Meta and Microsoft are reportedly pulling back on how much they lean on Anthropic’s Claude models inside their own engineering organizations, redirecting employees toward homegrown coding tools instead. The shift, first detailed by The Information and summarized by Yahoo Finance, pymnts.com, CyberSecurityNews, and The Times of India on October 5 and 6, 2026, does not amount to either company ending its commercial relationship with Anthropic. It is, instead, a story about internal cost control, tool preference, and what happens when two of the biggest Claude customers start asking whether they even need Claude for everyday coding work.
The reporting lands at a sensitive moment for Anthropic, which is reportedly pushing toward a market debut (see our coverage of what Anthropic’s IPO prospectus actually says) while two of its largest enterprise accounts quietly trim their internal usage. Here is what the reports actually say, what they stop short of saying, and why the distinction matters for anyone trying to read the AI coding tools market right now.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
What the Reports Actually Say
According to The Information’s reporting, as relayed by Yahoo Finance and CyberSecurityNews, both Meta and Microsoft have reduced the number of employees actively using Claude for internal coding work over the course of 2026. Neither company has reportedly canceled a contract with Anthropic or ended Claude’s availability to customers. The pullback described in the coverage is specifically about internal employee tooling, not about Claude’s presence on Microsoft Foundry, Azure, or Meta’s customer-facing products.
That distinction is easy to miss in headlines, but it is the central fact of this story. Microsoft reportedly continues to make Claude available through Microsoft Foundry and within parts of its Copilot and Microsoft 365 product family, according to the same reporting chain. What changed is employee-facing policy: both companies are reportedly steering engineers toward internal alternatives rather than Anthropic’s coding assistant for day-to-day development work.
Meta’s Claude Code Usage Reportedly Cut in Half
The most concrete figure in the reporting involves Meta’s internal headcount using Claude Code. The Information’s reporting, cited by The Times of India and CyberSecurityNews, says the number of Meta employees actively using Claude Code fell from roughly 60,000 earlier in 2026 to about 30,000 more recently, a decline of roughly half. The reports do not pin down an exact date for either measurement, and they do not separate how much of the drop came from a deliberate policy shift versus Meta’s broader 2026 workforce changes, including the layoffs covered in our report on Zuckerberg’s $145 billion AI spending bet.
What is clear from the reporting is the scale of money involved even as usage fell. One report, cited by CyberSecurityNews, says Meta spent more than $105 million on Claude Code during a single 28-day period this year. That figure is described as a period-specific expenditure rather than a disclosed annual contract value, so it should not be read as Meta’s total yearly Anthropic bill. It does, however, show that even a halved user base at Meta represents a meaningful chunk of Anthropic’s enterprise coding revenue.
In place of Claude, Meta is reportedly directing engineers toward internal tools identified in the coverage as MetaCode and Muse Code, alongside Meta’s own Muse Spark model line, which has reportedly displaced some Claude-based workflows. The reporting does not specify model versions, parameter counts, or exactly how Meta divides coding work between these tools, so any claim beyond “Meta is pushing internal alternatives” would be reading more into the story than the sourcing supports.
Microsoft Cuts Its Anthropic Spending Projection by More Than a Third
Microsoft’s side of the story centers on a budget number rather than a headcount. The Information’s reporting, as summarized by Yahoo Finance and The Times of India, says Microsoft had projected spending at least $1 billion internally on Anthropic technology during 2026. That projection was later cut by more than a third, according to the same sourcing. It is worth stressing that this $1 billion figure is described as an internal spending forecast, not a disclosed contract value and not necessarily a binding minimum commitment under Microsoft’s Anthropic partnership.
Inside Microsoft’s Cloud and AI organization specifically, the reporting describes a sharper tightening: monthly AI spending limits per employee reportedly dropped from roughly $100,000 to about $10,000 in most cases. The reports do not clarify whether this was a formal company-wide policy, a departmental ceiling, or an individual discretionary allowance, so it is best read as a sign of budget discipline rather than a precise, audited company-wide rule.
