Microsoft is reportedly cancelling most Claude Code access for engineers in its Experiences and Devices division by June 30, 2026, shifting teams working on Windows, Microsoft 365, Outlook, Teams, and Surface toward GitHub Copilot CLI as the company tries to rein in internal AI coding costs. The decision is more than a procurement tweak. It is a rare glimpse into what happens when the world’s most aggressive AI software company runs into the same metered-billing problem now facing every large engineering organization. Microsoft is not backing away from AI coding; it is trying to domesticate it.
June 30 is doing a lot of work here. Microsoft’s fiscal year ends that day, and a hard cutoff for a fast-growing, usage-based tool is exactly the sort of move finance teams make when enthusiasm has outrun budget design. The company can describe the shift as standardization, platform alignment, or security hygiene, and all of those arguments may be true. But the calendar makes the cost-control motive hard to ignore.
Claude Code reportedly arrived inside Microsoft’s Experiences and Devices group in December 2025. That is not a fringe corner of the company. It is the orbit of Windows, Microsoft 365, Teams, Outlook, and Surface — the products that define Microsoft for hundreds of millions of users and most enterprise administrators.
The reported pattern is familiar: engineers tried the tool, many liked it, usage spread, and the bill became harder to treat as experimental overhead. The surprise is not that Microsoft would prefer its employees to use GitHub Copilot CLI. The surprise is that a company with Microsoft’s AI balance sheet still found it necessary to yank a popular rival coding assistant on a fiscal deadline.
That makes this episode useful beyond Redmond gossip. It tells IT leaders that the AI coding market is entering its second phase. The first phase was about whether these tools work. The second is about whether they can be governed without turning every developer workstation into a variable-cost endpoint.
That does not mean Claude Code was failing. In fact, the reported urgency of the migration suggests the opposite. Tools rarely become budget emergencies unless people are actually using them.
For developers, the distinction between “approved tool” and “preferred tool” matters. Coding assistants are not like expense-report software, where a company can switch providers and most employees grumble for a week. They sit inside the loop of thinking, editing, testing, debugging, and committing. Once an engineer builds muscle memory around a command-line assistant, the assistant becomes part of the workflow rather than a detachable accessory.
This is where Microsoft’s decision becomes more complicated than a home-team preference. The company is asking engineers to change the instrument they use while the performance is already underway. That may be necessary, but it is not free.
GitHub Copilot CLI gives Microsoft a tighter surface for governance, telemetry, integration, and product feedback. It also lets Microsoft shape the tool around its own repositories, security expectations, and internal practices. But every migration carries a tax, and in this case the tax is paid by engineers who must re-map daily habits at speed.
That is the trap of token economics. The better the assistant is, the more it gets used. The more it gets used, the more context users provide. The more context they provide, the more expensive each session can become.
In normal SaaS, heavy adoption is usually a vendor success story and a customer productivity story. In usage-metered AI, heavy adoption can be both of those things and a budget fire. The marginal cost does not disappear just because the tool feels magical.
That is why the Microsoft move will resonate with enterprise IT teams. The hard part is not proving that AI can generate code, summarize logs, draft tests, or explain unfamiliar APIs. The hard part is predicting the cost of doing those things every hour across thousands of engineers who are incentivized to move quickly.
Microsoft is effectively saying that an AI assistant must fit inside an operating model before it can become standard engineering infrastructure. That is a more mature position than the marketing suggests. It is also less glamorous.
There is nothing wrong with standardization. Enterprises standardize because fragmented tooling creates duplicated spend, inconsistent security posture, scattered logs, weaker audit trails, and support headaches. In regulated environments, “the engineers like it” is not enough to justify uncontrolled adoption of an external AI coding agent.
But the migration order is the tell. If the issue were only philosophical alignment, Microsoft could have wound access down gradually, team by team, after Copilot CLI reached parity in the eyes of its internal users. A June 30 cutoff suggests a sharper constraint.
That constraint may be cost, competitive positioning, governance, or all three. The useful reading is not that Microsoft panicked. It is that Microsoft recognized an internal contradiction: it wants AI coding everywhere, but not at any price, and not through a tool it does not control.
