HomeArticles

Chapter 1 of Analytics databases: ClickHouse, Postgres and warehouses

GCP BigQuery Pricing 2026: The Real Cost Guide

How BigQuery pricing works: on-demand versus edition slots, hidden costs from streaming, storage and exports, a comparison with Athena, and cost cuts.

BigQuery pricing basics: $6.25/TiB on-demand, $0.02 and $0.01/GiB storage tiers, free batch load and export to Cloud Storage.

gcp bigquery pricing 2026 isn’t just a table of per‑TB numbers anymore. It’s a decision tree with hidden branches – editions, slot reservations, capacity commitments, storage tiers, and the elephant in the room: data ingestion and export costs. By the end of this guide you’ll know exactly what to pay, what to avoid, and how to structure your BigQuery spend.


What Changed in BigQuery Pricing in 2026?

Google Cloud revamped its BigQuery pricing in late 2024 and the changes solidified through 2025‑2026. The old “pay per byte scanned” model is still there, but it’s only one option among four. The new structure looks like this:

  • On‑demand (analysis pricing): $6.25 per TB scanned (US multi‑region) – yes, it went up from $5 in 2023.
  • Edition‑based slots: Basic, Standard, Enterprise, Enterprise Plus. Each comes with minimum 100 slots. Google Cloud Pricing 2026 breaks down the hourly rates.
  • Capacity commitments: 1‑year or 3‑year flex slots give 30–50% discount over monthly commits.
  • Storage: $0.02 per GB per month for active, $0.01 for long‑term (90 days inactive).

Most people think “on‑demand is fine for startups.” They’re wrong. If your query volume is predictable – even vaguely – slot‑based pricing beats on‑demand by a mile. The real trick is matching the right edition to your workload.

Editions: What You Actually Get

Edition Slots (min) Features Best for
Basic 100 Standard SQL, streaming, BI Engine Ad‑hoc queries, small teams
Standard 100 +auto‑scaling, materialized views Production dashboards
Enterprise 100 +Omni, multi‑cloud, high availability Data mesh, cross‑cloud
Enterprise Plus 100 +all features, reserved throughput, SLA Mission‑critical, real‑time

The Enterprise edition is where most serious workloads land. It includes BigQuery Omni (query across AWS/Azure) and the new dynamic scaling that kicks in within 30 seconds – a huge win compared to the old fixed‑slot model where you’d either idle or queue.


The Hidden Costs That Eat Your Budget

Scanning data is not the only cost. Four hidden cost drivers eat budgets:

1. Streaming Inserts

Every row you stream into BigQuery costs $0.05 per 200 MB (inserted payload). That’s $256 per 1 TB. If you have high‑volume event streams – say 50K events/second – you can easily rack up $10K/month just on inserts. Solution: batch your inserts into 10‑minute windows or use Dataflow with autoscaling. The runtime costs of Dataflow often net lower than streaming row charges.

2. Storage‑as‑you‑go vs Default Table Expiration

BigQuery storage is cheap on paper. But if you never set table expiration and keep all historical data, the long‑term tier kicks in after 90 days. Fair enough. However, the active storage for tables you query regularly – the clustering, partitioning, and metadata – still costs. A 10‑TB table that you partition by day and cluster by user_id will cost ~$2,000/month in storage if you have 30 days of active data and 270 days in long‑term. Sounds fine until you realise your query pattern only needs the last 7 days. Then you’re paying rent on data you never touch.

Pro tip: Use partition pruning and clustering intelligently. If your queries filter on date >= CURRENT_DATE - 7, BigQuery skips other partitions – but you still pay storage for them. Delete or move old partitions to a cheaper bucket like Cloud Storage (nearline/coldline) and query using external tables. More on that under gcp storage options comparison later.

3. Data Export (Yes, That Costs Too)

Exporting results from BigQuery to Cloud Storage is free. Exporting to outside GCP? $0.01 per GB. That sounds small until you’re moving 200 TB to a Snowflake instance or feeding an ML pipeline on AWS. GCP vs AWS 2026 | Which Cloud Platform Is Better? mentions this as a common lock‑in point that startups overlook.

4. Query Complexity Multipliers

BigQuery bills by the bytes processed, but not all bytes are equal. Queries that do SELECT * scan the entire table – that’s the obvious trap. What about queries with complex joins, subqueries, or user‑defined functions? They don’t increase the scanned bytes, but they increase slot usage. If you’re on slot‑based pricing, a slow query that consumes 500 slots for 3 minutes costs the same as a fast query that uses 500 slots for 30 seconds. Google doesn’t care about wall clock – it cares about slot‑seconds consumed.