Microsoft is reportedly asking employees to shift toward its own tools, particularly GitHub Copilot and GitHub Copilot CLI, which pulls engineers toward OpenAI’s models inside Microsoft’s own product rather than Anthropic’s. For context on how Microsoft’s own agent tooling has evolved alongside this shift, see our guide to setting up Microsoft Agent Framework.
Claude Still Has a Seat at Microsoft’s Table
None of this means Microsoft is walking away from Anthropic as a platform partner. The companies announced a strategic partnership in November 2025 that made Claude models available through Microsoft Foundry and Azure, continuing Claude’s presence in parts of the Copilot family. The October reporting describes a reduction in employee-directed internal usage, not a termination of that distribution deal. Microsoft reportedly continues to offer Claude to its own customers even as it nudges its own staff toward internal and OpenAI-backed tools for day-to-day coding.
Reported Spending and Usage Shifts at a Glance
| Company | Metric | Earlier 2026 | Recently Reported | Reported Change |
|---|---|---|---|---|
| Meta | Employees actively using Claude Code | ~60,000 | ~30,000 | ~50% decline |
| Meta | Claude Code spend in a single 28-day period | n/a disclosed | >$105 million | Not a disclosed annual figure |
| Microsoft | Projected 2026 internal Anthropic spending | At least $1 billion (forecast) | Cut by >33% | More than one-third reduction |
| Microsoft | Cloud & AI org monthly AI spend limit per employee | ~$100,000 | ~$10,000 (most cases) | ~90% reduction |
Source: The Information, as reported by Yahoo Finance, CyberSecurityNews, and The Times of India, October 2026. Figures are reported estimates from internal sources and should be treated as directional rather than audited totals.
Why Two of Anthropic’s Biggest Customers Are Pulling Back
The reporting does not include a named executive explaining the rationale in detail, but the pattern across both companies points to the same underlying pressure: cost. Running frontier coding assistants across tens of thousands of engineers is expensive, and both Meta and Microsoft have spent 2026 publicly tightening AI infrastructure budgets elsewhere, even while increasing total AI capital spending. Building or expanding internal alternatives, like Meta’s MetaCode and Muse Code or Microsoft’s push toward GitHub Copilot CLI, lets each company keep engineering productivity gains in-house while cutting a line item that was reportedly consuming over $100 million in a single month at Meta alone.
There is also a strategic dimension. Meta has invested heavily in its own model lineup, and routing engineers back to internal tools reinforces the case that Meta’s own models are good enough for internal use, not just external products. Microsoft’s push toward GitHub Copilot does something similar for OpenAI’s standing inside Microsoft’s own walls, at a moment when Microsoft has been diversifying its model relationships rather than relying on any single outside lab.
Historical Context: How Anthropic Became Big Tech’s AI Supplier
Anthropic’s rise as an enterprise coding and infrastructure supplier has been fast. The Microsoft-Anthropic strategic partnership announced in November 2025 made Claude available through Microsoft Foundry and Azure, giving Anthropic a direct channel into Microsoft’s enterprise customer base at a time when Microsoft was also deepening its OpenAI relationship. Meta, for its part, built out large-scale internal adoption of Claude Code across its engineering organization over the following months, reportedly reaching tens of thousands of active users by sometime in 2026.
That rapid internal adoption inside two of the world’s largest tech employers became a kind of implicit validation for Anthropic, useful marketing even if neither company talked about it publicly at the time. The October 2026 reporting complicates that narrative. It does not erase Anthropic’s commercial relationships with either company, but it does show that heavy internal usage by Big Tech engineering teams is not necessarily permanent, especially once a company’s own tools mature enough to compete.