For WindowsForum readers, that distinction matters. The same company telling customers to adopt AI across the software lifecycle is now modeling the less exciting half of adoption: limiting, measuring, and redirecting usage before it overwhelms the budget.
That makes the switch a serious engineering-management test. AI coding assistants are not merely code generators. They are increasingly used to draft tests, inspect failures, propose migrations, explain legacy behavior, and generate scaffolding around unfamiliar codebases. Remove one assistant and replace it with another, and the output may look similar in a demo while feeling very different in practice.
Developers notice small differences quickly. One tool may be better at navigating a monorepo. Another may be better at shell commands. One may produce safer incremental patches, while another may be more aggressive about rewriting. In a consumer app, these differences are annoyances. In a Windows or Office codebase, they affect velocity, review load, and confidence.
Microsoft’s gamble is that Copilot CLI can absorb the workflow without creating enough friction to matter. That is plausible, especially because GitHub is deeply embedded in modern development practices and Microsoft can tune the product internally. But the short-term hit is real even if the long-term platform logic is sound.
This is also an internal dogfooding moment. If GitHub Copilot CLI can satisfy Microsoft’s own high-pressure engineering groups, that becomes a powerful proof point. If it cannot, Microsoft will learn the problem before its customers do.
That makes a Claude Code pullback awkward. Microsoft can say it supports an ecosystem while still preferring its own tool internally. Both things can be true. But optics matter when developers are the audience, and developers tend to respect tools that win on merit more than tools that win by policy.
Claude Code’s reported popularity inside Microsoft complicates the story. If engineers adopted it because it solved problems well, then the company is not just swapping suppliers. It is overriding revealed preference.
Still, Microsoft’s position is defensible. Internal engineering is not a neutral marketplace. Companies standardize on compilers, security scanners, cloud platforms, package registries, and source-control policies all the time. AI coding agents are simply becoming important enough to be pulled into the same governance machinery.
The question is whether Microsoft can make the mandated tool feel like the best tool. If Copilot CLI becomes excellent because Microsoft’s own engineers push it hard, this episode may be remembered as a necessary consolidation. If not, it will feed the suspicion that AI strategy is being shaped more by balance sheets than by builders.
A coding assistant that costs a predictable per-user fee can be budgeted like collaboration software. A coding agent that burns through usage charges based on context windows, retries, tool calls, and background tasks must be managed more like cloud infrastructure. That means quotas, alerts, chargebacks, procurement reviews, and uncomfortable conversations about whether a generated patch was worth the run.
Cloud computing already taught this lesson once. The first wave of cloud adoption sold flexibility and speed. The second wave introduced FinOps because flexibility without discipline became surprise spending. AI coding is now entering its own FinOps era.
The irony is that the same practices engineers use to get better AI results can make costs harder to forecast. Supplying more files improves context. Asking for multiple approaches improves quality. Running agents through test failures improves usefulness. But each step can add cost, especially when multiplied across large teams.
Microsoft’s reported cutoff is a warning that AI coding tools cannot be evaluated only by demo quality. They must be evaluated by total operating cost, auditability, data handling, integration, and whether the company can explain the bill to finance without resorting to vibes.
That is not anti-developer bureaucracy. It is what turns a promising assistant into infrastructure. If an AI tool can write code that ships into Windows or Microsoft 365, then access and observability are not optional.
The governance layer must also distinguish between different kinds of usage. Generating boilerplate, writing unit tests, explaining stack traces, and drafting documentation may deliver strong returns at modest risk. Letting an agent roam a large codebase, invoke tools, rewrite major components, and iterate through failures is a different cost and risk profile.
The mistake many companies will make is treating all prompts as equal. They are not. A short explanation request and a multi-step agentic refactor may both appear as “AI coding usage,” but they belong in different budget and review categories.
Microsoft is better positioned than most to build that governance into its own stack. GitHub, Azure, Microsoft Entra, Defender, Purview, and the Microsoft 365 administrative universe all give it pieces of the control plane customers will eventually demand. Claude Code may be an excellent tool, but excellence alone does not solve the enterprise control problem.
