Snowflake vs Databricks vs BigQuery pricing 2027: credit, DBU and TiB costs

Summary. Snowflake's Service Consumption Table, reissued effective 9 November 2026, prices on-demand Platform Credits at $2.00 (Standard), $3.00 (Enterprise) and $4.00 (Business Critical) in AWS US East and at $2.70, $4.00 and $5.40 in the London regions of AWS, Azure and Google Cloud. AI features bill separately in AI Credits at $2.00 (Global) or $2.20 (Regional), outside your Platform Credit discount. Databricks on AWS (Premium) charges $0.70 per DBU for serverless SQL in US East and $0.95 in London, $0.15 per DBU for classic Jobs plus your own EC2 bill, and $0.35 or $0.39 per DBU for serverless Jobs. BigQuery on-demand costs $6.25 per TiB scanned in the US multi-region and $7.8125 in London (europe-west2), with the first 1 TiB free each month; Enterprise edition slots cost $0.06 and $0.078 per slot-hour. In our worked example for a 250-person company holding 10 TB, list prices run from about $866 a month (BigQuery on-demand, US) to $6,266 (Databricks serverless SQL, London). Our view: choose on workload fit, then budget roughly a 30% premium (27% to 34% in our example) if the data must stay in London.

About this article. For finance, data and platform leads in the US and UK setting a 2027 data platform budget or renewal. On 12 November 2026 we checked Snowflake's current consumption table against archived earlier versions, Snowflake's cost and AI inference documentation, the Databricks pricing pages and the price data behind them, Databricks documentation, and Google's BigQuery pricing page and release notes. All prices are list prices in US dollars, before tax and negotiated discounts, and vendors change them often. This is general information, not financial advice, and no vendor paid for inclusion. Researched with AI assistance and reviewed by our editorial team.

What changed for 2027 data platform budgets

Three things matter. First, Snowflake now bills AI on its own meter. The table splits compute into Platform Credits (warehouses, Snowpark Container Services, Openflow and Postgres compute) and AI Credits (Cortex AI functions, agents, search and other AI features). AI Credits first appear in the table effective 1 April 2026, when Cortex AI functions were still priced in ordinary credits; the 2 March 2026 version has no AI Credits. By the table effective 10 June 2026, functions such as AI_COMPLETE were priced in AI Credits per million tokens. The table now says an order form's Snowflake Credits or Credits means Platform Credits, and that the Platform Credit discount does not apply to AI Credits. AI Credits have their own capacity tiers: the first discount starts at $1.2 million of annual contract value, and the floor is $1.88 (Global) and $2.07 (Regional) at $40 million and above. The 9 November reissue only added AI model rows to the 4 November version; credit and storage prices did not move. If your 2026 contract assumed one pool of discounted credits for everything, check how AI usage is billed at renewal.

Second, the London premium is large and uneven. Snowflake charges 33% to 35% more per credit in London than in US East. Databricks serverless SQL costs 36% more in London, while classic SQL and classic Jobs rates are the same in both regions. BigQuery on-demand is 25% higher in London and edition slots are 30% higher.

Third, Databricks is running a promotion on its Lakehouse RT warehouse (Beta) until 31 January 2027. It bills 30% fewer DBUs while the price per DBU stays the same, so a workload that would use 10 DBUs in an hour is billed for 7. Budget the full DBU count for anything that runs after that date. Our Lakehouse RT comparison covers when it can replace a separate real-time database.

Snowflake: Platform Credits, AI Credits and storage

Snowflake bills compute in credits per hour by warehouse size: a standard X-Small uses 1 credit per hour, Small 2, Medium 4, Large 8 and X-Large 16. Gen 2 warehouses use 1.35 times those rates on AWS and Google Cloud and 1.25 times on Azure. Billing is per second after a 1-minute minimum each time a warehouse starts or resumes; Interactive warehouses carry a 60-minute minimum.

The cloud services layer (query compilation, metadata, authentication) costs 4.4 credits per hour, but Snowflake waives it on any day it stays at or under 10% of that day's warehouse credits. Serverless features such as Snowpipe, tasks and search optimisation bill separately.

