BigQuery Graph now runs GQL queries on on-demand pricing

55 minutes ago

7 min read

BigQuery Graph now runs GQL queries on on-demand pricing, Data Studios original cover

Google Cloud added an entry to the BigQuery release notes on November 6, 2026 stating that customers can now use BigQuery Graph and run Graph Query Language (GQL) queries when they use on-demand pricing. Until this change, the BigQuery Graph documentation said that running GQL queries required a reservation on the Enterprise or Enterprise Plus edition, and that on-demand projects could reach a graph only through the GRAPH_EXPAND function, which returns a relational view to be queried with SQL.

For analysts and data teams the practical effect is that a project billed per TiB scanned can now model tables as a property graph and query paths with MATCH patterns without first buying slot capacity. On-demand queries are billed for the data they scan, at the rate listed on the BigQuery pricing page, with the first 1 TiB per month free for each account. A team that wants to test fraud rings, customer relationships or supply-chain dependencies on its existing tables can therefore start with the billing model it already has.

The change also comes less than three months after Google Cloud moved in the opposite direction. In August 2026 the release notes announced that core graph processing would be restricted to the Enterprise and Enterprise Plus editions, with on-demand billing and Standard edition support being deprecated. The November 6 entry restores GQL access for on-demand users, while several documentation pages, last updated on October 27, 2026, still describe the reservation requirement.

··········

WHAT GOOGLE CLOUD ANNOUNCED FOR BIGQUERY GRAPH ON NOVEMBER 6.

The release note is short. It is filed as an announcement, it links to the BigQuery Graph overview, and it states that BigQuery Graph and GQL queries can now be used with on-demand pricing. It does not mention a preview label, an allowlist or a regional restriction, and it does not publish a separate price for graph queries. The on-demand rate that applies is therefore the general BigQuery on-demand compute price, which is billed on bytes processed.

BigQuery Graph itself has been generally available since August 31, 2026. It lets users define a property graph over existing BigQuery tables or views, with node tables and edge tables, and query it in GQL, which Google Cloud describes as compatible with the ISO GQL and ISO SQL/PGQ standards and shared with Spanner Graph. Graph definitions do not copy data. According to the documentation, storage is charged only once for the underlying tables, however many graphs are built on top of them.

The release note does not say whether the April 26, 2027 cut-off announced in August for on-demand core graph processing has been withdrawn or still applies. Teams that plan production workloads on on-demand graph queries should treat that date as an open question until Google Cloud updates the release notes or the edition documentation.

··········

BIGQUERY GRAPH BILLING CHANGES IN 2026.

The release notes allow the sequence of decisions to be reconstructed with dates. The table summarizes the entries that concern billing and availability.

........

Date (2026)	Release note entry	Billing effect
April 9	BigQuery Graph in Preview	Graph modeling and analysis at scale
August 20	Deprecation notice	Standard and on-demand core graph processing to end April 26, 2027
August 27	Enforcement announcement	GQL requires an Enterprise or Enterprise Plus reservation; allowlisted users keep older billing until April 26, 2027
August 31	General availability	BigQuery Graph GA, plus CALL and path functions
November 6	Announcement	GQL queries allowed under on-demand pricing

........

Graph measures followed a separate rule. The August entries stated that measures remain available for Enterprise, Enterprise Plus and on-demand queries, but not for Standard edition. Measures are aggregates such as SUM, AVG, COUNT, MIN and MAX declared with MEASURE() in a node or edge table's PROPERTIES clause, and they are read through GRAPH_EXPAND rather than directly from GQL.

The BigQuery editions comparison page, as fetched on November 7, 2026, still lists "No access to BigQuery Graph" for the Standard edition. The November 6 entry concerns on-demand pricing only, so a project that runs its queries on a Standard edition reservation should not assume it has gained GQL access.

··········

HOW GQL QUERIES WORK FOR AN ON-DEMAND BIGQUERY PROJECT.

The documented workflow does not change with the billing model. A user with the BigQuery Data Editor role on the dataset, or a custom role with the bigquery.propertyGraphs permissions, creates a property graph that points at existing tables. The example below follows the structure of the financial graph used in Google Cloud's "Create and query a graph" documentation, with Person and Account nodes and Owns and Transfers edges.

CREATE OR REPLACE PROPERTY GRAPH graph_db.FinGraph
  NODE TABLES (
    graph_db.Account,
    graph_db.Person
  )
  EDGE TABLES (
    graph_db.PersonOwnAccount
      SOURCE KEY (id) REFERENCES Person (id)
      DESTINATION KEY (account_id) REFERENCES Account (id)
      LABEL Owns,
    graph_db.AccountTransferAccount
      SOURCE KEY (id) REFERENCES Account (id)
      DESTINATION KEY (to_id) REFERENCES Account (id)
      LABEL Transfers
  );

A GQL query then starts with the GRAPH clause and describes the path to match. The documentation's example finds the people who received transfers from accounts owned by one person.