Code volume is not the same as productivity. A tool can generate many lines and still increase review burden. It can accelerate simple tasks while introducing subtle defects. It can help senior engineers move faster while giving junior engineers plausible-looking code they do not fully understand.
That does not mean the tools are overhyped. It means the measurement problem is harder than the marketing. The right question is not “How much code did AI write?” but “Which work became faster, safer, cheaper, or more reliable because AI participated?”
Microsoft’s internal shift should be read in that context. If Claude Code created enough value to become popular, the company now has to preserve that value while changing the cost envelope. That is the real test of enterprise AI: not whether a tool can impress a developer, but whether it can survive procurement, governance, security, and operational measurement.
A company can tolerate fuzzy ROI during a pilot. It cannot tolerate fuzzy ROI at platform scale.
Command-line coding assistants live or die by trust. Developers need them to understand project structure, avoid destructive changes, explain commands before running them, and recover gracefully when they are wrong. A slick integration is not enough if the assistant creates more cleanup than leverage.
The opportunity for GitHub is substantial. Microsoft’s internal engineering base can become an unusually demanding proving ground. Windows and Microsoft 365 engineers can surface edge cases that most product teams would never encounter. If Copilot CLI improves under that pressure, customers benefit.
But there is risk in captive adoption. When a tool wins because employees are told to use it, product teams can mistake compliance for love. Microsoft will need to watch not just usage numbers, but friction signals: abandoned sessions, manual rewrites, review churn, escalations, and quiet workarounds.
Developers are pragmatic. If Copilot CLI works well, the resentment fades. If it does not, the policy becomes a drag on morale and productivity.
Losing broad access inside Microsoft would sting, but it does not necessarily weaken Claude Code’s reputation. In some circles, it may strengthen it. Developers often interpret corporate bans on popular tools as evidence that the tools were too effective, too expensive, or too threatening to internal priorities.
That is not always fair. Security, compliance, and cost can justify restrictions even on excellent products. But developer folklore is powerful, and “Microsoft engineers preferred Claude until finance intervened” is exactly the kind of story that travels.
For Anthropic and other AI coding vendors, the lesson is clear. Capability is not enough. Enterprise buyers will demand predictable pricing, administrative controls, repository-aware policies, and audit trails that make large-scale deployment less frightening.
The next battle will not be only about model quality. It will be about packaging model quality in a way CFOs can approve and CISOs can defend.
This is the shape of the moment:
Microsoft’s AI Coding Lesson Arrives on the Last Day of the Fiscal Year
June 30 is doing a lot of work here. Microsoft’s fiscal year ends that day, and a hard cutoff for a fast-growing, usage-based tool is exactly the sort of move finance teams make when enthusiasm has outrun budget design. The company can describe the shift as standardization, platform alignment, or security hygiene, and all of those arguments may be true. But the calendar makes the cost-control motive hard to ignore.Claude Code reportedly arrived inside Microsoft’s Experiences and Devices group in December 2025. That is not a fringe corner of the company. It is the orbit of Windows, Microsoft 365, Teams, Outlook, and Surface — the products that define Microsoft for hundreds of millions of users and most enterprise administrators.
The reported pattern is familiar: engineers tried the tool, many liked it, usage spread, and the bill became harder to treat as experimental overhead. The surprise is not that Microsoft would prefer its employees to use GitHub Copilot CLI. The surprise is that a company with Microsoft’s AI balance sheet still found it necessary to yank a popular rival coding assistant on a fiscal deadline.
That makes this episode useful beyond Redmond gossip. It tells IT leaders that the AI coding market is entering its second phase. The first phase was about whether these tools work. The second is about whether they can be governed without turning every developer workstation into a variable-cost endpoint.
