Snowflake (SNOW)
Runs analytics, applications, transactions, and AI agents across governed data.
Something off on this page? Send us feedback.
Snowflake rents companies one managed online place to store, prepare, study and share their data. Its mature analytics tools still pay most of the bills; data-moving tools, AI helpers and databases for live apps are the push beyond that core. Customers commit money ahead of time, but Snowflake earns it only when they use the platform.
Item facts: FY2026 · year ended Jan 31, 2026, from filings, earnings calls and company pages.
Judgment weights, not filed revenue
The band summarizes business focus and direction. ~ marks estimates.
8 in detail · 8 more below
Snowflake for Analytics
The mature engine, worth roughly three-fifths of FY2026 revenue. Customers pay whenever their teams store and examine data; the watchpoint is whether newer uses grow without weakening this dependable core.
Competes with Databricks SQL (Databricks) · BigQuery (Google Cloud) · Amazon Redshift (AWS)
In plain English
Picture a vast spreadsheet that never lives on anyone's laptop. A company pours its sales, inventory or patient data into Snowflake, then employees ask questions such as what sold yesterday or where supplies are running low. Snowflake keeps the machinery behind those searches out of sight.
The bill rises with the amount stored and the computing power used to answer questions. That makes frequent searches valuable: a signed contract is only a promise, while actual use turns into Snowflake's revenue. This is the old, unglamorous engine that funds the newer bets.
Snowflake for Data Engineering
Roughly a fifth of FY2026 revenue comes from moving and cleaning data before anyone studies it. Openflow adds ready-made connections; lasting growth depends on those data routes feeding work people keep doing.
Competes with Lakeflow (Databricks) · Data Factory and Fabric Data Engineering (Microsoft)
In plain English
Before data can answer a question, someone has to collect it from many systems, straighten it out and keep it arriving. Snowflake's engineering tools do that assembly-line work: they pull information in, reshape it and pass a clean version onward to analysts, apps and AI helpers. Openflow adds managed connections to outside systems.
Companies pay as these jobs use storage and computing power. More sources and more frequent updates mean more use, but an idle route earns little. The business grows when a newly built route becomes the plumbing for something employees or applications depend on every day.
Snowflake for Applications and Collaboration
About one-eighth of FY2026 revenue is tied to apps and controlled data exchange. Outside providers booked over $100M through the marketplace in the first half of 2026, but Snowflake earns from the resulting platform use, not those sales.
Competes with Databricks Apps and Marketplace (Databricks) · Fabric and OneLake (Microsoft)
In plain English
A guarded meeting room is the useful picture. Two companies can work with the same information without either one carrying away an uncontrolled copy, while software makers can place an app beside the customer's data instead of hauling the data elsewhere. Snowflake also runs a marketplace that helps buyers find those apps and data sets.
Money arrives when all this activity consumes Snowflake storage and computing power. The providers' sales are their money, not Snowflake's. Snowflake wins only if a useful selection attracts buyers, those buyers return, and the sharing rules are trusted enough for companies to keep sensitive information there.
Snowflake AI
AI tools reached roughly a $100M yearly revenue pace late in FY2026, still tiny beside the core. CoCo helps programmers and CoWork helps office workers; the test is repeated paid use at sustainable delivery cost.
Competes with Agent Bricks and Genie (Databricks) · Copilot in Microsoft Fabric (Microsoft) · BigQuery AI (Google Cloud)
In plain English
Here the software does more than fetch a table: it drafts an answer or helps write computer code using information the company already keeps in Snowflake. CoCo is the helper for programmers, while CoWork is aimed at people doing ordinary office research and analysis.
Each request burns Snowflake credits, the prepaid units customers use across the platform. Snowflake then pays outside companies for answering software and computing power. That creates revenue quickly when people experiment, but the real prize is a habit: useful requests repeated often enough that the customer keeps buying credits and Snowflake can deliver them economically.
Transactions (Unistore and Snowflake Postgres)
A very small FY2026 business extends Snowflake into databases that update live apps, not just study the past. Snowflake Postgres arrived after year-end; winning depends on proving it reliable enough for everyday reads and writes.
