Meta and Microsoft are reducing some internal reliance on Anthropic’s Claude as both companies expand their own AI tools, according to a report from The Information, highlighting how quickly the enterprise AI market can shift even among major strategic partners.
The report says Microsoft lowered its projected internal spending on Anthropic technology by more than a third after encouraging employees to use less Claude and spend more time with Microsoft’s own AI systems. It also says the number of Meta employees using Claude Code has fallen sharply as Meta promotes its internal coding tools.
Microsoft’s change is about internal usage, not customer access
The reported reduction should not be read as Microsoft abandoning Anthropic. Anthropic models remain available through Microsoft products and cloud channels, and customer usage is a separate business from Microsoft’s own employee spending.
That distinction matters because large AI vendors increasingly operate as both partners and competitors. Microsoft can distribute Anthropic models to customers while simultaneously trying to reduce its own dependence on an outside model provider.
For enterprise buyers, the arrangement is a reminder that the AI stack is becoming deliberately multi-model. Companies may use one provider for coding, another for search or reasoning, and an internal model for sensitive workflows.
Meta is pushing employees toward Meta-built coding tools
The Information reported that Meta’s internal Claude Code users fell to roughly 30,000 from about 60,000 earlier in the year. Layoffs account for part of that reduction, but the report also points to adoption of Meta’s own coding systems, including MetaCode and Muse Code.
If sustained, the shift would give Meta something strategically valuable: direct internal feedback on its own models and developer tools. Large employee populations can act as proving grounds for systems the company may later offer to outside customers.
The story is bigger than one model vendor
Anthropic has benefited from strong enterprise demand for Claude, particularly in software development. But customers at the scale of Microsoft and Meta have powerful incentives to avoid permanent dependence on any single external model.
Cost is one factor. Data governance, latency, customization and bargaining power also matter. A company that can route workloads between several providers, or shift them to an in-house model, is in a stronger position than one whose workflows are tightly coupled to a single API or coding assistant.
Enterprise AI is becoming a routing problem
The emerging pattern is not necessarily “winner takes all.” It is model routing: organizations select different systems for different tasks and change those choices as price, performance and internal capabilities evolve.
That makes enterprise adoption more volatile than headline partnership announcements can suggest. A vendor can remain an important supplier while losing specific internal workloads to a customer’s homegrown tools.
For Anthropic, the reported changes are worth watching because coding and enterprise usage are central to Claude’s commercial position. For Microsoft and Meta, the moves show why frontier-model capability alone may not determine long-term market power. Distribution, internal tooling and the ability to switch workloads can matter just as much.
Source and verification note
The primary supporting record for this report is October 5 investigation by The Information. Internal company usage and spending figures are attributed to the publication’s reporting, not independently audited company disclosures. The figures concern reported employee usage, not the continued availability of Claude to external customers.