The Tool Microsoft Owns Was Always Going to Get the Benefit of the Doubt
There is an obvious corporate logic to pushing engineers toward GitHub Copilot CLI. Microsoft owns GitHub. GitHub Copilot is central to Microsoft’s developer strategy. If Microsoft’s own engineers are building critical products with a competing assistant, that creates a credibility problem as much as a cost problem.That does not mean Claude Code was failing. In fact, the reported urgency of the migration suggests the opposite. Tools rarely become budget emergencies unless people are actually using them.
For developers, the distinction between “approved tool” and “preferred tool” matters. Coding assistants are not like expense-report software, where a company can switch providers and most employees grumble for a week. They sit inside the loop of thinking, editing, testing, debugging, and committing. Once an engineer builds muscle memory around a command-line assistant, the assistant becomes part of the workflow rather than a detachable accessory.
This is where Microsoft’s decision becomes more complicated than a home-team preference. The company is asking engineers to change the instrument they use while the performance is already underway. That may be necessary, but it is not free.
GitHub Copilot CLI gives Microsoft a tighter surface for governance, telemetry, integration, and product feedback. It also lets Microsoft shape the tool around its own repositories, security expectations, and internal practices. But every migration carries a tax, and in this case the tax is paid by engineers who must re-map daily habits at speed.
Token Economics Punish Success
Traditional enterprise software trained CIOs to think in seats, renewals, and negotiated discounts. AI coding assistants break that mental model. A seat license might open the door, but the expensive part often happens after adoption, when engineers begin feeding tools larger contexts, running agents through repeated attempts, and asking for increasingly complex changes.That is the trap of token economics. The better the assistant is, the more it gets used. The more it gets used, the more context users provide. The more context they provide, the more expensive each session can become.
In normal SaaS, heavy adoption is usually a vendor success story and a customer productivity story. In usage-metered AI, heavy adoption can be both of those things and a budget fire. The marginal cost does not disappear just because the tool feels magical.
That is why the Microsoft move will resonate with enterprise IT teams. The hard part is not proving that AI can generate code, summarize logs, draft tests, or explain unfamiliar APIs. The hard part is predicting the cost of doing those things every hour across thousands of engineers who are incentivized to move quickly.
Microsoft is effectively saying that an AI assistant must fit inside an operating model before it can become standard engineering infrastructure. That is a more mature position than the marketing suggests. It is also less glamorous.
The Internal Memo Is Less Important Than the Migration Order
The reported internal framing leans on toolchain unification and a move toward Microsoft-controlled AI development workflows. That is the clean version of the story. It is also the version every large company tells when it wants to turn an emergency into a strategy.There is nothing wrong with standardization. Enterprises standardize because fragmented tooling creates duplicated spend, inconsistent security posture, scattered logs, weaker audit trails, and support headaches. In regulated environments, “the engineers like it” is not enough to justify uncontrolled adoption of an external AI coding agent.
But the migration order is the tell. If the issue were only philosophical alignment, Microsoft could have wound access down gradually, team by team, after Copilot CLI reached parity in the eyes of its internal users. A June 30 cutoff suggests a sharper constraint.
That constraint may be cost, competitive positioning, governance, or all three. The useful reading is not that Microsoft panicked. It is that Microsoft recognized an internal contradiction: it wants AI coding everywhere, but not at any price, and not through a tool it does not control.
For WindowsForum readers, that distinction matters. The same company telling customers to adopt AI across the software lifecycle is now modeling the less exciting half of adoption: limiting, measuring, and redirecting usage before it overwhelms the budget.
Windows and Microsoft 365 Engineers Are the Real Test Case
The affected teams reportedly sit near products that cannot tolerate casual disruption. Windows, Microsoft 365, Outlook, Teams, and Surface are not internal dashboards. They are mass-market, enterprise-critical, security-sensitive platforms with immense compatibility burdens.That makes the switch a serious engineering-management test. AI coding assistants are not merely code generators. They are increasingly used to draft tests, inspect failures, propose migrations, explain legacy behavior, and generate scaffolding around unfamiliar codebases. Remove one assistant and replace it with another, and the output may look similar in a demo while feeling very different in practice.