The result: poorly written queries on slot‑based editions cost more than the same queries on on‑demand. A dashboard query that does a CROSS JOIN on two 1‑billion‑row tables is a typical example.


BigQuery Pricing vs AWS Athena and Azure Synapse in 2026

Let’s compare real numbers. All prices are for US East / us‑central1 as of July 2026.

Query Cost (per TB scanned)

Service On‑demand per TB Reserved slots (1‑yr) / equivalent
BigQuery $6.25 ~$2.80/TB (Enterprise, 1000 slots)
AWS Athena $5.00 + S3 costs ~$3.50/TB (with workgroups, no reserv.)
Azure Synapse $7.00 ~$4.00/TB (with reserved capacity)

BigQuery’s on‑demand is 25% more expensive than Athena per TB. But BigQuery’s reserved slots can bring that down to $2.80 – half of Athena’s reserved. Cloud Computing Cost: AWS vs. Azure vs. GCP Pricing in 2026 notes that for predictable workloads, GCP wins. For pure ad‑hoc queries with no commitment, Athena is cheaper.

Storage Cost (per GB per month)

Service Active Long‑term
BigQuery $0.02 $0.01
S3 (Athena) $0.023 $0.012 (on Infrequent Access)
Azure Blob (Synapse) $0.018 $0.01

BigQuery is slightly cheaper than S3 for active storage, but the game changes when you consider that Athena charges per query plus S3 data retrieval. If your data lives in S3 and you rarely query it, Athena is cheaper. If you query the same data daily, BigQuery’s built‑in storage and caching win.

The Verdict

Match the service to your query volume, not your ego.

For startups, the best gcp data warehouse solution for startups depends on cash vs predictability. If you have $5K/month to spend and need a warehouse that scales to 100 TB, BigQuery with 100 flex slots (Basic edition) is unbeatable. If you’re bootstrapping and running 10 queries a day, start with on‑demand and move to slots when you hit $2K/month.


How to Calculate and Optimize Your BigQuery Cost

Four steps to BigQuery cost: profile jobs, model on-demand vs slots, trial slots for 30 days, then commit.

You don’t need a finance degree. You need two things: a good query profile and the right reservation strategy.

Step 1: Understand Your Query Profile

Run this query on INFORMATION_SCHEMA to see your top spenders:

SELECT
  project_id,
  user_email,
  SUM(total_bytes_processed) / 1e12 AS total_TB,
  SUM(total_slot_ms) / 1000 / 60 / 60 AS total_slot_hours,
  SUM(total_cost) AS estimated_cost_dollars
FROM `region-us.INFORMATION_SCHEMA.JOBS`
WHERE creation_time > TIMESTAMP_SUB(CURRENT_TIMESTAMP(), INTERVAL 30 DAY)
  AND state = 'DONE'
GROUP BY project_id, user_email
ORDER BY estimated_cost_dollars DESC
LIMIT 10;

This will show you which users and which queries are burning money.

Step 2: Choose Between On‑Demand and Slots

Rule of thumb: if your monthly query TB exceeds 20 TB, slot‑based pricing pays off. But the real threshold depends on your slot consumption. Use the Google Cloud Pricing Calculator to model your workload. Don’t guess.

Here’s a Python script to estimate costs based on job history:

import json
from google.cloud import bigquery

client = bigquery.Client()
query = """
SELECT
  SUM(total_slot_ms) / 1000 / 60 / 60 AS total_slot_hours,
  SUM(total_bytes_processed) / 1e12 AS total_TB,
  COUNT(*) AS query_count
FROM `region-us.INFORMATION_SCHEMA.JOBS`
WHERE creation_time > TIMESTAMP_SUB(CURRENT_TIMESTAMP(), INTERVAL 30 DAY)
  AND state = 'DONE'
"""
job = client.query(query)
result = list(job.result())[0]

total_slot_hours = result["total_slot_hours"]
total_TB = result["total_TB"]
query_count = result["query_count"]

# On‑demand cost
on_demand_cost = total_TB * 6.25

# Slot cost (example: Enterprise 100 slots @ $0.55/hour)
slot_hourly = 0.55
total_slot_cost = total_slot_hours * slot_hourly

print(f"On‑demand: ${on_demand_cost:.2f}")
print(f"Slot‑based (100 flex): ${total_slot_cost:.2f}")

Run this. If slot cost < on‑demand, buy a commitment.