Snowflake's on-demand price per credit by region and edition:

AI Credits cost $2.00 (Global) or $2.20 (Regional) on demand whatever your edition, and Cortex Inference may not qualify for capacity AI Credit pricing.

Storage is priced per compressed terabyte per month: $23.00 in AWS US East, $24.00 in AWS London and Azure UK South, and $23.00 in GCP London. Capacity orders that Snowflake designates as qualifying and that exceed $1.2 million a year get lower storage tiers on AWS and Azure; GCP London stays at $23.00. Data transfer to the internet or another cloud costs $90 per TB from both AWS US East and AWS London, and $20 per TB to another region on the same cloud.

As an AI example, AI_COMPLETE with claude-sonnet-4-6 costs 1.80 AI Credits per million input tokens and 9.00 per million output tokens. The table does not define the Global and Regional rates; the contract adviser Deal IQ describes $2.00 as the global-routing rate and $2.20 as the rate when requests are pinned to the home region. On that reading, keeping inference in-region costs 10% more per AI Credit before any discount. Our Genie versus Cortex Analyst cost comparison works through AI Credit costs for one use case.

Best for: SQL-first analytics teams with many dashboard users, bursty workloads that suit per-second billing and auto-suspend, and organisations that want the same platform on AWS, Azure or Google Cloud.

Databricks: DBUs by workload, plan and region

Databricks bills in Databricks Units (DBUs). The price per DBU depends on the workload (SQL, Jobs, pipelines, all-purpose notebooks), the plan (Premium or Enterprise on AWS and Google Cloud; Azure Databricks Premium corresponds to Enterprise elsewhere) and whether compute is serverless or classic. Serverless rates include the cloud instance cost. Classic rates do not: you also pay AWS, Azure or Google for the virtual machines in your account.

Databricks on AWS, Premium plan, list price per DBU on 12 November 2026:

On the Enterprise plan, classic Jobs rises to $0.20 per DBU and all-purpose compute from $0.55 to $0.65, while serverless SQL stays at $0.70 and $0.95. Classic declarative pipelines cost $0.20 (Core), $0.25 (Pro) and $0.36 (Advanced) per DBU. Serverless Jobs run in a Performance Optimized or Standard mode; Databricks says Standard can be up to 70% cheaper for some workloads, with slower start-up. Our note on the Databricks dbt performance mode change explains that trade.

Serverless SQL warehouses consume DBUs by size. Databricks' calculator data lists 4 DBUs per hour for 2X-Small, 6 for X-Small, 12 for Small, 24 for Medium, 40 for Large and 80 for X-Large. A Medium warehouse therefore costs $16.80 per hour in US East (24 × $0.70) and $22.80 in London (24 × $0.95). Serverless warehouses add clusters of that size when queries queue, so a busy warehouse can bill a multiple of these figures.

Default storage is billed in Databricks Storage Units: 1 DSU per GB-month, where Databricks counts a GB as 2^30 bytes, plus DSUs for operations (on AWS, 0.2174 DSU per 1,000 write-class requests and 0.0174 per 1,000 read-class requests). Tables kept in your own S3 bucket appear on your AWS bill instead.

Best for: teams with heavy Python, Spark or ML pipelines, a preference for open table formats in their own cloud account, and engineers who will tune classic job clusters (including spot instances).

BigQuery: on-demand TiB, edition slots and storage

BigQuery offers two compute models. On-demand charges per TiB scanned, with the first 1 TiB per month free. Capacity pricing (BigQuery editions) charges per slot-hour, billed per second with a 1-minute minimum unless you opt in to fluid scaling, and can be combined with 1-year or 3-year commitments. Fluid scaling, which removes the minimum for autoscaling reservations, became generally available on 3 June 2026.

BigQuery compute list prices in the US multi-region and London:

Storage is priced per GiB per month, with the first 10 GiB free. You choose logical (uncompressed) or physical (compressed) billing per dataset, and tables or partitions unmodified for 90 consecutive days move to the long-term rate.

Google states that if you pay in a currency other than US dollars, the prices listed in your currency on its Cloud Platform SKUs apply.

