Microsoft and Meta are scaling back internal reliance on Anthropic’s Claude and directing employees towards their own artificial intelligence tools, reflecting growing concerns around AI costs, product development priorities and long-term platform strategy.
According to a report by The Information, both companies remain commercial partners of Anthropic and continue providing access to Claude through customer-facing offerings. Internally, however, they are encouraging greater adoption of proprietary AI systems developed within their own ecosystems.
The shift highlights a broader trend emerging across the technology industry, where organisations are balancing access to leading third-party AI models against the strategic advantages of building and refining their own platforms.
Microsoft prioritises internal AI adoption
Microsoft's internal AI strategy appears increasingly focused on driving usage of its own tools while managing rising expenditure on external models.
According to The Information, Microsoft had projected spending at least $1 billion this year on Anthropic’s technology. The report said the company has since reduced that estimate by more than one-third and is encouraging employees to use Microsoft-developed AI products wherever possible.
The move follows Microsoft's earlier rollout of Claude Code, Anthropic’s coding assistant, to thousands of employees beginning in December.
While the tool reportedly gained traction among engineering teams, Microsoft later shifted its focus towards GitHub Copilot CLI, its own AI-powered coding solution.
The report noted that Microsoft's Experiences + Devices division, which oversees products including Windows, Microsoft 365 and Surface, was preparing to discontinue most Claude Code licences by the end of June.
Beyond cost considerations, Microsoft reportedly wants engineers to spend more time working with its own AI products and contributing feedback to improve them.
Despite the internal transition, Microsoft continues to support Anthropic's models across several business offerings. Claude remains available through Microsoft Foundry and is also used in selected workflows across Microsoft 365 and Copilot products, according to the report.
Meta expands adoption of proprietary coding platforms
Meta is pursuing a similar strategy by increasing reliance on internally developed AI coding tools.
According to The Information, the number of Meta employees using Claude Code has fallen to around 30,000, down from approximately 60,000 earlier this year.
Part of the decline coincides with workforce reductions affecting roughly 10% of Meta's employees. However, the report said the company is also actively promoting coding platforms built around its own AI models.
Key developments include:
- MetaCode has surpassed 30,000 internal users
- Muse Code, an internal coding assistant being tested as a Claude Code alternative, has attracted more than 6,000 users
- Meta continues investing in proprietary AI infrastructure to support internal software development
The company’s growing emphasis on in-house AI tools aligns with broader efforts to strengthen its own AI ecosystem while reducing dependence on external providers.
AI economics reshape enterprise decisions
The internal changes at Microsoft and Meta come as organisations gain a clearer understanding of the financial implications of large-scale AI deployment.
According to an earlier report by Axios, an AI consultant described a case where a client received a monthly Claude-related bill of approximately $500 million after failing to establish sufficient usage controls on employee licences.
The example underscores a challenge facing enterprises worldwide, including those in India, as AI adoption accelerates across software development, operations and knowledge work functions.
For technology leaders, the debate is increasingly shifting from AI access to AI economics, governance and long-term value creation.
Strategic balancing act continues
The latest developments do not signal a breakdown in relationships between Microsoft, Meta and Anthropic. Instead, they reflect a more nuanced approach in which technology companies continue offering third-party AI models to customers while simultaneously strengthening adoption of their own platforms internally.
As AI becomes embedded deeper into enterprise workflows, companies are likely to evaluate not only model performance but also cost efficiency, ecosystem control and product differentiation. For large technology firms investing billions in proprietary AI capabilities, internal adoption may become as strategically important as external revenue generation.