Meta and Microsoft Slash Claude Usage as Token Costs Soar, Pivot to In-House Models

Meta cut internal Claude Code users by 50% to 30,000 while Microsoft reduced Claude spending by over a third, both replacing Anthropic's model with proprietary AI coding tools to control token costs that reached billions annually, even as Anthropic reports 7x ARR growth to $65 billion.

Machine Heart
Machine Heart
Machine Heart
Meta and Microsoft Slash Claude Usage as Token Costs Soar, Pivot to In-House Models

Meta Halves Internal Claude Code Adoption

According to The Information , Meta's active Claude Code users dropped from 60,000 to 30,000 — a 50% decline. Part of the drop stems from spring layoffs that reduced headcount by roughly 10% from 78,000 employees. However, the report emphasizes a deliberate shift: Meta is actively steering developers toward its own AI coding tools.

Meta has built a full internal AI coding stack. Muse Code , powered by Meta's proprietary models and positioned as a direct Claude Code competitor, began external testing in August and already counts over 6,000 internal users. A second internal-only tool, MetaCode , has surpassed 30,000 users. As Claude Code usage fell, Meta's homegrown tools rapidly filled the gap.

Cost Pressure Behind the Shift

A June internal memo revealed Meta's internal AI tool spending was on track to reach billions of dollars annually. In a recent 28‑day period, Meta spent $105 million on Claude Code alone, making it the company's most popular third‑party coding assistant. At that run rate, the annual bill would be unsustainable.

Microsoft Cuts $1 Billion Annualized Claude Bill by Over a Third

By May, Microsoft's internal use of Anthropic technology — including Claude Code, Copilot's Claude model, and security‑scanning models — was projected to exceed $1 billion per year. The company responded by slashing that budget by more than 33%, lowering the per‑employee monthly token allowance from $100,000 to $10,000 .

Microsoft is also reducing direct Claude Code subscriptions and pushing developers to access Claude through GitHub Copilot . Copilot's automatic routing mode preferentially sends requests to OpenAI models, which — given Microsoft's commercial partnership with OpenAI — lowers Microsoft's effective model costs. In an August memo, Microsoft developer‑tools chief Jay Parikh instructed every division to manage its own token budget and continuously optimize model usage.

Anthropic Says Revenue Impact Is Minimal

Anthropic has told investors that its annual recurring revenue (ARR) has grown roughly 7× since January , from about $9 billion to $65 billion . The company now has approximately 6,000 enterprise customers spending over $100,000 per year (a 4× increase) and more than 100 large accounts spending above $10 million annually.

Regarding the Meta and Microsoft concentration risk: in 2025, when Anthropic's full‑year revenue was $4.6 billion, the two giants together contributed about $1.1 billion (25%). At the current $65 billion run rate, that same $1.1 billion represents less than 2% of revenue.

Meta Claude Code user decline chart
Meta Claude Code user decline chart
Meta internal AI coding tool adoption
Meta internal AI coding tool adoption
Microsoft Claude spend reduction
Microsoft Claude spend reduction
Anthropic ARR growth and customer metrics
Anthropic ARR growth and customer metrics
Original Source

Signed-in readers can open the original source through BestHub's protected redirect.

Sign in to view source
Republication Notice

This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactadmin@besthub.devand we will review it promptly.

MicrosoftClaudetoken economicsAnthropicMetaAI coding assistantscompetitive strategyenterprise AI adoption
Machine Heart
Written by

Machine Heart

Professional AI media and industry service platform

0 followers
Reader feedback

How this landed with the community

Sign in to like

Rate this article

Was this worth your time?

Sign in to rate
Discussion

0 Comments

Thoughtful readers leave field notes, pushback, and hard-won operational detail here.