Competitive Comparison: Claude Code vs. Internal Tools vs. GitHub Copilot
The reporting frames this as a three-way competitive dynamic inside each company rather than a clean head-to-head benchmark, since none of the sources disclose comparative performance data between Claude Code, MetaCode, Muse Code, or GitHub Copilot CLI. What the reports do establish is the direction of internal preference at each company.
| Tool | Backing Company / Lab | Reported Role in This Story | Reported Availability |
|---|---|---|---|
| Claude Code | Anthropic | Usage reportedly cut at Meta and Microsoft | Still available via Microsoft Foundry and Azure |
| MetaCode / Muse Code | Meta | Reportedly gaining internal share at Meta | Internal Meta tool |
| Muse Spark model line | Meta | Reportedly displacing some Claude-based workflows | Internal/Meta AI products |
| GitHub Copilot / Copilot CLI | Microsoft (OpenAI-backed) | Reportedly promoted internally at Microsoft | Commercially available to all GitHub customers |
For readers comparing coding assistants directly, our breakdown of Claude Opus 5.5 API pricing and setup is a useful reference point for what Anthropic is still charging external developers, even as two of its largest internal accounts scale back.
Market Impact: What This Means for Anthropic
The available reporting does not include a quantified stock-price reaction for Meta or Microsoft tied specifically to this story, and neither does it include a named analyst’s published assessment of the shift. That absence is itself informative: this looks like an internal-operations story rather than a market-moving disclosure, at least so far. Both stocks continue to trade on broader AI capital spending narratives rather than on internal tool-usage figures at this scale.
For Anthropic, the practical risk is narrower than the headlines suggest but still real. Losing internal seats at Meta and Microsoft does not touch Anthropic’s external enterprise contracts or its Claude availability on Microsoft’s own cloud platform. It does, however, remove two of the most visible proof points that Claude Code can scale inside a giant, cost-sensitive engineering organization, right as Anthropic is reportedly preparing for a market debut. Anthropic has also been widening its outside footprint: on October 8, 2026, it announced a $150 million, three-year commitment to the Genesis Mission and said it would make Claude available to participating federal agencies. Our earlier coverage detailed how Anthropic’s own IPO materials flagged existential AI risk as a disclosed factor, and enterprise customer concentration is exactly the kind of detail that IPO investors tend to probe.
Reports about Anthropic’s IPO push remain unverified in key respects. A separate report cited an alleged prospectus figure of roughly $4.6 billion in revenue against a $42 billion loss, with valuation chatter reaching as high as $2 trillion, but none of this has been confirmed through a regulatory filing in the reporting reviewed for this story. Readers should treat those IPO numbers as speculative until Anthropic or a verified filing confirms them.
Key Dates in the Anthropic-Big Tech Relationship
| Date | Reported Event | Source |
|---|---|---|
| November 2025 | Microsoft and Anthropic announce strategic partnership bringing Claude to Microsoft Foundry and Azure | Industry reporting |
| First half of 2026 | Meta’s internal Claude Code usage reportedly reaches roughly 60,000 employees | The Information |
| Throughout 2026 | Microsoft projects at least $1 billion in internal Anthropic spending for the year | The Information, via Yahoo Finance |
| October 5-6, 2026 | Reports surface that Meta’s Claude Code usage has fallen to roughly 30,000 employees and Microsoft cut its spending projection by more than a third | The Information, Yahoo Finance, pymnts.com, CyberSecurityNews, The Times of India |
| October 2026 | Unverified reports of an Anthropic IPO prospectus circulate, citing disputed revenue, loss, and valuation figures | Unconfirmed reports |
What This Means for Enterprise AI Buyers
For other enterprises watching from the outside, the lesson is less about Anthropic specifically and more about how any AI coding vendor relationship can shift once internal alternatives mature. Companies that adopted Claude Code early, in the same wave that pushed Meta past 60,000 internal users, are now the ones with enough internal data and engineering capacity to build credible substitutes. Smaller enterprises without the resources to build a MetaCode or a Muse Code equivalent face a different calculation, and for most of them, external tools like Claude Code or GitHub Copilot remain the only realistic option.
That split helps explain why Anthropic’s commercial relationships with Meta and Microsoft, including Claude’s continued presence on Microsoft Foundry, have not been affected by this internal pullback. Microsoft still has an incentive to sell Claude access to its own cloud customers even while trimming its own employees’ use of the tool. The economics of being a platform are different from the economics of being an internal user, and this story is really about the latter.
