· Data Platforms · 2 min read
Demystifying BigQuery Slots: Flat Rate vs On-Demand
BigQuery pricing can be confusing. We explain what a 'Slot' actually is and how to choose between paying per query or paying for capacity.
When you run a query in Google BigQuery, you aren’t renting a server; you are renting thousands of tiny CPUs for a few seconds. These units of compute are called Slots.
Understanding the Slot
A Slot is a virtual CPU dedicated to executing SQL.
- A query might need 10 slots to read a small table.
- A complex join on a Petabyte table might need 2,000 slots.
The Pricing Models
1. On-Demand (The Default)
You pay $5 per Terabyte of data scanned.
- Pros: Zero commitment. Easy to understand.
- Cons: Unpredictable bills. If a junior analyst writes a
SELECT *on a massive table, it could cost $100 instantly. Also, you are capped at 2,000 slots (“Burst”), so huge queries might run slow.
2. Capacity (Editions)
You pay for a fixed number of slots per hour (e.g. 100 slots).
- Pros: Predictable monthly bill.
- Cons: If your queries need 200 slots and you only bought 100, the query runs slower (it queues).
The Autoscaling Sweet Spot
Google now offers BigQuery Editions (Standard, Enterprise, Enterprise Plus) which allow Autoscaling. You set a min and max (e.g. Min 0, Max 1000). You pay for what you use, but with a safety cap to prevent bill shock.
Recommendation: Start with On-Demand. Move to Editions only when your monthly bill exceeds $10,000 or you need guaranteed performance.
BigQuery costs out of control? We help enterprises master their GCP spend. Get a FinOps review.