Best for: teams already on Google Cloud, spiky ad hoc querying, and organisations that do not want to size or suspend clusters. On-demand pricing rewards partitioned and clustered tables; a careless full-table scan is billed in full.

Head-to-head: the numbers that drive the bill

The table summarises the list prices above.

Which platform fits: a decision table

Worked example: a mid-size analytics bill in US East and London

The company: a 250-person B2B software firm with a six-person data team and 10 TB of compressed data. Its BI workload runs 10 hours a day on 22 working days (220 hours a month). Its nightly ELT run takes 3 hours a day, every day (90 hours a month).

Our assumptions: Snowflake Enterprise runs a Medium warehouse for BI and a Small one for ELT. Databricks runs a Medium serverless SQL warehouse for BI and a Small one for dbt-style ELT, each on a single cluster. On BigQuery, on-demand queries scan 80 TiB a month; under Enterprise edition they average 100 slots across the same 310 hours. Snowflake's cloud services stay within the 10% daily allowance, and all storage is active (the worst case). These are budgeting assumptions: warehouse sizes are not equivalent across vendors, and we did not benchmark the platforms. For build and staffing costs, see our data engineering cost guide.

Snowflake Enterprise. BI: 4 credits × 220 hours = 880 credits. ELT: 2 credits × 90 hours = 180 credits. Total 1,060 credits.

• US East: 1,060 × $3.00 = $3,180 compute + 10 TB × $23 = $230 storage = $3,410.

• London (AWS): 1,060 × $4.00 = $4,240 + 10 × $24 = $240 = $4,480.

• On Standard edition the same usage costs $2,350 in US East (1,060 × $2.00 + $230) and $3,102 in London (1,060 × $2.70 + $240).

Databricks serverless SQL (Premium). BI: 24 DBU × 220 hours = 5,280 DBU. ELT: 12 DBU × 90 hours = 1,080 DBU. Total 6,360 DBU. Storage: 10 TB is about 9,313 binary GB, so 9,313 DSU.

• US East: 6,360 × $0.70 = $4,452 + 9,313 × $0.023 = $214.20 = $4,666.20, before storage operation charges.

• London: 6,360 × $0.95 = $6,042 + 9,313 × $0.024 = $223.51 = $6,265.51.

BigQuery on-demand. Billable scan: 80 TiB minus 1 free = 79 TiB. Storage: 10 TB is about 9,313 GiB, less 10 free = 9,303 GiB of active physical storage.

• US: 79 × $6.25 = $493.75 + 9,303 × $0.04 = $372.12 = $865.87.

• London: 79 × $7.8125 = $617.19 + 9,303 × $0.052 = $483.76 = $1,100.95.

BigQuery Enterprise edition. 100 slots × 310 hours = 31,000 slot-hours.

• US: 31,000 × $0.06 = $1,860 + $372.12 = $2,232.12.

• London: 31,000 × $0.078 = $2,418 + $483.76 = $2,901.76.

BigQuery's crossover here is about 300 TiB. In the US, $1,860 of slot-hours buys 297.6 billable TiB on demand ($1,860 ÷ $6.25), so with the free TiB the crossover is about 299 TiB a month. In London it is about 310 TiB ($2,418 ÷ $7.8125 = 309.5, plus the free TiB). The crossover moves with your slot-hours, and dashboards that rescan wide tables every 15 minutes can pass 300 TiB quickly.

Snowflake and Databricks hinge on hours, not data. Both bills are almost entirely compute. Halving BI hours through auto-suspend or caching cuts the US Snowflake Enterprise bill by about 39% (to $2,090) and the Databricks bill by about 40% (to $2,818.20). Storage is under 7% of either total.

The London premium is close to a third. A UK-resident deployment costs 27% to 34% more for the same usage in this scenario. If the data need not stay in the UK, weigh that premium against the compliance work a US or EU region would create.

US and UK notes: regions, currency, tax and residency

United States. Mainstream US regions carry the base price on all three platforms. Government regions cost more: Snowflake's AWS US East 1 Commercial Gov region starts at $4.80 per credit (Business Critical only), and AWS US Gov West 1 at $5.60.

