In a strategic shift, Meta and Microsoft are scaling back the use of Anthropic’s Claude AI among their employees, opting instead to promote their proprietary coding solutions. This development was reported on October 5 by The Information, highlighting a focus on optimizing internal expenses and workflows while maintaining customer access to Claude through Microsoft’s offerings.
The Drive to Control AI Costs
This move underscores an increasing need to control AI-related expenditures, as coding assistants become integral to daily software development processes. Despite being key customers of Anthropic, Meta and Microsoft are also its competitors. The adjustments in their AI usage reflect a broader trend of aligning AI spending with product strategies and enhancing control over developer tools.
Microsoft’s Revised AI Spending Strategy
Initially, Microsoft anticipated its internal spending on Anthropic technologies would surpass $1 billion annually. However, this expectation has been significantly adjusted, with projected spending now reduced by over one-third. Employees have been encouraged to minimize their use of Claude and prioritize Microsoft’s own tools, such as GitHub Copilot and OpenAI models.
Within Microsoft’s cloud and AI divisions, monthly AI spending limits have reportedly decreased from $100,000 per employee to about $10,000. These figures represent spending caps rather than actual expenditures. Although this change has frustrated some engineers, it does not indicate the removal of Claude from Microsoft’s customer products, where spending on Anthropic models continues to rise.
Meta’s Transition to In-House AI Tools
Similarly, Meta has seen a decrease in Claude Code users, from around 60,000 earlier this year to approximately 30,000. While layoffs contributed to this decline, the shift towards Meta’s proprietary AI tools, such as MetaCode and Muse Code, is a more significant factor. MetaCode alone reportedly has over 30,000 internal users, while Muse Code exceeds 6,000. Testing of Muse Code with external customers began in August.
Despite the reduced user base, Meta’s investment in Claude Code remains substantial, with over $105 million spent in a 28-day period. This indicates that a drop in users does not necessarily equate to decreased spending.
Security Considerations and Future Outlook
For security teams, the transition to different coding assistants does not eliminate the risks associated with automated file and credential access. Previously, vulnerabilities in Claude Code were addressed by Anthropic, while Microsoft’s tools have faced similar issues, such as the patched RoguePilot vulnerability.
These developments highlight the importance of maintaining stringent permission controls and thoroughly reviewing AI actions. Despite the budget tightening and competitive landscape, Anthropic’s reported $65 billion annualized revenue signals enduring demand for AI solutions.
As AI strategies evolve, both Meta and Microsoft are poised to refine their tools and offerings, ensuring they remain at the forefront of technological advancement.