Step 3: Use Capacity Commitments (the real savings)

Google offers three commitment types:

  • Monthly (flex): no long‑term commitment, but no discount over slots.
  • 1‑year: 30% discount on the slot hourly rate.
  • 3‑year: 50% discount.

The catch: you pay for the minimum number of slots every hour, even if you don’t use them. So don’t over‑commit. Start with 100 flex slots, monitor utilisation for 30 days, then lock in a 1‑year.

Here’s a Terraform snippet to create a 1‑year commitment of 200 Enterprise slots in us‑central1:

resource "google_bigquery_capacity_commitment" "commitment" {
  location      = "us-central1"
  plan          = "ANNUAL"
  slot_count    = 200
  edition       = "ENTERPRISE"
}

Note: as of 2026, you can mix commitments across editions (e.g., 100 Basic + 100 Enterprise). That’s useful if your ETL team needs throughput while your analysts need only cheap queries.


GCP Storage Options Comparison: BigQuery vs Cloud Storage vs Spanner vs Bigtable

When people talk about gcp storage options comparison, they often forget that BigQuery is a warehouse, not a general‑purpose data lake. But it integrates tightly with other storage.

Option Latency Cost/GB/month Query cost Use case
BigQuery storage 1–2 seconds $0.02 active / $0.01 long‑term Included in query price Analytics, BI
Cloud Storage (Standard) milliseconds (on hot data) $0.0208 External table query = $0.001 per MB scanned Data lake, backups
Cloud Storage (Nearline) 1 sec retrieval $0.01 Same as above Log archives (30–90 day access)
Cloud Storage (Coldline) 90 sec retrieval $0.004 Same as above Compliance, rarely accessed
Bigtable <10 ms $0.65 per GB (node‑based) Read/write per row Real‑time, high‑throughput events

For startups on a budget, the sweet spot is to store raw data in Cloud Storage (Nearline or Coldline) and load only what you frequently query into BigQuery partitions. This combines cheap storage with fast analytics. It’s the best gcp data warehouse solution for startups when you don’t have unlimited cash.


FAQ

1. Is BigQuery still pay‑per‑query in 2026?

Yes, the on‑demand model still exists. But Google heavily pushes slot‑based pricing. For most production workloads, it’s cheaper to use slots with a capacity commitment.

2. What is the minimum BigQuery slot commitment?

100 slots for any edition. You can buy flex slots (month‑to‑month) or annual/3‑year.

3. How much does a 100‑slot Basic edition cost?

About $0.20/hour per slot, so $0.20 × 100 × 730 hours = $14,600/month. But that covers all query computation – no per‑TB charges.

4. Does BigQuery charge for loading data?

Loading via batch jobs (e.g., from Cloud Storage) is free. Streaming inserts cost $0.05 per 200 MB.

5. Can I use BigQuery with AWS or Azure data?

Yes – BigQuery Omni lets you query data across AWS S3 and Azure Blob without moving it. The pricing is per TB scanned, and you pay egress if you move results out.

6. What’s the cheapest way to store historical data in BigQuery?

Use long‑term storage ($0.01/GB/month). Better yet, export old partitions to Cloud Storage Coldline ($0.004/GB) and create an external table for the rare queries.

7. How do I monitor BigQuery costs in real time?

Use the INFORMATION_SCHEMA tables (as shown above) and set budget alerts in the GCP console. The BigQuery admin panel also tracks slot utilisation.

8. Is BigQuery cheaper than Redshift in 2026?

For most analytic workloads, yes. Redshift’s reserved instances start at ~$0.25/GB/hour, while BigQuery’s per‑slot cost is lower for comparable performance. AWS vs Azure vs GCP Cost Comparison 2026 shows BigQuery 20–30% cheaper for heavy BI workloads.


The Bottom Line

BigQuery bottom line: start on-demand, move to slots when predictable, match edition to needs, cluster and partition.

gcp bigquery pricing 2026 is not a single menu. It’s a family of pricing plans that reward planning. A common mistake is startups jumping into on‑demand because “it’s simple.” That simplicity costs you 2–3x over time.

If you’re a startup, start with on‑demand, but run the Python script above after your first 30 days. The moment your monthly query TB crosses 20, buy a 100‑slot flex commitment. Don’t overthink the edition – Standard is sufficient for 80% of workloads. Upgrade to Enterprise only if you need multi‑cloud or high availability SLAs.

And for the love of your budget: cluster and partition everything. A well‑designed table can cut query costs by 90% compared to a raw SELECT * load.

Moving to the right edition and optimising the schema is where the savings are. The tools are there – you just have to use them.

Want to talk this through?

Sivaro Enterprise →