How This Fits the Broader AI Agent and Safety Backdrop
The timing also lands amid broader scrutiny of how AI coding and agent tools behave inside large organizations. Regulators have been probing AI agent behavior more aggressively this year, including the FTC’s review of agent conduct at major labs covered in our report on the FTC’s three-company AI agent probe. Meanwhile, OpenAI has been pushing its own enterprise agent products, detailed in our coverage of the Dots agent launch, and announced GPT-6 in ChatGPT on October 7, 2026, adding another competitive angle to how Microsoft frames its internal tooling choices. None of these developments are confirmed to be directly connected to the Meta and Microsoft usage cuts, but they form the competitive backdrop against which this decision is playing out.
Predictions: Where This Goes Next
- Expect Meta and Microsoft to keep publicizing internal tool wins (MetaCode, Muse Code, GitHub Copilot CLI) as proof their own AI investments are paying off, independent of what they still buy from outside labs.
- Anthropic’s external enterprise and Azure/Foundry distribution deals are unlikely to be affected in the near term, since the reported cuts are specifically about internal employee usage, not platform contracts.
- Watch for Anthropic to emphasize external, non-Big-Tech enterprise customers in any IPO materials, since customer concentration among a handful of giant, cost-sensitive internal buyers is a flag investors will probe.
- Other large tech employers with mature internal AI coding efforts, not just Meta and Microsoft, will likely face similar internal build-vs-buy pressure on their own Claude or Copilot seat counts over the next two to three quarters.
- If Anthropic’s IPO process moves forward, expect its prospectus to address enterprise customer concentration directly, given how closely investors will compare it to the existential-risk disclosures already reported in earlier filings coverage.
What to Watch For Next
The clearest open question is whether Meta’s and Microsoft’s internal pullback is a temporary budget-tightening exercise or the start of a longer structural shift away from third-party coding assistants at the largest AI labs’ biggest customers. Neither company has issued a public statement beyond what has been reported through internal sources, and Anthropic has not publicly addressed the figures directly in the material reviewed for this story. Readers should expect follow-up reporting from The Information and other outlets in the coming weeks as both companies finalize 2027 AI tooling budgets.
Frequently Asked Questions
Did Meta and Microsoft cancel their contracts with Anthropic?
No. The reporting describes a reduction in internal employee usage of Claude, not a cancellation of either company’s commercial or platform relationship with Anthropic. Claude reportedly remains available through Microsoft Foundry and Azure.
How much did Meta’s Claude Code usage actually drop?
The Information’s reporting, cited by The Times of India and CyberSecurityNews, says Meta’s active Claude Code users fell from roughly 60,000 earlier in 2026 to about 30,000 more recently, a decline of roughly half.
What tools is Meta using instead of Claude Code?
Reports identify MetaCode and Muse Code as internal tools gaining ground, alongside Meta’s Muse Spark model line, though the reporting does not detail exactly how work is divided between them.
How much is Microsoft cutting its Anthropic spending?
Microsoft reportedly projected spending at least $1 billion internally on Anthropic technology in 2026, and that projection has since been cut by more than a third, according to The Information’s reporting as summarized by Yahoo Finance.
Is this story about Anthropic losing customers?
Not directly. It is about Meta and Microsoft reducing internal employee use of Claude for coding tasks. Both companies reportedly continue to make Claude available to their own external customers through existing platform deals.
Does this affect Anthropic’s rumored IPO plans?
There is no confirmed direct link in the reporting. However, reduced internal usage at two of Anthropic’s largest corporate customers could become a talking point for investors evaluating customer concentration risk if and when Anthropic formally files for a public offering.
Is GitHub Copilot replacing Claude at Microsoft?
Reports say Microsoft is directing employees toward GitHub Copilot and GitHub Copilot CLI as internal alternatives, but Claude remains available within parts of Microsoft’s Copilot and Microsoft 365 product family for select use cases.
Where did this story originate?
The Information first reported the internal usage and spending figures. Yahoo Finance, pymnts.com, CyberSecurityNews, and The Times of India subsequently summarized and corroborated the core numbers in their own coverage on October 5 and 6, 2026.
