Microsoft and Meta are scaling back Claude Code: What we know about the strategy shift

TVS News AI, editorially supervised, presents Microsoft and Meta are scaling back Claude Code: What we know about the strategy shift.

Around 60,000 Meta employees were regularly working with Anthropic's coding agent Claude Code at the beginning of this year. Since then, the number has shrunk to about 30,000 according to internal reports – nearly cut in half. The changes at Microsoft are similarly drastic: The company already announced in May 2026 that it would cancel numerous Claude Code licenses in the coming months. Internal AI spending forecasts have also plummeted according to t3n – instead of the originally targeted around one billion US dollars, only about a third less is now budgeted for Anthropic's AI.

The t3n report (author: Marvin Fuhrmann, October 17, 2026, 4:35 PM) is based on an article by The Information and paints a picture of a strategic shift at two of the world's largest tech employers: Instead of continuing to roll out AI tools from Anthropic, the internal use of Claude Code is being deliberately scaled back.

Why Microsoft already pulled the lever in May

That Microsoft is restricting Claude Code usage is not new – the company had already initiated the move in May 2026. At that time, developers were informed that numerous licenses for Anthropic's coding helper would be canceled in the coming months. At the same time, internal forecasts for AI spending also dropped drastically – an indication that Microsoft is putting the brakes on Anthropic overall.

At Meta, user numbers dropped from 60,000 to 30,000

At Meta, the strategic shift is reflected in a concrete number: While around 60,000 employees were integrated into the daily workflow with Claude Code at the beginning of 2026, now only about 30,000 remain. Part of this decline is explained by Meta's ongoing layoffs – but that is not the primary driver, according to internal sources quoted by t3n.

Copilot, Muse Spark, Muse Code: What's being used instead

Both companies want their developers to increasingly rely on their own AI tools. At Microsoft, Copilot remains the central alternative. Meta is currently building several in-house solutions: The relatively new AI model Muse Spark is designed to support employees in their daily work, the coding agent Muse Code is currently being tested internally – and could become a competitor to Claude Code in the medium term. For internal use, Meta has also long relied on an agent called Meta Code.

What this means for you – and for Anthropic

For external developers and Microsoft customers, nothing changes for the time being: The restrictions only affect the internal policies of the two companies. Microsoft customers can continue to use Anthropic models via Microsoft services without being subject to new limits – the report emphasizes this explicitly.

It remains to be seen how the move will affect Anthropic itself in the medium to long term. Microsoft is one of the largest investors in the AI company – a corporation that is simultaneously scaling back licenses for its coding agents. Precisely this constellation of investment and simultaneous restriction makes the coming months a litmus test for Anthropic.

Source: t3n