🔍 Read the full analysis: Thinking Of Leaving Claude? Consider These Switching Costs on ThorstenMeyerAI.com
Get smart everyday buys 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 and Microsoft have reduced some employees’ use of Anthropic’s Claude tools, directing work toward products and models they already have. The reported changes concern internal use, not an end to Claude access or a finding that Claude performs worse. The companies’ existing alternatives make their moves different from a typical enterprise switch, which can involve evaluation, engineering, training and quality costs.
Meta and Microsoft have reportedly reduced some internal use of Anthropic’s Claude tools and directed employees toward alternatives, according to a report by The Information on Oct. 5. The reported moves show how large companies can redirect AI work when costs rise and substitutes are already available, but they do not establish that Claude performed worse or that either company has ended its access to the service.
The Information reported that Meta reduced the number of employees using Claude Code from about 60,000 earlier this year to about 30,000. The report said Meta has been steering staff toward its internal coding tools, MetaCode, with more than 30,000 internal users, and Muse Code, with more than 6,000. Those figures describe reported internal use; the source material does not specify how the user counts were measured or over what exact periods.
Microsoft had reportedly projected more than $1 billion a year in internal spending on Anthropic technology, including Claude Code, Claude models in Copilot and Claude Mythos. The report says Microsoft has since cut that projection by more than a third and is directing employees toward GitHub Copilot and OpenAI models. The material also says Microsoft continues to use Anthropic models for customer-facing Copilot features, and that customer spending on Claude through Microsoft platforms is growing. The reported internal shift is not the same as withdrawing Claude from products or customers.
The reported reasons include rising token costs, tighter spending controls and the availability of in-house or affiliated alternatives. Neither company is reported to have said the change followed a finding that Claude was lower quality. A separate account cited in the source material says some Microsoft team budgets fell from around $100,000 per month to around $10,000; that detail is attributed to a single report and should not be treated as a company-wide figure.
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.
Why Enterprise Switching Has a Cost
The reported changes matter because an AI bill is only one part of the cost of moving work between models. Companies may need to rerun evaluations, revise prompts and tool definitions, rebuild integrations and give employees time to learn a different system. For coding assistants, performance also depends on connections to an editor, repository and team practices, not just the model itself.
There can be less visible costs, too. Moving providers may disrupt cached context used by agent workflows, changing both performance and cache-related charges. If a replacement handles a company’s tasks less well, the effect may appear as more review, rework or errors, rather than a clear failure. Those costs need to be measured alongside token spending to judge whether a switch saves money overall.
Meta and Microsoft reportedly had alternatives ready, giving them a different starting point from most buyers. The source material argues that their scale may make the savings worth the migration work; it estimates that cutting more than a third from a projection above $1 billion would mean more than $300 million in annual savings if the full projection and reduction are comparable. That is an estimate based on reported figures, not a confirmed realized saving. Smaller organizations may face substantial switching costs relative to their potential savings.
As an affiliate, we earn on qualifying purchases.
Internal Tools Versus Customer Products
The report concerns employee use and internal spending, not a general change in customer access to Claude. Microsoft is reported to keep using Anthropic models in customer-facing Copilot features, while Meta and Microsoft have their own reasons to promote alternatives: Meta develops models and coding tools, and Microsoft owns GitHub Copilot and backs OpenAI. These business ties are relevant context, but they do not by themselves explain every decision or establish that the alternatives are better.
The distinction also matters for how the figures should be read. A projected annual budget is not the same as money already spent, and an internal user count is not a measure of product quality or customer demand. The available account describes companies reallocating some work among available tools; it does not report a blanket decision by either company to abandon Anthropic.
For buyers, the practical issue is whether an alternative can handle their own workflows. A model that works well in one company’s engineering environment may not perform the same way elsewhere. The source material recommends maintaining multiple model options, building an evaluation set of representative tasks, and keeping prompts and application logic in a layer the buyer controls. Those are recommendations, not steps the report confirms Meta or Microsoft followed in every case.
enterprise AI model training software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What the Report Does Not Establish
The reported figures do not show how much Anthropic technology the companies currently use across all internal and customer-facing work, or how the reported user counts and budget projections were calculated. The material does not provide a detailed timeline for each change, identify which teams or tasks moved, or say whether the projected spending reduction has translated into realized savings.
It is also unclear how the alternatives compare with Claude on the companies’ actual tasks, and whether productivity, review time or error rates changed after employees were redirected. The report does not establish that either company has ended its relationship with Anthropic. Claims about quality, total cost and the scale of the shift remain unconfirmed in the material provided.
As an affiliate, we earn on qualifying purchases.
Measure the Cost of Each Move
The next useful evidence would be company statements or further reporting clarifying current Anthropic spending, the scope of employee changes and whether budget cuts became actual savings. Comparable information on task quality, review work and productivity would help distinguish a spending reduction from a successful transfer of work.
For other organizations, the immediate next step is not necessarily to switch providers. Buyers can test alternatives on a limited share of real work, maintain evaluations for their own tasks, and track token expense alongside accepted results, review and rework. Whether Meta’s and Microsoft’s reported moves lead to broader changes in their Anthropic use or customer offerings is not yet clear.
As an affiliate, we earn on qualifying purchases.
Key Questions
Have Meta and Microsoft stopped using Claude?
No such complete cutoff is established. The report concerns reductions or changes in internal use. The source material says Microsoft continues to use Anthropic models in customer-facing Copilot features.
Did the companies say Claude performed worse?
The material does not report a quality-based explanation. It identifies costs, spending controls and available alternatives as the reported drivers, and does not establish how the tools compare on specific tasks.
Why might switching AI tools cost more than expected?
Teams may need to repeat evaluations, adapt prompts and integrations, train employees and review output more closely. Changes to cached context and the quality of results can also affect total cost.
Do the reported numbers show how much the companies saved?
No. Microsoft’s figure is described as a reduced spending projection, not confirmed realized savings. The reported user counts and budget details also do not provide a full accounting of total costs.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