GRAPH graph_db.FinGraph
MATCH
  (person:Person {name: "Dana"})-[own:Owns]->
  (account:Account)-[transfer:Transfers]->(account2:Account)<-[own2:Owns]-(person2:Person)
RETURN
  person.name AS owner,
  transfer.amount AS amount,
  person2.name AS transferred_to
ORDER BY person2.name;

Before November 6, an on-demand project could not run this second statement. Results can be shown as a graph in BigQuery Studio, and BigQuery notebooks accept the %%bigquery --graph magic for visual output. Graph data can also feed conversational analytics, which Google Cloud brought to spreadsheets with conversational analytics in Connected Sheets.

··········

ON-DEMAND VS ENTERPRISE RESERVATION COSTS FOR GRAPH QUERIES.

On the BigQuery pricing page, viewed on November 7, 2026, on-demand compute is listed at $6.25 per TiB, with the first 1 TiB per month free. Charges are rounded up to the nearest MB, with a minimum of 10 MB processed per table referenced and per query. Capacity pricing is listed per slot-hour: $0.04 for Standard, $0.06 for Enterprise and $0.10 for Enterprise Plus on pay-as-you-go, billed per second with a one-minute minimum. The page presents the slot prices with a location selector and does not state which location the default figures describe, so buyers outside the US multi-region should check the rate for their own region.

The illustrative example below compares one month for a small analytics team. It assumes 4 TiB scanned by all on-demand queries in the account, and, for the alternative, 100 Enterprise slots kept active for 4 hours on each of 22 working days. The figures were computed in Python and ignore storage, which is billed the same way under both models.

........

Scenario (illustrative)	Calculation	Monthly compute cost
On-demand, 4 TiB scanned	(4 − 1 free) × $6.25	$18.75
On-demand, 20 TiB scanned	(20 − 1 free) × $6.25	$118.75
Enterprise, 100 slots × 88 hours	8,800 slot-hours × $0.06	$528.00
Enterprise Plus, 100 slots × 88 hours	8,800 slot-hours × $0.10	$880.00

........

At these list prices, the 88-hour Enterprise block costs the same as about 84.5 TiB of billed on-demand scanning, or 85.5 TiB including the free tier. Below that volume, on-demand is cheaper for this usage pattern; above it, or when queries run around the clock, a reservation becomes competitive. Real reservations use baselines, autoscaling and optional commitments, so the slot side of the comparison should be rebuilt with the team's actual configuration.

Graph queries can scan more data than their result suggests, because multi-hop patterns read node and edge tables repeatedly. Google Cloud's graph query best practices recommend starting traversal from low-cardinality nodes, specifying labels for every node and edge, preferring directional patterns, using a single MATCH statement where possible and using ANY or ANY SHORTEST when only connectivity is needed. Under on-demand billing these choices affect the invoice directly, in the same way that partitioning and clustering do for SQL queries.

··········

WHAT BIGQUERY ADMINS SHOULD CHECK BEFORE MOVING GRAPH WORK TO ON-DEMAND.

The first check is the billing model of each project that will run graph queries. A project assigned to an Enterprise or Enterprise Plus reservation keeps using slots; the November 6 change concerns projects that are not assigned to a reservation and pay per TiB.

The second check is cost control. On-demand users can set a maximum bytes billed value on queries and custom quotas on daily usage, which cap the effect of an inefficient path pattern. A first round of graph queries on a test dataset, with the bytes processed read from the job details, gives a realistic basis for the monthly estimate before the workload moves to production tables.

The third check concerns measures. GQL statements cannot reference measure properties directly; GRAPH_EXPAND returns the relational view in which measures are aggregated once per key. The documentation warns that changes to a graph do not invalidate cached GRAPH_EXPAND results, so cached results should be disabled when the graph definition is edited.

Teams already running AI functions on BigQuery data, such as AI.PARSE_DOCUMENT for PDF parsing, or querying data in other clouds through cross-cloud caching for S3 and Azure, can combine those tables with graph definitions, since a graph is built over ordinary tables and views. Edition requirements for each of those features remain separate and should be checked one by one.

··········

BIGQUERY GRAPH ON-DEMAND PRICING: CONFIRMED FACTS AND OPEN POINTS.

Google Cloud's release notes confirm that, as of November 6, 2026, BigQuery Graph and GQL queries can be used with on-demand pricing. BigQuery Graph has been generally available since August 31, 2026, graph storage is billed once on the underlying tables, and on-demand compute is listed at $6.25 per TiB after the first free TiB each month.

Several points depend on the customer's setup: the slot price for the chosen region, whether a project is assigned to a reservation, and whether it runs on Standard edition, which the editions page still lists without access to BigQuery Graph. Google Cloud has not stated whether the April 26, 2027 restriction announced in August still applies to on-demand billing, and the graph overview and measures pages had not yet been updated to reflect the change when they were checked on November 7, 2026.

··········

·····

DATA STUDIOS

·····

datastudios.org