Every Labs generation makes real, paid AI calls. The spending cap is the account's AI budget in USD: the sum of every ledger row (each paid step writes one โ tokens, model, cost) is the account's lifetime spend, and when it reaches the cap, generation stops. The cap is checked at the door AND before every step, so a long run cannot blow through the budget on one check; a run stopped mid-flight parks as failed with reason spending_cap_exceeded.
The cap is per-account configuration on the Labs grant (platform default $0.10), raised by a platform operator. Raising it un-parks nothing by itself โ regenerate the parked deliverable once headroom exists. The full money trail is the account's Usage & spend page.
Related: Deliverable ยท Process