Google Cuts AI Model Prices as Companies Track Token Costs

Google announced price reductions on its Gemini 3.5 Flash model and stated that the change targets companies facing rising AI token expenses. The company reported that monthly usage of its AI products reached 3.2 quadrillion tokens, up seven times from the prior year.

Google CEO Sundar Pichai said companies are already exceeding annual token budgets in May. He added that moving 80 percent of workloads at top Google Cloud customers to a combination of Flash and frontier models could save more than $1 billion annually.

Performance differences between leading AI models have narrowed, moving competition toward infrastructure and inference costs. Google stated it controls the full stack of chips, data centers, cloud services, and models, allowing it to reduce internal compute expenses by 50 to 75 percent compared with rivals that rely on external providers.

The company noted that its approach mirrors the infrastructure strategy it used to scale Search, where faster and cheaper operations supported wider adoption. Google said it has built custom systems with off-the-shelf components for more than 25 years to maintain speed while limiting costs.

Some companies have reported difficulty justifying continued AI expenditures. Uber's chief operating officer stated that ballooning AI costs are becoming harder to support. Venture capitalist Chamath Palihapitiya said his firm 8090 stopped using Cursor after token expenses grew too large.

Google said smaller AI providers are raising prices, prompting customers to review total spending. The company positioned its lower-cost models as an alternative for organizations that no longer require the absolute highest capability.