Competes with Lakebase (Databricks) · Amazon Aurora (AWS) · Fabric Databases (Microsoft)
In plain English
Think of the difference between studying yesterday's shop receipts and running today's checkout. Analytics handles the first job; these databases aim to handle both. Unistore lets an app change records and examine them in one place, while Snowflake Postgres offers a familiar kind of database used by many application builders.
The customer pays as live reads, writes and stored copies consume the platform. Once a checkout, booking system or other working app depends on a database, moving it is painful, so reliability and speed matter more than novelty. For now this is a foothold, not a material revenue engine.
Observe
Bought just after FY2026, Observe is expected to add about one point to FY2027 product growth. It searches the exhaust from software systems for failures; the watchpoint is whether the answers justify the heavy data-handling cost.
Competes with Observability Platform (Datadog) · Dynatrace platform (Dynatrace)
In plain English
When a company website slows or breaks, its machines leave behind a blizzard of status notes and measurements. Observe gathers that evidence and helps the people who keep systems working trace what went wrong, much as a mechanic follows drips and rattles back to the failing part. It runs on Snowflake while remaining a product customers can buy directly.
More complicated systems produce more evidence, so storing and searching it creates Snowflake use. The danger is the same abundance: machine records can multiply faster than budgets. Customers keep paying only when Observe shortens costly outages enough to make all that collection worthwhile.
Capacity Agreements
At July 31, 2026, customers' remaining promised use had grown by almost a third, with just over half expected within a year. Snowflake gets paid only as customers actually use that capacity.
Competes with Databricks SQL (Databricks) · Microsoft Fabric and OneLake (Microsoft) · BigQuery (Google Cloud)
In plain English
Most customers reserve a pool of Snowflake use for the next one to four years. It resembles buying a thick book of ride tickets: signing the deal fills the book, but Snowflake records a sale only as the tickets are torn out for computing, storage or moving data.
Large renewals show that major customers intend to stay, yet promises do not guarantee a steady pace. A customer can speed up, slow down or leave capacity unused. The growing pool gives Snowflake visibility; actual use decides when that pool becomes money.
Professional Services and Other
Consulting, technical help and training brought in $211.6M in FY2026 but cost more to deliver than customers paid. Its job is to get customers using the platform, not to become the profit engine.
Competes with AWS Professional Services (Amazon) · Forward Deployed Engineering (Databricks)
In plain English
The unglamorous setup crew helps a new customer move old data, train employees and get its first Snowflake jobs working. Think of buying a complex kitchen: the installer is useful because the appliances start earning their keep sooner, even if installation itself is not the lucrative part.
Customers pay for Snowflake staff and partners to guide the move. Delivery requires people, and in FY2026 that work cost Snowflake more than it brought in. The payoff comes later if a smooth setup leads the customer to run more analytics, data routes, apps or AI requests on the platform.
Snowflake for AnalyticsThe mature engine, worth roughly three-fifths of FY2026 revenue. Customers pay whenever their teams store and examine data; the watchpoint is whether newer uses grow without weakening this dependable core.The mature engine, worth roughly three-fifths of FY2026 revenue. Customers pay whenever their teams store and examine data; the watchpoint is whether newer uses grow without weakening this dependable core.
In plain English
Picture a vast spreadsheet that never lives on anyone's laptop. A company pours its sales, inventory or patient data into Snowflake, then employees ask questions such as what sold yesterday or where supplies are running low. Snowflake keeps the machinery behind those searches out of sight.
The bill rises with the amount stored and the computing power used to answer questions. That makes frequent searches valuable: a signed contract is only a promise, while actual use turns into Snowflake's revenue. This is the old, unglamorous engine that funds the newer bets.
Competes with Databricks SQL (Databricks) · BigQuery (Google Cloud) · Amazon Redshift (AWS)
Snowflake for Data EngineeringRoughly a fifth of FY2026 revenue comes from moving and cleaning data before anyone studies it. Openflow adds ready-made connections; lasting growth depends on those data routes feeding work people keep doing.Roughly a fifth of FY2026 revenue comes from moving and cleaning data before anyone studies it. Openflow adds ready-made connections; lasting growth depends on those data routes feeding work people keep doing.
