This article discusses the recent shift in AI subscription pricing models, moving away from unlimited access to metered billing. It highlights key changes announced by Anthropic, OpenAI, and GitHub, and their implications for enterprise AI budgets.
Key Announcements and Changes
- Anthropic:
- Splitting Claude subscriptions into interactive and automated usage pools.
- Automated usage (agentic tools) now requires separate credit meters.
- Enterprise contracts are transitioning to usage-based plans.
- A new tokenizer in Claude Opus 4.7 effectively increased costs by up to 27%.
- OpenAI:
- Offering two free months of Codex access to users switching from Claude.
- GitHub:
- Transitioning Copilot plans to usage-based billing on June 1, 2026, using AI Credits.
Reasons for the Shift
- Unsustainable Costs: AI labs were selling compute at a fraction of its cost.
- Agentic Workloads: Autonomous agents generate significantly more requests than human users, making flat-fee subscriptions unsustainable.
- IPO Timelines: OpenAI and Anthropic are on IPO timelines and need to demonstrate a path to profitability.
Impact on Enterprises
- Budget Overruns: Companies like Uber and ServiceNow have already exceeded their AI budgets.
- Increased Spending: KPMG data shows U.S. organizations projecting nearly double their AI spending.
- Two Emerging Coalitions:
- OpenAI (Microsoft-backed): Unified, unrestricted enterprise plans.
- Anthropic (Google, Amazon-aligned): Tiered, capacity-controlled subscriptions with per-token billing.
Recommendations
The article reiterates the importance of:
- Auditing token consumption.
- Modeling costs at 2x to 10x current prices.
- Building vendor optionality.
- Discussing AI budgets with CFOs proactively.
Conclusion
The era of unlimited AI subscriptions is ending, with significant implications for enterprise AI budgets and vendor selection. The shift to metered billing is driven by unsustainable costs and the need for AI companies to demonstrate profitability as they approach IPOs. Companies need to adapt to these changes by carefully managing their AI usage and costs.