🔍 Read the full analysis: Estimating AI Migration Costs: What The Claude Shift Could Involve on ThorstenMeyerAI.com
Get hardware and tech essentials 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 are steering some employees toward in-house or other AI coding tools, amid spending controls and cost concerns. The report does not say either company has ended Claude access or found it inferior; for most businesses, copying the shift could entail evaluation, engineering, integration and productivity costs.
Meta and Microsoft are steering some employees away from Anthropic’s Claude tools and toward alternatives they own or already use, according to an Oct. 5 report by The Information. The reported moves concern internal employee use, not a withdrawal of Claude from customer products, and point to a practical question for other businesses: what would switching models actually cost?
The Information reported that Meta reduced the number of employees using Claude Code from about 60,000 earlier this year to about 30,000. It said Meta is directing staff toward its own coding products: MetaCode, with more than 30,000 internal users, and Muse Code, with more than 6,000. These figures are reported usage counts; the source material does not specify a comparison period for the alternative-tool user totals.
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 company has since cut that projection by more than a third, according to the report, and is steering employees toward GitHub Copilot and OpenAI models. The report also describes tighter token budgets. One account puts some monthly team budgets at about $10,000, down from around $100,000, but that figure is attributed to a single report.
The reported explanations include rising token costs, spending controls and companies’ preference for tools they own or back. The source material does not report either company saying Claude performed worse. It also says Microsoft continues to use Anthropic models in customer-facing Copilot features and that customer spending on Claude through Microsoft’s platforms is growing. The internal changes therefore do not, on the information available, amount to a full departure from Anthropic.
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 moves matter because a lower model bill does not automatically mean lower operating costs. Companies changing providers may need to rerun evaluations to check that existing workflows still meet performance and safety requirements. If a team has not built a representative test set, it may have no reliable way to compare whether a new model works as well on its actual tasks.
There can also be engineering work to adapt prompts, tool definitions and agent systems, plus time spent rebuilding integrations with editors, repositories and team processes. Employees may be less productive while learning a different tool. For agent workloads, changing providers can reset cached context or change cache pricing. And if a new model needs more human review or produces more rework, those costs may outweigh savings visible in a token bill.
Meta and Microsoft have existing alternatives, and the scale of their reported spending makes potential savings substantial. A cut of more than a third from a projection exceeding $1 billion would imply a reduction of more than roughly $330 million if the entire projected spend and comparison held. That is a calculation from the reported figures, not a confirmed realized saving. Smaller companies may face a different equation: the source material offers no measured switching-cost data, so it cannot establish whether a company spending $20,000 monthly would save money by moving.
As an affiliate, we earn on qualifying purchases.
Why These Buyers Had Alternatives
The reported changes involve companies with a strong reason and the resources to use substitutes. Meta builds its own models and coding tools; Microsoft owns GitHub Copilot and is OpenAI’s largest backer, according to the source material. Both can shift internal work toward products they develop or support. That makes their decisions different from those of a company that depends on one external provider and has no tested replacement.
The distinction between internal deployment and customer access also matters. A company can reduce employee use of a supplier’s product while continuing to sell or use that supplier’s technology in services for customers. The report, as summarized in the source material, describes that pattern at Microsoft. It does not establish that either company has ended its broader relationship with Anthropic.
The broader operational issue is how easily a buyer can reroute work when prices, budgets, product terms or availability change. Maintaining more than one production-ready model can reduce dependence, but it also takes resources to build and test. The reported actions show what two large buyers did with alternatives already in place; they do not show that the same approach is economical for every organization.
As an affiliate, we earn on qualifying purchases.
What the Report Does Not Establish
The available source material does not provide company statements confirming the reported usage counts, spending projection or budget changes, nor does it detail how those figures were calculated. The reported reductions concern internal use and projected spending; they do not establish the exact amount spent, the savings realized, or the share of employee work moved to each alternative.
It is also unclear how the tools compare on the companies’ specific tasks, whether employees retain access to Claude, and what transition costs Meta or Microsoft incurred. The reported reasons center on cost and in-house options, but that does not rule out other considerations. No independent evaluation results or productivity measurements are supplied, so the report cannot settle which tool performs better or whether switching improved overall value.
AI model migration cost calculator
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
How Businesses Can Measure a Move
For other buyers, the next step is to measure a proposed change against real work rather than infer a result from the headline numbers. Teams can keep a second model family running on a limited share of tasks, maintain a set of representative evaluations, and record pass criteria before making a broad shift. That creates evidence about quality and integration before a change becomes urgent.
Organizations can also keep prompts, tool definitions and business logic in a layer they control, and track cost per accepted result alongside token spending, review time and rework. These steps do not guarantee that switching will pay off; they make the costs and trade-offs easier to see. The next confirmed developments would be updated company figures or direct statements clarifying the scope, reasons and results of the reported changes.
As an affiliate, we earn on qualifying purchases.
Key Questions
Have Meta and Microsoft stopped using Claude?
The report describes reduced or redirected internal employee use, not a complete end to Claude access. The source material says Microsoft continues to use Anthropic models in customer-facing Copilot features.
Why are the companies reportedly shifting work?
The reported reasons include token costs, tighter spending controls and existing alternatives. The source material does not report either company saying Claude performed worse.
What costs can a company incur when changing models?
Potential costs include retesting workflows, adapting prompts and integrations, employee retraining, cache changes, and added review or rework if results differ. The actual cost depends on the company’s systems and tasks.
Does the report show that switching saves money?
No. It reports changes to internal use and a spending projection, but supplies no complete accounting of realized savings or transition costs. Each buyer would need to measure total cost and task quality for its own use.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
