing to unveil its own AI coding model at its Build developer conference, marking a strategic pivot away from relying on external AI partners like Anthropic and OpenAI. The move reflects growing pressure in the AI coding space, where competition is intensifying and companies are reassessing the true return on their massive AI investments.

Why Is Microsoft Building Its Own Coding Model?

According to reports, Microsoft is working on multiple AI models for different tasks, including coding, image creation, speech, reasoning, and transcription. The coding-focused model is expected to become a major part of GitHub Copilot, Microsoft's AI-powered assistant for developers. Currently, GitHub Copilot uses technology from several AI companies, including OpenAI, Anthropic, and Google. But newer coding assistants, particularly Anthropic's Claude Code, have started attracting strong developer interest, prompting Microsoft to accelerate its own alternatives.

The timing is significant. Microsoft has been internally testing Anthropic's Claude Code across thousands of employees over the past several months, allowing even non-engineering staff to experiment with AI-assisted software development. However, the company is now reportedly planning to reduce internal usage of Claude Code by the end of June and encourage teams to move toward its own Copilot-based command line tools instead.

What's Driving the Broader Shift Away From Third-Party AI Tools?

Microsoft's move reflects a larger industry pattern. Corporate leaders across the United States are increasingly questioning whether their massive spending on AI is producing meaningful returns. According to Ali Ansari, chief executive of AI training company Micro1, companies may be overestimating what AI can currently achieve. Ansari noted that "the reality of AI right now is that it only works for coding," suggesting that many businesses assume AI tools work equally well across all departments, but that expectation often does not match reality.

Ansari

Cost considerations are playing a significant role in these decisions. A Financial Times report indicated that reducing third-party AI software usage could help Microsoft lower operating expenses as the company enters its next financial year. The situation extends beyond Microsoft. Amazon shut down an internal AI leaderboard called Kirorank after some employees used AI tools to generate unnecessary activity to improve their rankings, which increased computing costs. One business reportedly spent about $500 million in a single month after failing to place limits on employee use of Anthropic's Claude AI tools.

How Are Companies Managing AI Costs and Efficiency?

• Internal Tool Standardization: Microsoft is transitioning employees from third-party tools like Claude Code to its own Copilot CLI tool, centralizing AI development within its ecosystem and reducing external licensing costs.
• Usage Monitoring and Limits: Companies like Amazon are implementing stricter controls on employee AI tool usage to prevent unnecessary consumption and associated computing expenses.
• Strategic Acquisitions: Microsoft has been exploring possible acquisitions of AI startups to strengthen its internal expertise and reduce dependence on external AI partners for advanced capabilities.

Financial factors are also influencing workforce decisions. Major companies including Meta and Amazon have cut thousands of jobs in 2026, with most attributed to AI. Anuj Kapur, CEO of CloudBees, suggested that workforce cuts may simply be "the only lever they can pull" to offset their AI bills, indicating that companies are struggling to balance AI investments with profitability.

Anuj Kapur, CEO of CloudBees

What New Capabilities Are Emerging in Claude Code?

While Microsoft is building alternatives, Anthropic continues advancing Claude Code with new features. On May 28, Anthropic launched dynamic workflows in Claude Code, which allows the system to construct and execute task-specific structures at runtime rather than following a fixed sequence. Instead of handling problems linearly, Claude decomposes tasks, assigns subtasks to multiple parallel subagents, and synthesizes results after internal comparison and verification.

Dynamic workflows are currently available in research preview across the Claude Code CLI, Desktop, and VS Code extension for Max, Team, and Enterprise plans. The feature is also accessible via the Claude API, Amazon Bedrock, Vertex AI, and Microsoft Foundry. The company explicitly noted that dynamic workflows can consume significantly more tokens than a typical Claude Code session, advising users to test the tool on limited-scope tasks to understand costs.

The feature is designed for complex tasks that typically take hours or days. By running tens to hundreds of subagents simultaneously in one session, Claude can explore more possibilities at once while using adversarial verification, where independent agents attempt to refute findings before surfacing results. According to Anthropic, dynamic workflows support migrations across hundreds of thousands of lines of code, using existing test suites as the completion standard.

What Does This Mean for the AI Coding Market?

Anthropic's rapid growth has added fresh pressure to the competition. The AI startup recently announced a new funding round worth $65 billion, which pushed its valuation to nearly $965 billion, placing it ahead of several major AI rivals. Anthropic also revealed that its annualized revenue run rate has climbed to around $47 billion, demonstrating how quickly demand for AI coding tools is growing across the industry.

Microsoft's strategy signals confidence from investors. The company's shares moved higher after reports about the upcoming AI announcements surfaced. If Microsoft successfully positions itself as a stronger independent AI player rather than relying mainly on outside partners, it could reshape the competitive landscape. The Build conference, scheduled for next week, will be a critical moment to see whether Microsoft can deliver on its ambitions to compete directly with Anthropic and OpenAI in the AI coding space.

" }