Microsoft has ended its corporate licenses for Anthropic’s Claude Code within its Experiences & Devices Division, the unit responsible for Windows, Office, Teams, and Surface. Engineers were mandated to migrate to GitHub Copilot CLI by June 2026, according to AD HOC NEWS; the decision came after costs spiraled due to the token-based pricing model, which proved unsustainable for internal use. GitHub Copilot is also shifting to usage-based billing in July 2026, with U.S. prices increasing by 20% to 37%.

A Costly Transition for Microsoft and GitHub

According to AD HOC NEWS. GitHub has faced 257 incidents between May 2025 and April 2026, including 48 classified as severe; February 2026 was the worst month on record with 37 interruptions. The CI/CD service Actions was the hardest hit, accounting for 57 of these incidents, as the underlying issue is a delayed migration to Microsoft’s Azure cloud, as GitHub still relies on older data centers that cannot scale effectively. Chief Technology Officer Vlad Fedorov initially pledged to increase capacity tenfold but admitted in February that the target was insufficient, with GitHub now needing thirty times its current scale.

Anthropic Launches Claude Opus 4.8 Amid Market Competition

On the same day that Microsoft made its decision, Anthropic released Claude Opus 4.8, a new version of its large language model, According to Anthropic, the model offers improvements in coding, reasoning, and knowledge work. It also introduces an “effort control” feature that allows users to adjust how much computing power the model uses for a given task, helping to manage costs. The fast mode of Opus 4.8 is three times cheaper than previous versions, as noted by Business Insider.

Opus 4.8 was also launched alongside a “dynamic workflows” feature in Claude Code, enabling the model to handle large-scale coding tasks; Early testers reported that the model was more reliable and had better judgment when performing agentic tasks, according to Anthropic. The company also emphasized that Opus 4.8 outperformed prior models and GPT-5.5 on the Super-Agent benchmark, completing every case end-to-end.

The Ongoing Battle for AI Market Share

Business Insider notes that Anthropic and OpenAI are in a high-stakes race for technological and market dominance; Anthropic’s latest valuation has jumped from $380 billion in February to a higher figure, while OpenAI’s last valuation was $852 billion in March. The release of Opus 4.8 coincides with OpenAI’s GPT-5.5 model, launched in April; Anthropic also introduced a new feature called “effort control,” which gives users more control over resource allocation and helps manage AI spending.

According to Ecosistema Startup. The competition between Anthropic and OpenAI intensified on February 5, 2026, when both companies launched their latest coding models — GPT-5.3-Codex and Claude Opus 4.6, simultaneously. These models marked the start of what the industry is calling ‘the wars of AI for code,’ a battle for enterprise software development market share worth billions of dollars. GPT-5.3-Codex was reportedly used in its development process to improve its own training and deployment, according to OpenAI CEO Sam Altman.

Microsoft’s decision to cancel its internal use of Claude Code after just six months highlights the financial risks of token-based pricing models — the company had exhausted its annual AI budget quickly, according to Ecosistema Startup. Anthropic moved away from flat-rate pricing to a token-based model, which can lead to unexpected costs, especially for coding agents that consume more resources, though the model processes large files, performs multiple iterations, and generates extensive code outputs, all of which contribute to rising expenses. These costs are often hidden until the final bill is received, making budgeting difficult for enterprises.