Microsoft has cut its projected internal spending on Anthropic technology by more than a third from an earlier estimate of at least $1 billion, Investing.com reported on 5 October 2026. The reduction concerns Microsoft’s own use of Claude as it directs employees towards its AI tools.
Key points
- Microsoft’s internal Anthropic spending projection has fallen by more than a third from at least $1 billion.
- Monthly AI spending limits in its cloud and AI division reportedly fell from roughly $100,000 to about $10,000 per employee in most cases.
- Meta’s Claude Code users fell from roughly 60,000 earlier in the year to about 30,000.
- Customer spending on Anthropic models through Microsoft’s enterprise platforms continues to grow.
Microsoft lowers Claude budgets for employees
Earlier this year, Microsoft projected at least $1 billion in internal spending on Anthropic technology for 2026. Its lower projection followed instructions to staff to use less Claude and rely more on Microsoft’s offerings, according to the 5 October report. The figure is a spending estimate for Microsoft’s operations, rather than a measure of what customers pay to use Anthropic models on its platforms.
Within Microsoft’s cloud and AI division, monthly AI spending limits fell from roughly $100,000 to about $10,000 per employee in most cases, Cyber Security News reported on 6 October. Those amounts are limits on permitted spending, rather than amounts spent by every employee. Some engineers were frustrated by the reduced scope to test models.
Microsoft is steering staff towards tools including GitHub Copilot and OpenAI models. AI Affairs previously reported that Microsoft combined its Copilot products to focus on corporate customers. The employee spending controls concern a separate decision about which tools Microsoft’s own staff use.
Customer spending on Anthropic models through Microsoft’s enterprise platforms continues to grow, Investing.com reported. Microsoft’s reduced internal projection and that reported growth describe different sources of demand for Claude: purchases for employee use and spending by customers through Microsoft’s platforms.
Meta shifts Claude Code users to its tools
At Meta, the number of employees using Claude Code has fallen from roughly 60,000 earlier in 2026 to about 30,000, Investing.com reported. The report attributed part of the decline to spring layoffs affecting about 10% of Meta’s 78,000-person workforce, while identifying the company’s push towards its own AI tools as the main factor.
Meta’s alternatives include MetaCode, an internal coding tool with more than 30,000 users, and Muse Code, which has over 6,000 employee users. Both use Meta models. Meta began testing Muse Code with external users in August, the 5 October report said.
Meta spent more than $105 million on Claude Code during one 28-day period this year, Cyber Security News reported. The $105 million is expenditure over the reported 28 days, rather than an annual spending figure.
Meta’s reported user counts cover different tools: about 30,000 employees using Claude Code, more than 30,000 using MetaCode and over 6,000 using Muse Code. Its shift towards those alternatives is taking place alongside continued spending on Anthropic’s coding assistant.
Anthropic retains Microsoft platform demand
The changes at Microsoft and Meta concern tools used by their employees. Both companies remain Anthropic customers, and Microsoft continues to offer customers access to Claude through its products. At Microsoft, the reported growth in customer spending contrasts with the reduction in its own projected purchases.
AI Affairs previously reported its $100 million commitment to train 10,000 enterprise AI engineers. Cyber Security News put Anthropic’s reported annualised revenue pace at $65 billion.