In plain English
Before data can answer a question, someone has to collect it from many systems, straighten it out and keep it arriving. Snowflake's engineering tools do that assembly-line work: they pull information in, reshape it and pass a clean version onward to analysts, apps and AI helpers. Openflow adds managed connections to outside systems.
Companies pay as these jobs use storage and computing power. More sources and more frequent updates mean more use, but an idle route earns little. The business grows when a newly built route becomes the plumbing for something employees or applications depend on every day.
Competes with Lakeflow (Databricks) · Data Factory and Fabric Data Engineering (Microsoft)
Snowflake for Applications and CollaborationAbout one-eighth of FY2026 revenue is tied to apps and controlled data exchange. Outside providers booked over $100M through the marketplace in the first half of 2026, but Snowflake earns from the resulting platform use, not those sales.About one-eighth of FY2026 revenue is tied to apps and controlled data exchange. Outside providers booked over $100M through the marketplace in the first half of 2026, but Snowflake earns from the resulting platform use, not those sales.
In plain English
A guarded meeting room is the useful picture. Two companies can work with the same information without either one carrying away an uncontrolled copy, while software makers can place an app beside the customer's data instead of hauling the data elsewhere. Snowflake also runs a marketplace that helps buyers find those apps and data sets.
Money arrives when all this activity consumes Snowflake storage and computing power. The providers' sales are their money, not Snowflake's. Snowflake wins only if a useful selection attracts buyers, those buyers return, and the sharing rules are trusted enough for companies to keep sensitive information there.
Competes with Databricks Apps and Marketplace (Databricks) · Fabric and OneLake (Microsoft)
Snowflake AIAI tools reached roughly a $100M yearly revenue pace late in FY2026, still tiny beside the core. CoCo helps programmers and CoWork helps office workers; the test is repeated paid use at sustainable delivery cost.AI tools reached roughly a $100M yearly revenue pace late in FY2026, still tiny beside the core. CoCo helps programmers and CoWork helps office workers; the test is repeated paid use at sustainable delivery cost.
In plain English
Here the software does more than fetch a table: it drafts an answer or helps write computer code using information the company already keeps in Snowflake. CoCo is the helper for programmers, while CoWork is aimed at people doing ordinary office research and analysis.
Each request burns Snowflake credits, the prepaid units customers use across the platform. Snowflake then pays outside companies for answering software and computing power. That creates revenue quickly when people experiment, but the real prize is a habit: useful requests repeated often enough that the customer keeps buying credits and Snowflake can deliver them economically.
Competes with Agent Bricks and Genie (Databricks) · Copilot in Microsoft Fabric (Microsoft) · BigQuery AI (Google Cloud)
Transactions (Unistore and Snowflake Postgres)A very small FY2026 business extends Snowflake into databases that update live apps, not just study the past. Snowflake Postgres arrived after year-end; winning depends on proving it reliable enough for everyday reads and writes.A very small FY2026 business extends Snowflake into databases that update live apps, not just study the past. Snowflake Postgres arrived after year-end; winning depends on proving it reliable enough for everyday reads and writes.
In plain English
Think of the difference between studying yesterday's shop receipts and running today's checkout. Analytics handles the first job; these databases aim to handle both. Unistore lets an app change records and examine them in one place, while Snowflake Postgres offers a familiar kind of database used by many application builders.
The customer pays as live reads, writes and stored copies consume the platform. Once a checkout, booking system or other working app depends on a database, moving it is painful, so reliability and speed matter more than novelty. For now this is a foothold, not a material revenue engine.
Competes with Lakebase (Databricks) · Amazon Aurora (AWS) · Fabric Databases (Microsoft)
ObserveBought just after FY2026, Observe is expected to add about one point to FY2027 product growth. It searches the exhaust from software systems for failures; the watchpoint is whether the answers justify the heavy data-handling cost.Bought just after FY2026, Observe is expected to add about one point to FY2027 product growth. It searches the exhaust from software systems for failures; the watchpoint is whether the answers justify the heavy data-handling cost.
In plain English
When a company website slows or breaks, its machines leave behind a blizzard of status notes and measurements. Observe gathers that evidence and helps the people who keep systems working trace what went wrong, much as a mechanic follows drips and rattles back to the failing part. It runs on Snowflake while remaining a product customers can buy directly.
