SAN FRANCISCO — Google's Gemini application now charges users based on compute consumed, departing from its previous daily prompt limits. The change, announced at Google I/O, factors in chat length, features used and prompt complexity.
The shift reflects the rising cost of delivering AI services at scale. Microsoft's GitHub Copilot moved to usage-based billing in June, and Anthropic's Claude Enterprise made a comparable change in April.
Enterprises are absorbing higher costs as token consumption grows — a dynamic Uber Chief Operating Officer Andrew Macdonald said is becoming difficult to justify. Companies are moving from promotional or subsidized AI packages to standard production pricing as deployments mature.
Brands and agencies are tracking how the pricing changes affect their budgets. Some are also building strategies to pass AI costs through to clients, balancing internal value against rising subscription fees.
Full Glass Wine Co. a company using enterprise AI, has set aside new budget for these tools. Co-founder and co-CEO Neha Kumar said the investment lets her team move faster.
The company uses ChatGPT and Gemini to summarize communications. Its sommeliers, for example, use ChatGPT to generate wine-pairing notes — work that previously required a copywriter for a full day. Kumar said the company has built AI costs into its budget.
Marketing technology startup Pomo is watching rising enterprise AI costs as it scales. Co-founder and CEO Praneet Dutta said the company relies on AI coding tools, with engineers using OpenAI's Codex and Anthropic's Claude Code. Pomo also builds on the Databricks enterprise AI platform, which is an investor in the startup.
Dutta said building a company of Pomo's scale would not be possible without these AI tools.
CIOs and other technology leaders have spent much of 2026 focused on controlling AI costs as adoption spreads across their organizations, balancing the push for innovation against budget pressure.
