Google says Gemini API spend caps take 10 minutes to take effect

• Gemini API spend caps have roughly ten minutes of enforcement latency, down from more than 24 hours, according to Google's explanation.
• Budget alerts send emails; spend caps enforce limits after a delay, as Google clarified in a reply.
• Immediate enforcement remains unavailable, a caveat Google acknowledged in a follow-up.

The Gemini API billing guide warns that batch completions and agent sessions can overrun project caps. Meanwhile, cost-breakdown graphs can still take up to 24 hours to update.

Google's Logan Kilpatrick says the current experimental version has about ten minutes of latency, compared with more than 24 hours originally. The team is working on a real-time version, he said, without giving a release date.

Charges can continue accumulating while billing data is being processed, according to Google's documentation. Ten minutes is a long leash for a runaway workload.

Kilpatrick distinguished two controls in the exchange:

• Budget alerts: email notifications about spending.
• Gemini API spend caps: enforced limits, currently subject to the ten-minute delay.

Cloud Billing also supports using budget alerts to automate cost-control responses, a separately configured response to a notification.

The controls operate at project and billing-account level, with different scopes, according to Google's billing documentation:

• Project caps: user-set monthly limits in AI Studio under Spend → Monthly spend cap → Edit spend cap. Project editors, owners and admins can set them.
• API keys: no independent billing settings. Usage from every key in a project counts toward that project's cap and the billing account's total.
• Billing-account caps: tier-based monthly limits aggregate spending across linked projects. Reaching the account limit pauses Gemini API service across those projects until the next billing cycle, starting on the first of the month.
• Eligibility: spend caps are unavailable for invoiced, or offline, accounts.

Moving a project to another billing account preserves its configured project cap but resets its accumulated spend to $0 for the new billing cycle, the same guide states.

A prepaid balance also has a delayed cutoff. Google's Prepay rules specify that a zero balance stops all API keys across every project linked to the billing account; requests then fail with HTTP 402 Payment Required until credits are added.

The guide warns that Prepay billing has approximately ten minutes of billing-pipeline latency, allowing charges beyond the available balance before usage is halted.