Spend caps and budgets
Cap what your organization can be charged in a month on Gigadrive Network, and know exactly what stops when the cap engages.
A budget caps what your organization can be charged in a calendar month. When month-to-date spend reaches it, Gigadrive Network stops the work that would add to the bill instead of letting it grow.
Two ways spending stops
| Budget cap | Included compute exhausted | |
|---|---|---|
| Applies to | Any plan | Plans that never bill usage overage, which today means Starter |
| Measured on | Rated spend across every metric, month to date | Active CPU, provisioned memory and Function invocations only |
| Warns at | 80% of the cap | 80% of any of those three allowances |
| Engages at | 100% of the cap | 100% of any of those three allowances |
| What stops | New deployments, running Functions, AI Gateway requests | New deployments and running Functions |
| How it clears | Raise the budget, or the next billing period starts | Upgrade the plan, or the next billing period starts |
They are deliberately separate, because the remedy is different. A budget is a number you chose and can change. An exhausted allowance is the shape of the plan, and the way out is a plan with more of it.
Setting a budget
The card is at /{organization}/settings/billing, called Spend budget. Enter an amount in euros in
the Monthly budget field and save. Leave the field empty to fall back to your plan's default cap,
which is €25 on Starter and no cap on Pro and Enterprise.
Saving re-arms enforcement against the new number, so raising the budget after it has engaged lifts the block immediately, without waiting for the next period.
Spend is measured as rated euros, not raw counters, so one number covers every figure on Usage metrics at once. Usage is rated on every plan, which is why Starter has a default cap despite never producing a usage bill.
A cap suspends running Functions
Unlike a quota, which only refuses new allocations, an engaged budget cap pauses
compute your organization is already running. Visitors then get HTTP 503 with the "Compute limit reached" page and
the error code EDGE_ORGANIZATION_COMPUTE_LIMIT.
What happens when a cap engages
Deployments are refused. In the console:
Your organization has reached its monthly billing cap. Raise your budget to keep deploying.Through the API, the same refusal is HTTP 402:
{
"error": "Organization billing cap reached. Raise your budget to create new deployments."
}Running Functions are suspended, and one that traffic tries to wake afterwards is refused rather
than started. AI Gateway requests are refused with HTTP 429 and the error code
billing_cap_reached. Paid build-cache overage is paused for as long as the cap is engaged, so the
cache keeps working but its ceiling drops back to the allowance your plan includes.
See Build cache.
An exhausted compute allowance produces the same shape with different wording, pointing at the plan rather than the budget:
{
"error": "Included compute for this billing period is used up. Upgrade your plan to create new deployments."
}Timing
Enforcement runs once an hour, against usage that is at most about an hour old. Neither the warning nor the cap fires more than once per period.
After you raise a budget, a Function can take up to 30 seconds to start waking again.
Emails go to the organization's billing recipients: one at 80% of the budget, one when the cap engages, and the equivalent pair for included compute. Each is sent at most once in a billing period.
AI Gateway budgets are separate
The organization budget is a single monthly ceiling on everything. The AI Gateway has its own budgets on top of it, scoped to the organization, one application or one user, and checked per request before a model is called. Use those to bound model spend specifically. See Budgets.