More complicated systems produce more evidence, so storing and searching it creates Snowflake use. The danger is the same abundance: machine records can multiply faster than budgets. Customers keep paying only when Observe shortens costly outages enough to make all that collection worthwhile.
Competes with Observability Platform (Datadog) · Dynatrace platform (Dynatrace)
Capacity AgreementsAt July 31, 2026, customers' remaining promised use had grown by almost a third, with just over half expected within a year. Snowflake gets paid only as customers actually use that capacity.At July 31, 2026, customers' remaining promised use had grown by almost a third, with just over half expected within a year. Snowflake gets paid only as customers actually use that capacity.
In plain English
Most customers reserve a pool of Snowflake use for the next one to four years. It resembles buying a thick book of ride tickets: signing the deal fills the book, but Snowflake records a sale only as the tickets are torn out for computing, storage or moving data.
Large renewals show that major customers intend to stay, yet promises do not guarantee a steady pace. A customer can speed up, slow down or leave capacity unused. The growing pool gives Snowflake visibility; actual use decides when that pool becomes money.
Competes with Databricks SQL (Databricks) · Microsoft Fabric and OneLake (Microsoft) · BigQuery (Google Cloud)
Professional Services and OtherConsulting, technical help and training brought in $211.6M in FY2026 but cost more to deliver than customers paid. Its job is to get customers using the platform, not to become the profit engine.Consulting, technical help and training brought in $211.6M in FY2026 but cost more to deliver than customers paid. Its job is to get customers using the platform, not to become the profit engine.
In plain English
The unglamorous setup crew helps a new customer move old data, train employees and get its first Snowflake jobs working. Think of buying a complex kitchen: the installer is useful because the appliances start earning their keep sooner, even if installation itself is not the lucrative part.
Customers pay for Snowflake staff and partners to guide the move. Delivery requires people, and in FY2026 that work cost Snowflake more than it brought in. The payoff comes later if a smooth setup leads the customer to run more analytics, data routes, apps or AI requests on the platform.
Competes with AWS Professional Services (Amazon) · Forward Deployed Engineering (Databricks)
Named in filings, launches and programs
- Snowflake Horizon CatalogEcosystemOne place to find data and apply access rules across Snowflake and outside stores.
- Snowflake Open CatalogPlatformA managed index for Iceberg tables, an open way of organizing large data files so different tools can use them.
- Snowflake OpenflowService · RampingManaged connections that pull data from outside systems into the data-engineering tools.
- Snowpark Container ServicesPlatformRuns packaged software and AI models beside customer data without sending that data elsewhere.
- Snowflake AI Data Cloud for IndustriesProduct lineVersions of the platform packaged for fields such as finance, healthcare, retail and government rather than a separate business division.
- Powered by Snowflake Startup ProgramEcosystemSupport and free platform use for startups whose apps may later bring lasting customer activity.
- Select Star technologyProduct · AnnouncedPlanned data-finding technology for Horizon Catalog; its purchase was announced but not confirmed complete.
- Natoma / Cortex AI Gateway integrationProduct · AnnouncedPlanned secure connections for AI helpers; integration was discussed, but the purchase was not separately confirmed complete.
Snowflake Horizon CatalogEcosystem
One place to find data and apply access rules across Snowflake and outside stores.
Snowflake Open CatalogPlatform
A managed index for Iceberg tables, an open way of organizing large data files so different tools can use them.
Snowflake OpenflowService · Ramping
Managed connections that pull data from outside systems into the data-engineering tools.
Snowpark Container ServicesPlatform
Runs packaged software and AI models beside customer data without sending that data elsewhere.
Snowflake AI Data Cloud for IndustriesProduct line
Versions of the platform packaged for fields such as finance, healthcare, retail and government rather than a separate business division.
Powered by Snowflake Startup ProgramEcosystem
Support and free platform use for startups whose apps may later bring lasting customer activity.
Select Star technologyProduct · Announced
Planned data-finding technology for Horizon Catalog; its purchase was announced but not confirmed complete.
Natoma / Cortex AI Gateway integrationProduct · Announced
Planned secure connections for AI helpers; integration was discussed, but the purchase was not separately confirmed complete.