OpenAI and Anthropic face spending-driven growth slowdown

CNBC reports that enterprise customers are constraining AI budgets and shifting away from aggressive usage patterns branded as "tokenmaxxing." The article quotes D.A. Davidson analyst Gil Luria saying, "Some of their largest enterprise customers may start limiting their out-of-control token spend." CNBC profiles startup Lindy and quotes CEO Flo Crivello saying the company moved 100% of its traffic off Anthropic's Claude models to Chinese provider DeepSeek, and that the switch sharply reduced Lindy's cost curve. CNBC also reports that open-source models are emerging as cheaper alternatives and that Microsoft, Amazon, and Google are proposing efficiency-focused offerings.

The reported shift foregrounds three technical pressures practitioners already face: inference cost per token, model selection for cost-performance tradeoffs, and engineering effort to integrate lighter-weight or open-weight backends. Companies undertaking comparable transitions typically prioritize latency-cost curves, embedding dimensionality versus retrieval frequency, and batching or quantization to reduce token spend.

Reporting frames this as a potential growth headwind for API-driven incumbents because enterprise unit economics change when customers optimize for cost per useful output. CNBC's sourcing of an industry analyst and first-hand customer examples lends empirical weight to the trend away from unconstrained token consumption. Broader coverage from PYMNTS and Investing.com corroborates that OpenAI is actively weighing significant price cuts in response to pricing pressure from Anthropic's Claude Code momentum and open-source alternatives.

Track enterprise RFP language for cost-per-inference clauses, pricing moves from major model vendors, open-source adoption rates, and published token-usage metrics from large customers. Observers should also watch vendor product announcements that explicitly bill on efficiency or fixed-cost inference tiers.