Meta and Microsoft Cut Claude Use & Spending, Workday Faces a New AI Threat

Big tech is replacing Anthropic's Claude internally with first-party models, pressuring Anthropic's growth narrative ahead of its IPO.

• Beam's efficiency bet: Reflection's new 500B-parameter open-weight model Beam used the largest documented RL run (10,000+ GPUs, four weeks, 100M+ rollouts) to optimize token efficiency rather than raw benchmark-topping.
• Claude spending pullback: Microsoft cut internal Claude spend over 33% from a ~$1B peak by shifting engineers to free OpenAI access and internal MAI models; Meta is pushing its Muse code model instead.

Themes, sentiment and reasons are AI-written paraphrases, checked so they never quote an episode. They may contain errors. Not investment advice. Episodes remain the property of their creators.