A company reportedly accumulated a US$500 million bill from Anthropic’s Claude in a single month, as soaring enterprise AI costs have driven companies including Uber, Cisco and Walmart to introduce internal guard rails to prevent runaway inference spending. In mid-May, Microsoft announced the cancellation of its internal Claude Code licences to rein in operating costs. The article notes that most businesses have a cheaper alternative: open-weight models. It distinguishes between "open source" and "open weight," explaining that truly open-source AI models, such as Ai2’s OLMo, make source code, training data, and model weights public, while most AI labs release only the model weights, as seen with DeepSeek’s models, Meta’s Llama family, and certain versions of Ali.