AI

The AI Cost Reckoning Has Arrived — Microsoft Cancels Claude Code Licenses as Uber Burns Entire Annual AI Budget in Just Four Months

Microsoft is cancelling most internal Claude Code licences by June 30, migrating to GitHub Copilot CLI. Uber exhausted its annual AI budget in four months after 92% of engineers adopted AI coding tools. Sam Altman and Dario Amodei now admit their job-loss predictions were wrong.
AI cost reckoning Microsoft Uber enterprise spending

The era of unconstrained AI spending in enterprise technology may be coming to an abrupt end. Microsoft is cancelling the majority of its internal Claude Code licences by 30 June 2026, instructing developers to migrate to GitHub Copilot CLI. Uber has disclosed that its entire annual AI budget was exhausted in just four months after rolling out Claude Code to 5,000 engineers. And the CEOs who warned most loudly about AI job displacement — Sam Altman and Dario Amodei — are now admitting they were “pretty wrong.”

These are not isolated incidents. They represent a structural correction in how companies are absorbing the costs of AI-assisted development. The technology works — that much is beyond dispute. What is now being questioned is whether it works at a price that enterprise budgets can sustain.

What Happened at Microsoft

Microsoft’s decision to cancel Claude Code licences is both practical and strategic. Under Anthropic’s token-based pricing model, per-engineer monthly costs ranged from $500 to $2,000, depending on usage intensity. For a company with tens of thousands of engineers, that adds up to hundreds of millions of dollars annually — a line item that even Microsoft’s scale cannot absorb indefinitely.

The migration to GitHub Copilot CLI serves a dual purpose. It consolidates Microsoft’s internal developer tools onto its own platform, and it shifts pricing to a usage-based billing model that is expected to launch in June 2026. This gives Microsoft greater control over per-user costs and keeps the revenue from AI-assisted coding within its own ecosystem rather than sending it to Anthropic.

For developers at Microsoft, the transition means adapting to a different tool — one that is tightly integrated with GitHub’s infrastructure but may lack some of the capabilities that made Claude Code popular, particularly its ability to handle complex multi-file refactoring and extended reasoning tasks. When Anthropic launched Claude Opus 4.8 with major intelligence upgrades, it highlighted exactly the kind of advanced capability that Copilot CLI will need to match.

Uber’s Budget Implosion

Uber’s experience is perhaps the most vivid illustration of AI cost dynamics in 2026. When the company rolled out Claude Code access in December 2025, adoption was pull-driven — engineers opted in voluntarily rather than being assigned the tool. By December, 32 per cent of engineers were using it. By February, that figure hit 63 per cent. By April, the CTO reported that the entire annual AI budget was already spent.

The numbers are staggering. Uber has roughly 5,000 engineers. By April 2026, 92 per cent were using AI tools monthly. Nearly 70 per cent of code was AI-generated. Eleven per cent of pull requests were opened by autonomous agents. Over 65,000 code changes were submitted with AI assistance.

The productivity gains are real. Teams are shipping features faster, reducing time spent on boilerplate code, and catching bugs earlier. But the cost per unit of productivity has proved far higher than initial projections assumed. Token consumption scales with usage, and when engineers discover that an AI tool can handle tasks they previously spent hours on, they use it for everything — driving costs exponentially higher.

The Job Apocalypse That Never Arrived

Against this backdrop of surging costs and productivity gains, the predictions about mass AI-driven unemployment look increasingly disconnected from reality. Sam Altman, in an interview with Commonwealth Bank of Australia CEO Matt Comyn on 26 May, said he was “pretty wrong” about AI’s impact on entry-level white-collar jobs.

“I thought there would have been more impact on entry-level white-collar jobs being eliminated by now than has actually happened,” Altman acknowledged. He added that while he accepts criticism for contributing to fear, he maintains that the caution was warranted at the time.

Dario Amodei, CEO of Anthropic, has undergone a similar revision. Having previously claimed that AI could eliminate 50 per cent of white-collar jobs, he now says automation may actually expand the scope of work people do rather than replacing them. Goldman Sachs CEO David Solomon has been making a similar argument since late 2025, pointing to a century of American economic history where technology consistently created more jobs than it destroyed.

The layoffs at companies like Wix, which cut 20 per cent of its workforce citing the “AI revolution,” suggest that some displacement is occurring. But the pattern across the broader economy is augmentation, not replacement — workers using AI to do more, not companies replacing workers with AI.

What This Means for Indian Companies

Indian enterprises adopting AI coding tools are about to face the same cost reckoning. With AI tool adoption surging across Bengaluru, Hyderabad and Delhi, companies that scaled up usage without rigorous cost modelling may find their quarterly budgets overwhelmed.

The cost dynamics are particularly acute for Indian startups, where AI tool spending can quickly become a significant percentage of overall burn rate. A startup with 50 engineers using Claude Code at $1,000 per engineer per month is spending $600,000 annually on AI coding tools alone — a material expense that VCs and board members are increasingly scrutinising.

The mitigation strategies are emerging: usage caps per engineer, tiered access based on project priority, and hybrid approaches that combine premium AI tools for complex tasks with lighter-weight alternatives for routine coding. Microsoft’s shift to usage-based billing is designed to give enterprises more control, but it also means that costs scale linearly with productivity — a double-edged sword.

The Correction Is Healthy

The AI cost reckoning of mid-2026 is not a failure of the technology. It is a predictable correction in how organisations consume a fundamentally new type of resource. Token-based pricing creates incentives for unlimited consumption, and when powerful tools are given to talented engineers, consumption explodes.

The companies that navigate this phase successfully will be those that treat AI tool spending as a strategic investment requiring the same discipline as cloud infrastructure or headcount planning. The ones that do not will find themselves in Uber’s position — delivering impressive productivity metrics while blowing through budgets that were set for an entire year.

Anthropic’s growing portfolio of AI capabilities and the intensifying competition in enterprise AI tools suggest that pricing pressure will eventually bring costs down. But for the next 12 to 18 months, the industry is in a recalibration phase where the gap between AI capability and AI affordability will define the strategic choices of every technology-intensive organisation.

Ankit Thakur

Ankit Thakur

Ankit Thakur is an Editor at Daily Tips overseeing sports and entertainment coverage. A lifelong sports enthusiast with years of journalism experience, he covers cricket, kabaddi, football, esports, and gaming. He also manages the publication's entertainment vertical, bringing insider knowledge and passionate storytelling to every piece.

View all posts by Ankit Thakur →