Separate Google usage cost from payment timing

How do Google Prepay and Postpay work together?

Google’s Gemini API billing documentation says affected developer accounts are being transitioned from Postpay to Prepay. The deadline is account-specific: follow the cutover date in the notice Google sent your account rather than assuming one platform-wide date. Google’s split charging-cycle documentation explains the resulting structure: one Cloud Billing account linked to separate Google payments accounts. Gemini API usage from AI Studio is paid through the Prepay account, while other Google Cloud products and eligible services continue through the Postpay account.

Consider a project that calls Gemini and runs a Compute Engine worker. The Gemini calls accrue usage cost and reduce prepaid credits. The VM also accrues usage cost, but Google collects that amount later through Postpay. Both cost streams remain attached to the Cloud Billing account even though they are paid at different times.

That produces four distinct concepts:

Concept	What it means	Where CostCompass gets it
Accrued usage cost	The provider’s monetary value for consumed services	BigQuery billing export, or a token-based estimate before export setup
Payment or top-up	Money added to Prepay or collected through Postpay	Not counted as another usage event
Prepaid credit balance	Credits remaining for Gemini API usage	Not available to CostCompass; managed in AI Studio
Runway	Balance divided by recent burn rate	Not available for Google because the balance is unavailable

The Cloud Billing account is the common cost ledger. The payments accounts determine when those costs are paid. A top-up must not be added to the exported usage cost: doing that would count the same consumption twice.

This is a billing change, not an inference API migration. Existing Gemini API endpoints, request bodies, and response handling do not need to change merely because an account moves to Prepay.

What does CostCompass measure?

CostCompass supports two cost sources, chosen per Google card.

The default source is Cloud Monitoring. CostCompass reads Gemini API and Vertex AI token metrics for the projects you selected, applies stored model rates, and marks the result Estimated. This is an AI-only reading. It does not include Compute Engine, Storage, networking, a prepaid balance, or a top-up. It also cannot include explicit Gemini context-cache storage: Google charges for the cache’s token count and lifetime, but does not publish a cache-storage meter in Cloud Monitoring. CostCompass does not guess the missing duration or storage usage.

The second source is Google’s Standard usage cost export. Google states that a configured export includes all cost data accrued in the Cloud Billing account, including Gemini API usage in AI Studio. CostCompass queries that BigQuery table for the selected projects and groups the results by AI model or cloud service. Because it reads Google’s monetary cost rows rather than reconstructing every charge from Monitoring metrics, it includes provider-reported explicit context-cache storage charges when Google exports them.

The dashboard calls this reading Exact because the monetary amount comes from Google rather than from CostCompass’s token-rate calculation. More precisely, it is provider-reported accrued usage cost. The query sums the export’s cost field; it does not subtract separate credit rows, read the prepaid balance, or reconstruct cash payments. Cost details normally arrive after the usage and can take more than 24 hours to appear.

Prepay therefore does not break the existing query. The usage still accrues in Cloud Billing, and Google still exports it to BigQuery. What changed is the payments account that settles the Gemini portion.

What setup unlocks provider-reported cost?

If an AI-only estimate is enough, connect Google, choose the projects, and refresh. To include non-AI services and use Google’s reported monetary values, enable Standard usage cost export once:

Close

1. Open Billing → Billing export in Google Cloud.
2. On the BigQuery export tab, enable Standard usage cost.
3. Choose the project and dataset that will receive the export, then save.
4. In CostCompass, check for the export and refresh the Google card after data begins to arrive.

The connection remains read-only. CostCompass can list the projects and billing accounts you authorize, read Cloud Monitoring metrics, and query the selected billing-export table. It cannot buy credits, configure auto-reload, read the Gemini prepaid balance, or move an account between charging cycles.

What should you monitor under Prepay?

Google’s Gemini API billing documentation makes the operational boundary clear:

• Paid Gemini API usage deducts from purchased credits in near real time.
• When the Prepay balance reaches zero, paid Gemini API services linked to that billing account stop; projects do not automatically fall back to the free tier.
• Auto-reload can purchase more credits when the balance falls below a chosen threshold. A Monthly Auto-Charge Limit caps automatic reloads, but not manual purchases.
• Billing processing has a short delay, so long-running or fast workloads can consume beyond the displayed balance before Google halts them.
• Purchased credits expire 12 months after purchase.

These controls belong in AI Studio because they act on the payments account. CostCompass covers the complementary question: how quickly is the underlying usage cost accumulating across Google and the rest of the stack?

Use both surfaces deliberately. Keep enough prepaid credit or a suitable auto-reload policy in AI Studio to protect service continuity. Use the daily cost trend to find model changes, traffic ramps, and infrastructure growth before they become the next funding or payment surprise.

What does Forecast mean for a prepaid provider?

Forecast remains a consumption projection. CostCompass takes usage from the trailing seven days, converts it to a daily burn rate, extends that rate across every day in the next calendar month, and adds fixed subscriptions expected in that month.

It does not predict the next card charge. An auto-reload might happen tomorrow, several times in one month, or not at all. Forecast also does not mean days until empty. That calculation is runway, and runway requires a current prepaid balance that Google does not expose to CostCompass.

The distinction lets the shared dashboard keep one useful meaning across providers:

• Forecast: what next month’s consumption is likely to cost at the recent rate.
• Balance: provider-reported credit remaining, when the provider exposes it.
• Runway: how long that balance lasts at the recent burn rate.

For Google, Forecast and burn rate are available. Gemini balance and runway remain unavailable.

How do you track Google costs in one view?

The Google card groups daily AI usage by model and billing-export usage by model or cloud service. Older model usage stays visible rather than being folded into whichever model is current. Selected projects under one billing account share one card; projects under another billing account use another card.

Google does not sit alone. The same dashboard can combine its accrued usage cost with Claude, OpenAI, OpenRouter, and other connected services. Tracking AI costs across providers explains how those different source types fit into one total.

The result is intentionally narrower than a replacement for Google’s billing pages. AI Studio remains where you manage Gemini credits and payment history. Google Cloud remains where you configure the export and inspect Google-only reports. CostCompass keeps the accrued cost and forward consumption view beside every other provider you run.