🔍 Read the full analysis: What To Evaluate Before Your Company Switches From Claude on ThorstenMeyerAI.com
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
The Information reported on Oct. 5 that Meta and Microsoft have reduced or plan to reduce some internal use of Anthropic’s Claude tools, directing employees toward products they own or support. The reported moves reflect cost controls and available substitutes, not a stated finding that Claude performs worse. For other companies, switching requires evaluating engineering, integration, productivity and quality costs alongside model prices.
Meta and Microsoft are steering some employees toward alternatives to Anthropic’s Claude tools, according to a report by The Information on Oct. 5. The reported changes show how large buyers may redirect AI work when costs rise and internal or partner alternatives are available, but the report does not say either company concluded Claude was inferior.
Meta reportedly reduced the number of employees using Claude Code from about 60,000 earlier this year to about 30,000. The report says the company has encouraged staff to use its own coding products, MetaCode, which has more than 30,000 internal users, and Muse Code, which has more than 6,000. These figures describe reported internal use, not customer adoption or the end of Meta’s access to Claude.
Microsoft had reportedly projected annual internal spending of more than $1 billion on Anthropic technology, including Claude Code, Claude models used in Copilot and Claude Mythos. The Information reported that Microsoft cut that projection by more than a third and was steering employees toward GitHub Copilot and OpenAI models. The source material also says Microsoft continues to use Anthropic models for customer-facing Copilot features and that customer spending on Claude through Microsoft platforms is reported to be growing.
The reported reasons for the internal changes include token costs, tighter spending controls and the availability of tools the companies own or support. Neither company is reported to have said Claude performed worse. A separate detail about some Microsoft team budgets falling from roughly $100,000 to roughly $10,000 a month comes from a single account in the reporting and should not be treated as a company-wide policy.
Meta and Microsoft pulled back from Claude. Here’s what switching actually costs.
The Information reports both companies steering their own employees away from Claude. Read as a verdict on Claude, it misleads. Read as a demonstration of switching — and who can afford it — it’s the most useful enterprise-AI signal this month.
Staff steered to GitHub Copilot and OpenAI models; stricter token budgets. One unconfirmed report: some team budgets ~$100k → ~$10k/month.
Microsoft reportedly still spends heavily on Claude for customer-facing Copilot — and that spending is reported to be growing.
Reported drivers: rising token costs and owned alternatives. Neither company is reported to have called Claude worse.
Meta builds coding tools; Microsoft owns Copilot and backs OpenAI. This is ordinary vertical integration.
Keep a second vendor live on real work.
A few hundred tasks with pass criteria.
Logic, prompts, tools in your layer.
Tokens are the cheap half.
Know what you’d rebuild.
On the evidence reported, Meta and Microsoft didn’t reject Claude. They brought spending in-house where they could and kept buying where they couldn’t — Microsoft remains a large Anthropic customer for the products it sells. The signal is the mechanism: the most sophisticated buyers treat models as interchangeable suppliers behind a layer they control.Meta could halve its Claude usage because it had built somewhere else to go. Build somewhere else to go.
Switching Costs Go Beyond Model Prices
The reported moves matter to enterprise buyers because a lower model price does not, by itself, establish that a switch will save money. Companies also need to account for evaluation work, engineering changes, integration and employee time. These costs can offset savings, particularly when a replacement has not already been deployed.
Teams may need to rerun tests for each important workflow, revise prompts and tool definitions, and rebuild connections to editors, repositories or internal systems. Employees can also lose time while adapting to a different tool. For coding agents, changes in context caching and cache pricing may affect costs, while a quality difference can appear as extra review, rework or mistakes rather than an obvious model failure.
Meta and Microsoft reportedly have alternatives ready to use, giving them a different starting point from a company with one provider integrated across its workflows. The report’s figures do not show whether the companies’ savings outweighed all migration costs. They do, however, illustrate why buyers should compare the cost per accepted result, including human review, rather than token spending alone.
As an affiliate, we earn on qualifying purchases.
Why These Buyers Have Alternatives
Meta and Microsoft are not neutral observers of the AI market. Meta develops its own models and coding tools. Microsoft owns GitHub Copilot and is a major backer of OpenAI. That gives both companies products they can direct employees to use when managing internal spending. Their reported decisions therefore do not establish that a typical customer can switch as easily.
The distinction between internal use and customer products is also important. The reported reductions concern employees’ use of Claude tools; Microsoft is still reported to use Anthropic models in customer-facing Copilot features. A company can reduce its own consumption while continuing to offer a supplier’s technology to customers. The report is evidence of internal routing and cost management, not a blanket withdrawal of Claude from either company.
For organizations without a ready substitute, a change may require building and maintaining a second integration before it can deliver negotiating leverage or operational flexibility. A practical preparation is to keep more than one model family available for real work, maintain a representative evaluation set, and keep prompts and business logic in an application layer the company controls. Those steps reduce dependence, but they also require ongoing engineering and testing.
“Meta reportedly cut the number of employees using Claude Code from about 60,000 earlier this year to about 30,000.”
— The Information, in a report published Oct. 5
As an affiliate, we earn on qualifying purchases.
What the Report Does Not Establish
The available reporting does not establish the exact timing of every change, the amount ultimately spent by either company, or the full costs of moving employees to replacement tools. It also does not provide a controlled comparison of Claude with MetaCode, Muse Code, GitHub Copilot or OpenAI models across the companies’ internal tasks.
It remains unclear how much of the reported reduction reflects lasting changes in employee behavior, formal policy, or revised forecasts. The reported Microsoft team-budget figures come from a single account, and the source material does not establish that they apply broadly. Nor does the report quantify effects on employee output, review burden or error rates. Those limits make it difficult to infer whether the shifts represent a lasting technical judgment or primarily a response to costs and available alternatives.
As an affiliate, we earn on qualifying purchases.
Measure Before Changing Providers
Companies considering a move should first define the workflows they need to preserve and test candidate models on representative tasks. A useful evaluation records not only model output but also pass rates, correction time, review effort and total cost. Teams should run the alternative on actual work before ending their existing setup, so they can identify integration gaps and differences in output quality.
Buyers should also track provider prices, usage limits and caching terms as they change, and decide how much operational redundancy is worth maintaining. The next meaningful evidence will be whether Meta and Microsoft report sustained changes in internal use, spending or productivity; the Oct. 5 report does not settle those questions. For other companies, the immediate decision is not necessarily whether to leave Claude, but whether they can make a measured switch if cost, performance or availability changes.
As an affiliate, we earn on qualifying purchases.
Key Questions
Are Meta and Microsoft ending their use of Claude?
The report describes changes to internal employee use, not a complete end to access. Microsoft is also reported to continue using Anthropic models in some customer-facing Copilot features.
Did the companies say Claude performs worse?
No such claim is reported in the source material. The stated drivers are cost controls and available alternatives; the report does not provide a direct performance comparison.
Why might another company find switching harder?
Meta and Microsoft have alternatives ready to use. Other organizations may need to rebuild integrations, revise prompts, test workflows and account for employee learning time and quality risks.
What should a company measure before switching?
Test representative work on both systems and compare the full cost: model use, engineering, review, rework and accepted results. A maintained evaluation set helps distinguish actual performance from assumptions.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
