🔍 Read the full analysis: Meta And Microsoft’s Claude Decision: How To Assess 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 as they promote internal or affiliated alternatives. The reported shifts concern internal use, not a broad withdrawal of Claude from customer-facing products. They show how access to ready-made substitutes can make switching easier, while most companies still face costs in testing, integration, training and quality review.
Meta and Microsoft have reportedly reduced some employees’ use of Anthropic’s Claude tools while directing staff toward alternatives they own or already use, according to an Oct. 5 report by The Information. The reported changes concern internal use, not a general end to Claude access, and highlight how the cost of switching AI tools depends on whether a company already has a workable substitute.
The Information reported that Meta’s Claude Code use fell from about 60,000 employees earlier this year to about 30,000. The company has been promoting its own coding tools: MetaCode, which the source material says has more than 30,000 internal users, and Muse Code, with more than 6,000. The figures describe reported internal usage; the report does not establish that every employee who stopped using Claude moved to one of those products.
At Microsoft, the report said the company had projected more than $1 billion a year in internal spending on Anthropic technology, including Claude Code, Claude models used in Copilot and Claude Mythos. It reportedly reduced that projection by more than a third and has directed employees toward GitHub Copilot and OpenAI models. A separate account cited in the source material says some team budgets fell from about $100,000 a month to about $10,000; that detail is attributed to a single report.
The reported reasons include rising token costs, tighter spending controls and the availability of in-house or affiliated tools. Neither company is reported to have said the move was prompted by Claude’s quality. The source material also says Microsoft continues to spend on Anthropic models for customer-facing Copilot features and that customer spending on Claude through Microsoft platforms is growing. Those claims do not establish the full scale or terms of either company’s current Anthropic use.
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 Existing AI Alternatives Matter
The reported changes matter because Meta and Microsoft could redirect work to tools already deployed, rather than starting a replacement effort from scratch. Both companies have large engineering operations and credible alternatives: Meta has internal coding products, while Microsoft has GitHub Copilot and access to OpenAI models. That makes their experience a limited guide for companies without comparable resources.
For other buyers, the relevant question is not simply which model has the lowest price. Switching may require teams to rerun evaluations, adapt prompts and tool integrations, retrain users and check whether results remain accurate on real tasks. A cheaper model can carry higher costs if it creates more review or rework. Token spending is only one part of the cost; accepted output, employee time and operational reliability also matter.
The source material argues that companies can reduce dependence on one provider by maintaining more than one model option and keeping application logic in their own systems. That is a proposed approach, not proof that every business should run multiple models. Its value depends on whether the added integration and oversight costs are lower than the risks of relying on a single supplier.
As an affiliate, we earn on qualifying purchases.
The Reported Shift at Each Company
The development is about employees’ internal access and company spending plans, as described in reporting by The Information and summarized in the supplied source material. It should not be read as evidence that either company has ended its commercial relationship with Anthropic or that customers have lost access to Claude.
The companies have reasons to favor their own or affiliated products beyond a direct model comparison. Meta develops AI models and coding tools. Microsoft owns GitHub Copilot and is a major backer of OpenAI. That competitive position is relevant when interpreting their internal choices: a company may have commercial and operational reasons to promote a product it controls. The available information does not show how much each factor contributed to the reported decisions.
The broader switching-cost question is practical. Prompts, agent tools and integrations can be tuned to a specific model, and teams may need time to learn a replacement. Changes can also affect how cached context is used and priced. These costs vary by workflow, so the reported savings at very large companies cannot be assumed to apply to a smaller buyer.
As an affiliate, we earn on qualifying purchases.
What the Usage Figures Do Not Show
The supplied material does not include direct statements from Meta, Microsoft or Anthropic confirming the reported numbers or explaining the decisions. The exact timing, measurement methods and current usage levels are not clear. Nor does the account establish whether the reported employee counts refer to active users, accounts with access or another measure.
It is also unclear how much work moved to alternative tools, how much was discontinued, and whether the changes produced measurable savings without reducing output quality. The reported spending projection at Microsoft is not necessarily the same as actual spending. The source material provides no comparable performance results across Claude and the alternatives, and does not quantify switching costs for a typical company.
The source material refers to another analysis by SemiAnalysis about AI subscription limits changing by account and list-price reductions affecting subscription value, but does not provide enough detail here to verify its findings or connect them directly to the decisions at Meta and Microsoft. Those points should not be treated as evidence about either company’s specific contracts.
AI model switching cost calculator
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What Buyers Should Measure Next
The next useful evidence would be direct company responses, updated usage or spending figures, and performance data showing how the replacement tools fare on the work employees actually do. Until then, the reported cuts show a change in internal allocation, not a confirmed verdict on Claude’s quality or its long-term place in either company’s products.
For businesses reviewing their own AI contracts, the practical next step is to measure representative tasks before making a switch. Teams can compare models using the same evaluation set, track the time needed for human review and rework, and include integration and training costs alongside token prices. Maintaining a limited second-provider option may make future changes easier, but it also creates extra work and expense that should be measured.
Companies should also check whether prompts, tool definitions and business logic are portable, and whether contractual or data-handling requirements change across providers. A switch is only a saving if total costs and results improve over the relevant period. The available reporting does not say whether Meta or Microsoft has published that kind of comparison.
AI development tools for businesses
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
Did Meta and Microsoft stop using Claude?
No full withdrawal is established by the supplied reporting. It describes reduced or redirected internal use; the material says Microsoft continues to use Anthropic models in customer-facing Copilot features.
Why are the companies reportedly steering employees to other tools?
The reported factors are token costs, spending controls and available alternatives. The supplied material does not report either company saying Claude performed worse.
Does this mean Claude is lower quality than its alternatives?
No comparative performance results are provided. The reported decisions alone do not establish that Claude is better or worse on the companies’ tasks.
Why might switching AI models cost more than expected?
Teams may need to retest workflows, revise prompts and integrations, train employees and account for changed review or rework. The cost depends on the company’s systems and tasks; the report does not quantify it for typical businesses.
What should a company measure before changing providers?
Compare models on representative tasks and track total cost, including usage, integration, human review and rework. A test set with clear pass criteria can help show whether a lower token bill also delivers acceptable results.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