Developers notice small differences quickly. One tool may be better at navigating a monorepo. Another may be better at shell commands. One may produce safer incremental patches, while another may be more aggressive about rewriting. In a consumer app, these differences are annoyances. In a Windows or Office codebase, they affect velocity, review load, and confidence.
Microsoft’s gamble is that Copilot CLI can absorb the workflow without creating enough friction to matter. That is plausible, especially because GitHub is deeply embedded in modern development practices and Microsoft can tune the product internally. But the short-term hit is real even if the long-term platform logic is sound.
This is also an internal dogfooding moment. If GitHub Copilot CLI can satisfy Microsoft’s own high-pressure engineering groups, that becomes a powerful proof point. If it cannot, Microsoft will learn the problem before its customers do.
The Competitive Optics Are Awkward but Not Fatal
Microsoft’s AI posture has always involved a careful mix of partnership, ownership, and competition. The company has invested heavily in OpenAI, integrated AI across Windows and Microsoft 365, and turned GitHub Copilot into one of the most visible commercial coding assistants. At the same time, enterprise AI customers increasingly expect model choice, and GitHub itself has been moving toward a more multi-model world.That makes a Claude Code pullback awkward. Microsoft can say it supports an ecosystem while still preferring its own tool internally. Both things can be true. But optics matter when developers are the audience, and developers tend to respect tools that win on merit more than tools that win by policy.
Claude Code’s reported popularity inside Microsoft complicates the story. If engineers adopted it because it solved problems well, then the company is not just swapping suppliers. It is overriding revealed preference.
Still, Microsoft’s position is defensible. Internal engineering is not a neutral marketplace. Companies standardize on compilers, security scanners, cloud platforms, package registries, and source-control policies all the time. AI coding agents are simply becoming important enough to be pulled into the same governance machinery.
The question is whether Microsoft can make the mandated tool feel like the best tool. If Copilot CLI becomes excellent because Microsoft’s own engineers push it hard, this episode may be remembered as a necessary consolidation. If not, it will feed the suspicion that AI strategy is being shaped more by balance sheets than by builders.
The Bigger Warning Is for CIOs, Not Just Developers
Every CIO wants the AI productivity story. Fewer want the spreadsheet that comes after it. Yet the spreadsheet is where the next phase of AI adoption will be decided.A coding assistant that costs a predictable per-user fee can be budgeted like collaboration software. A coding agent that burns through usage charges based on context windows, retries, tool calls, and background tasks must be managed more like cloud infrastructure. That means quotas, alerts, chargebacks, procurement reviews, and uncomfortable conversations about whether a generated patch was worth the run.
Cloud computing already taught this lesson once. The first wave of cloud adoption sold flexibility and speed. The second wave introduced FinOps because flexibility without discipline became surprise spending. AI coding is now entering its own FinOps era.
The irony is that the same practices engineers use to get better AI results can make costs harder to forecast. Supplying more files improves context. Asking for multiple approaches improves quality. Running agents through test failures improves usefulness. But each step can add cost, especially when multiplied across large teams.
Microsoft’s reported cutoff is a warning that AI coding tools cannot be evaluated only by demo quality. They must be evaluated by total operating cost, auditability, data handling, integration, and whether the company can explain the bill to finance without resorting to vibes.
Governance Is the Product Feature Nobody Wants to Demo
The least glamorous part of AI coding may become the most valuable. Enterprises need governance controls that determine who can use which models, against which repositories, with what data, under what spending limits, and with what logs preserved for review.That is not anti-developer bureaucracy. It is what turns a promising assistant into infrastructure. If an AI tool can write code that ships into Windows or Microsoft 365, then access and observability are not optional.
The governance layer must also distinguish between different kinds of usage. Generating boilerplate, writing unit tests, explaining stack traces, and drafting documentation may deliver strong returns at modest risk. Letting an agent roam a large codebase, invoke tools, rewrite major components, and iterate through failures is a different cost and risk profile.
The mistake many companies will make is treating all prompts as equal. They are not. A short explanation request and a multi-step agentic refactor may both appear as “AI coding usage,” but they belong in different budget and review categories.