United Kingdom. All three platforms offer London: Snowflake on AWS, Azure UK South and GCP Europe West 2; Databricks on AWS Europe (London) and Azure UK South; and BigQuery in europe-west2. Snowflake and Databricks publish London prices only in US dollars, and none of the three publishes a sterling price list; BigQuery bills a sterling billing account at Google's sterling SKU prices. Snowflake's table makes the customer responsible for taxes such as VAT on top of its prices; treat the Databricks and Google figures as pre-tax too, and check how VAT appears on each invoice.

Data residency. Keeping the warehouse in London does not keep every byte in the UK. Snowflake's cross-region inference setting decides where AI requests run. New accounts in organisations created after 9 March 2026 default to any commercial region. Other commercial accounts that have not set it default to a same-cloud geography: AWS accounts in London get AWS_EU and Azure accounts in the UK get AZURE_EU, so requests can be processed in EU regions such as Frankfurt, Ireland or Paris as well as London. No option is limited to the UK, so keeping inference in London means disabling cross-region inference, which limits the models available. Snowflake says stored data stays in the account's region and prompts and responses move only transiently. Also check where replicas, failover accounts and support access sit.

Cost controls to set before the first invoice

1. Tag everything. Snowflake query tags and warehouse names, Databricks custom tags read through the billing system tables, and BigQuery labels on jobs and datasets.

1. Set suspend timers. Short auto-suspend on Snowflake BI warehouses, short auto-stop on Databricks serverless warehouses, and no always-on Interactive warehouses unless you need them.

1. Cap scans. In BigQuery, partition and cluster large tables and set a maximum bytes billed per query, or custom cost controls per user or project.

1. Price AI separately. Track Snowflake AI Credits as their own budget line, decide whether you need Regional or Global processing, and negotiate AI Credit pricing explicitly.

1. Avoid promotional anchoring. Model Databricks Lakehouse RT at the full DBU count, since the 30% reduction ends on 31 January 2027.

1. Measure, then commit. Run a 2 to 4 week pilot on real queries, read consumption from each vendor's billing views, and only then sign a Snowflake capacity deal, Databricks commitment or BigQuery CUD.

1. Plan the exit. Snowflake charges $90 per TB to the internet from AWS US East and London, so moving 10 TB out once costs $900 before any cloud provider fees. If any workloads serve EU customers, read our EU Data Act switching guide first.

How eCorpIT can help

eCorpIT is a Gurugram-based technology organisation, founded in 2021, assessed at CMMI Level 5 and MSME certified, with senior-led engineering teams working across AWS, Microsoft and Google platforms. Our data and analytics team turns your query history into credit, DBU and TiB forecasts, runs a short pilot on your own workloads, and sets up tagging, suspend policies and scan limits before you commit. Talk to us at /contact-us/.

Last updated: 12 November 2026.

References

1. Snowflake Service Consumption Table, effective 9 November 2026 — Snowflake

1. Databricks pricing card data (cards.json), read 12 November 2026 — Databricks

1. BigQuery pricing — Google Cloud

1. Snowflake Service Consumption Table, effective 1 April 2026 (archived copy) — Snowflake via Internet Archive

1. Snowflake Service Consumption Table, effective 2 March 2026 (archived copy) — Snowflake via Internet Archive

1. Snowflake Service Consumption Table, effective 10 June 2026 (archived copy) — Snowflake via Internet Archive

1. Snowflake Service Consumption Table, effective 4 November 2026 (archived copy) — Snowflake via Internet Archive

1. Databricks Lakehouse pricing — Databricks

1. Understanding compute cost — Snowflake Documentation

1. Snowflake pricing guide — Snowflake

1. Snowflake AI Credits and Cortex pricing explained, updated September 2026 — Deal IQ

1. Lakeflow Jobs pricing — Databricks

1. Databricks pricing calculator data for AWS (AWS.json) — Databricks

1. SQL warehouse sizing, scaling and queuing behaviour, updated 5 November 2026 — Databricks Documentation

1. Databricks storage pricing — Databricks

1. BigQuery release notes — Google Cloud

1. Parameters reference, CORTEX_ENABLED_CROSS_REGION — Snowflake Documentation

1. Cross-region inference — Snowflake Documentation