🔍 Read the full analysis: What Does It Cost To Switch From Claude? Meta And Microsoft Pulled Back on ThorstenMeyerAI.com
Get movie-night favorites delivered free — and shop member deals
- Fast, free delivery on millions of items
- Access to Prime Big Deal Days deals on October 6–7
- Prime Video, Amazon Music and more included
TL;DR
The Information reported on Oct. 5 that Meta reduced employee use of Claude Code and Microsoft cut a projected annual internal Anthropic spend by more than a third. The reported moves concern internal use, not an end to customer access, and the stated drivers were cost controls and available alternatives rather than Claude quality.
Meta and Microsoft are pulling some internal work away from Anthropic’s Claude tools, according to an Oct. 5 report by The Information, citing cost controls and the availability of alternatives. The reported shifts involve the companies’ own employees, not a decision to end Claude access for customers, and the report does not say either company found Claude performed worse.
Meta reportedly reduced the number of employees using Claude Code from about 60,000 earlier this year to about 30,000. The company has steered staff toward its own coding products: MetaCode, which the source material says has passed 30,000 internal users, and Muse Code, which has passed 6,000. Those figures describe internal users; the report does not establish how much work has shifted or compare the tools’ performance.
Microsoft had reportedly projected spending more than $1 billion a year internally 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 directed employees toward GitHub Copilot and OpenAI models. The source material also reports that Microsoft’s internal token budgets tightened, with one account describing some monthly team budgets falling from about $100,000 to about $10,000. That budget detail comes from a single report and should not be treated as a company-wide figure.
The reported changes do not amount to a withdrawal of Claude from Microsoft products. The source material says Microsoft continues to spend on Anthropic models for customer-facing Copilot features and that customer spending on Claude through Microsoft platforms is growing. The companies’ reported reasons for changing internal usage include token costs, spending limits and investment in tools they own or back; neither is reported as saying Claude’s quality prompted the moves.
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.
The Cost of Switching Models
The report matters less as a verdict on a single AI model than as a view of how large buyers manage vendor costs and dependence. Meta and Microsoft have alternatives ready: Meta has its own coding tools, while Microsoft owns GitHub Copilot and backs OpenAI. Their ability to redirect work reflects existing products and engineering capacity, not a simple switch available to every customer.
Changing providers can carry costs that do not appear in a model’s per-token price. Companies may need to rerun evaluations, adjust prompts and tool integrations, retrain staff, and monitor whether quality changes increase review or rework. For coding tools, workflows are often closely tied to editors, repositories and team conventions. Moving them can mean rebuilding those connections and accepting a period of lower productivity.
There can also be billing changes when teams move agent workloads. The source material points to cached context as a potentially important part of those costs, because a change of provider may reset caches or alter cache pricing. Whether a move saves money overall depends on the task, actual usage and the cost of engineering and review—not just the headline model price. The report provides no public accounting of Meta’s or Microsoft’s complete switching costs.
As an affiliate, we earn on qualifying purchases.
Why These Buyers Had Alternatives
Meta and Microsoft are not typical customers with a single AI supplier. Meta develops its own models and coding products. Microsoft sells GitHub Copilot and has a major relationship with OpenAI, while also using Anthropic models in some products. Those overlapping roles give both companies options when deciding which tools their employees use.
The distinction between internal adoption and customer-facing services is central. An employer can redirect its staff to an in-house or partner product while continuing to offer another supplier’s models to customers. The reported internal reductions therefore do not show that customers have lost access to Claude, or that Microsoft has stopped using Anthropic technology across its business.
The source material frames maintaining access to multiple models as a way to reduce reliance on one supplier’s prices, policies or availability. That is a proposed purchasing strategy, not a finding that every organization should move its work. The reported actions at Meta and Microsoft show that large companies with established substitutes can redirect some usage; they do not establish what the same decision would cost a smaller buyer.
“Meta reduced internal Claude Code use from about 60,000 employees earlier this year to about 30,000, according to the report.”
— The Information, in its Oct. 5 report
As an affiliate, we earn on qualifying purchases.
What the Report Does Not Show
The available reporting does not give a complete breakdown of how much Claude usage either company has eliminated, how quickly changes are being implemented, or how much money the moves will save after engineering and productivity costs. The reported Microsoft spending figure is a projection, not a confirmed annual bill, and the source material does not specify the period or scope behind every internal usage count.
It is also unclear how the alternative tools compare with Claude on the companies’ actual tasks, or whether employee productivity, review time or error rates changed after the shifts. Neither company is reported to have identified model quality as the reason for the move. The cost rationale and the interpretation that switching is easier for firms with substitutes should not be mistaken for a public performance comparison.
Customer arrangements are another separate issue. The source material says Microsoft continues using Anthropic models in customer-facing Copilot features and that customer spending on Claude through Microsoft platforms is growing, but it does not provide figures or clarify how those arrangements may change. Meta’s reported employee shift likewise does not establish a change to third-party access to Claude.
As an affiliate, we earn on qualifying purchases.
How Buyers Can Track the Shift
The next useful evidence will be any further company disclosures or reporting that clarifies actual spending, employee usage and customer-facing deployment. Until then, the figures should be read as reported internal changes, not as a complete account of either company’s Anthropic business or as proof that one model is better.
For other organizations weighing a change, the practical test is to measure the full cost of a representative workflow before moving it. That means comparing results on real tasks, including evaluation and integration work, staff review, rework and usage charges. Maintaining a second provider on some real workloads can make a later move less disruptive, but it also has costs; the source material does not quantify what that approach would save.
The broader question is whether the savings from an alternative exceed the expense and risk of changing a production workflow. Meta and Microsoft can draw on tools and teams they already have. For companies without those substitutes, the same reported price pressure may not make a switch worthwhile. Further details on savings and performance remain the key points to watch.
As an affiliate, we earn on qualifying purchases.
Key Questions
Did Meta and Microsoft stop using Claude?
No such broad exit is reported. The reported changes concern internal employee use and projected internal spending. The source material says Microsoft continues to use Anthropic models for customer-facing Copilot features.
Why are the companies reportedly reducing internal use?
The reported reasons are rising token costs, tighter spending controls and available alternatives. Neither company is reported as saying that Claude performed worse.
How much did Microsoft cut its projected spending?
The Information reportedly said Microsoft cut its projection by more than a third. The projection had been more than $1 billion a year in internal Anthropic technology spending; a final annual spending amount is not provided.
Will switching models save other companies money?
Not necessarily. Savings depend on usage prices as well as the costs of retesting, integration, staff adjustment, review and rework. The report does not establish the total switching costs or savings for either company.
What should a company measure before switching?
It should test both providers on representative work and track the full cost per accepted result, including usage, engineering, review and rework. The source material also recommends keeping workflows and evaluations usable across more than one model, but does not provide a universal cost estimate for doing so.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