Microsoft is better positioned than most to build that governance into its own stack. GitHub, Azure, Microsoft Entra, Defender, Purview, and the Microsoft 365 administrative universe all give it pieces of the control plane customers will eventually demand. Claude Code may be an excellent tool, but excellence alone does not solve the enterprise control problem.
The Productivity Debate Is Moving From Magic to Measurement
AI coding boosters often talk as if the productivity case is settled. In some narrow workflows, it may be. Test generation, repetitive migrations, documentation drafts, and first-pass code explanations can be obvious wins. But broad claims about AI writing a fixed percentage of code are less useful than they sound.Code volume is not the same as productivity. A tool can generate many lines and still increase review burden. It can accelerate simple tasks while introducing subtle defects. It can help senior engineers move faster while giving junior engineers plausible-looking code they do not fully understand.
That does not mean the tools are overhyped. It means the measurement problem is harder than the marketing. The right question is not “How much code did AI write?” but “Which work became faster, safer, cheaper, or more reliable because AI participated?”
Microsoft’s internal shift should be read in that context. If Claude Code created enough value to become popular, the company now has to preserve that value while changing the cost envelope. That is the real test of enterprise AI: not whether a tool can impress a developer, but whether it can survive procurement, governance, security, and operational measurement.
A company can tolerate fuzzy ROI during a pilot. It cannot tolerate fuzzy ROI at platform scale.
The Copilot CLI Mandate Puts Pressure Back on GitHub
GitHub Copilot CLI now has to do more than exist as Microsoft’s preferred alternative. It has to win back engineers who reportedly liked Claude Code enough to make the shutdown painful. That is a product challenge disguised as an internal policy.Command-line coding assistants live or die by trust. Developers need them to understand project structure, avoid destructive changes, explain commands before running them, and recover gracefully when they are wrong. A slick integration is not enough if the assistant creates more cleanup than leverage.
The opportunity for GitHub is substantial. Microsoft’s internal engineering base can become an unusually demanding proving ground. Windows and Microsoft 365 engineers can surface edge cases that most product teams would never encounter. If Copilot CLI improves under that pressure, customers benefit.
But there is risk in captive adoption. When a tool wins because employees are told to use it, product teams can mistake compliance for love. Microsoft will need to watch not just usage numbers, but friction signals: abandoned sessions, manual rewrites, review churn, escalations, and quiet workarounds.
Developers are pragmatic. If Copilot CLI works well, the resentment fades. If it does not, the policy becomes a drag on morale and productivity.
Anthropic Still Wins the Argument Even When It Loses the Account
There is another uncomfortable reading for Microsoft: Claude Code may have already won the conceptual battle. If the tool spread quickly inside one of the world’s most AI-invested software companies, that says something about Anthropic’s product-market fit with developers.Losing broad access inside Microsoft would sting, but it does not necessarily weaken Claude Code’s reputation. In some circles, it may strengthen it. Developers often interpret corporate bans on popular tools as evidence that the tools were too effective, too expensive, or too threatening to internal priorities.
That is not always fair. Security, compliance, and cost can justify restrictions even on excellent products. But developer folklore is powerful, and “Microsoft engineers preferred Claude until finance intervened” is exactly the kind of story that travels.
For Anthropic and other AI coding vendors, the lesson is clear. Capability is not enough. Enterprise buyers will demand predictable pricing, administrative controls, repository-aware policies, and audit trails that make large-scale deployment less frightening.
The next battle will not be only about model quality. It will be about packaging model quality in a way CFOs can approve and CISOs can defend.
Microsoft’s Retreat Draws the Map for Everyone Else
The most concrete lesson from Microsoft’s reported Claude Code pullback is that AI coding has crossed from experimentation into management. The tools are now useful enough to create dependency and expensive enough to create backlash. That combination forces decisions.This is the shape of the moment:
- Microsoft is reportedly ending most Claude Code access in its Experiences and Devices division on June 30, 2026, aligning the shutdown with the end of its fiscal year.
- Engineers working near Windows, Microsoft 365, Outlook, Teams, and Surface are being steered toward GitHub Copilot CLI instead of a rival AI coding assistant.
- The move appears driven by a mix of cost control, internal standardization, security governance, and Microsoft’s desire to strengthen its own GitHub-centered developer platform.
- Token-based AI pricing makes successful adoption financially volatile because heavier use can raise costs far faster than traditional per-seat software.
- The migration will test whether Copilot CLI can match the habits and expectations engineers built around Claude Code in real production workflows.
- Other enterprises should treat AI coding assistants less like ordinary SaaS subscriptions and more like cloud infrastructure that requires budgets, limits, telemetry, and review.
References
- Primary source: La Revue Tech
Published: 2026-06-12T00:03:07.328903
Loading…
larevuetech.fr - Related coverage: claudeapi.com
Uber Burned Through Its Full-Year AI Budget in 4 Months: 5,000 Engineers Averaged $500-$2,000 per Month, and 70% of Committed Code Was AI-Generated
Uber’s fiscal 2026 AI tooling budget was burned through by 5,000 engineers by April, with per-engineer monthly bills reaching $500-$2,000 and the COO calling it a “head-exploding” moment. Claude Code usage doubled in two months, and 84% of engineers moved into agentic coding. This article breaks...www.claudeapi.com
- Related coverage: winbuzzer.com
Microsoft Shifts Engineers from Claude Code to GitHub Copilot CLI
Microsoft is moving Experiences + Devices engineers from Claude Code to GitHub Copilot CLI by June 30, while keeping its broader Anthropic ties intact.
winbuzzer.com
- Related coverage: epcgroup.net
GitHub Copilot CLI Replacing Claude Code: What Microsoft Engineers Actually Do Now
Microsoft moved its Experiences and Devices engineering team from Claude Code to GitHub Copilot CLI by June 30, 2026. The actual day-to-day developer workflow change, what GitHub Copilot CLI does vs Claude Code, and the multi-model AI lesson every CIO should take from it.www.epcgroup.net
- Related coverage: insights.itdukes.com
Loading…
insights.itdukes.com - Related coverage: agent-wars.com
Uber's 2026 AI Budget Lasted Four Months. Claude Code Won.
Uber spent its entire 2026 AI budget by April after deploying Claude Code in December. With 95% of engineers using AI tools monthly and API costs of $500 to $2,000 per engineer, the CTO says the company is 'back to the drawing board' on funding. Claude Code dominates over Cursor, with 70% of...
www.agent-wars.com
- Related coverage: byteiota.com
Microsoft Claude Code Killed: Copilot CLI by June 2026
Microsoft is canceling Claude Code at Windows and Teams by June 30, forcing Copilot CLI instead. Token costs and platform strategy explain the real reason why.
byteiota.com
- Related coverage: windowsforum.com
Loading…
windowsforum.com - Related coverage: quasa.io
Microsoft Is Pulling Claude Code from Its Core Product Teams and Forcing a Switch to GitHub Copilot CLI
June 30, 2026 (the end of Microsoft’s current fiscal year). Developers have been told to transition to GitHub Copilot CLI in the coming weeks.quasa.io
- Related coverage: advancedai.com
Microsoft Drops Claude Code. Who Chooses Your AI Tools? | Advanced AI
Microsoft canceled internal Claude Code licenses with a June 30 deadline, routing engineers to GitHub Copilot CLI — while PwC simultaneously went all-in on Claude. The AI coding tool you rely on may not be yours to keep.www.advancedai.com
- Related coverage: forbes.com
Loading…
www.forbes.com - Related coverage: aiagentsfirst.com
Loading…
aiagentsfirst.com - Related coverage: techradar.com
Microsoft may discontinue Claude Code internally as it looks to push users towards GitHub Copilot
Claude Code has proven a hit among Microsoft's engineers, but the company now wants them to switch to GitHub Copilot CLI.www.techradar.com
- Related coverage: windowscentral.com
Microsoft is ditching Claude Code for Copilot CLI — but its own devs aren’t happy
Microsoft cancels Claude Code licenses in favor of GitHub Copilot CLI. Financial motives could be in play.
www.windowscentral.com