How Snowflake Makes its Money: Revenue Breakdown

How Does Snowflake Make its Money?

Snowflake Inc. (NYSE: SNOW) is a cloud-native data platform that enables organisations to store, query, share, and process large volumes of structured and semi-structured data across any cloud. The company generated $3.65 billion in total revenue for fiscal year 2025 (ending January 2025), up 27.6% from $2.86 billion in FY2024, with a net loss of $0.83 billion on a GAAP basis but a non-GAAP operating margin of approximately 7% (excluding ~$1.5B in stock-based compensation). Snowflake’s gross margin on product revenue is 73.5% — reflecting the economics of a cloud software platform layered over cloud infrastructure costs.

Snowflake’s business is almost entirely one product line: Product Revenue ($3.43B, 94% of total) from its consumption-based data platform, with the remaining 6% from Professional Services. The defining characteristic of Snowflake’s revenue model is that it is not a subscription business — customers do not pay a fixed monthly or annual fee for a set number of seats. Instead, customers pay for the compute and storage they actually consume, using Snowflake Credits as the unit of measurement. This consumption-based model means Snowflake’s revenue grows or shrinks with customer usage: more queries, more data transformation pipelines, more AI workloads run on Snowflake = more revenue; cost-optimisation or workload migration to competing platforms = immediate revenue reduction. The consumption model creates more volatile revenue but also means Snowflake has nearly unlimited upside with its largest customers as they grow their data workloads.

Snowflake runs exclusively on top of the three major public clouds — Amazon Web Services (AWS), Microsoft Azure, and Google Cloud Platform (GCP) — making it simultaneously a partner to and tenant of the very hyperscalers whose own data warehouse products (Redshift, Synapse, BigQuery) compete with it.

Key Takeaways

• Snowflake generated $3.65B in FY2025 revenue (+27.6%), split 94%/6% between product revenue (consumption-based) and professional services; product gross margin of 73.5% and non-GAAP operating margin of ~7% reflect a path to profitability that GAAP figures (-23.8% operating margin, -$0.83B net loss) obscure due to ~$1.5B in annual stock-based compensation (~41% of revenue)
• Consumption-based pricing is the defining revenue mechanic: Snowflake earns Snowflake Credits (billed in US dollars) based on virtual warehouse compute usage (per second) and managed storage (per terabyte/month); there are no per-seat fees, no minimum usage floors for most customers, and no cap on upside — a customer that triples its data processing workloads pays roughly triple the Snowflake bill; this contrasts with subscription SaaS where revenue is fixed regardless of usage
• Net Revenue Retention Rate (NRR) of 127% is the proof that existing customers are expanding usage: for every $100 Snowflake earned from a customer cohort in FY2024, that same cohort spent $127 in FY2025; NRR above 100% means Snowflake grows revenue even with zero new customer acquisition; at 127%, it is among the highest NRRs in enterprise software — though it has been decelerating from 158% in FY2023, which is the central concern for investors
• Remaining Performance Obligations (RPO) of $6.9B — committed future spend from signed contracts — represents approximately 1.9x annual revenue and provides strong near-term visibility; Snowflake’s large enterprise customers sign multi-year capacity contracts (often 3-year), committing to minimum spend levels; RPO growing faster than revenue indicates accelerating customer commitment; RPO growth rate is a leading indicator of future consumption trends
• Databricks is the primary existential competitive threat: Databricks (private, ~$62B valuation) offers a competing data lakehouse platform with a unified architecture for data engineering, analytics, and AI/ML — Snowflake’s traditional strength was SQL analytics and structured data warehousing; Databricks was stronger in data engineering and machine learning; the two are converging on each other’s territory; every enterprise data platform RFP is now a Snowflake vs. Databricks decision, with BigQuery, Redshift, and Microsoft Fabric as secondary alternatives
• Cortex AI is Snowflake’s primary growth catalyst: Snowflake Cortex embeds AI capabilities (LLM inference, vector search, text-to-SQL, document AI) directly into the Snowflake platform — meaning enterprises can run AI on their existing Snowflake data without extracting data to a separate AI infrastructure; Cortex consumption is incremental to traditional SQL query consumption; if enterprises adopt Cortex for AI workloads at scale, it would re-accelerate Snowflake’s consumption growth rate without requiring new customers
• The hyperscaler paradox: Snowflake runs on AWS, Azure, and GCP — paying each approximately $0.30–0.40 of every dollar of infrastructure costs to the clouds it competes with in data warehousing; AWS (Redshift, Athena), Azure (Synapse, Fabric), and GCP (BigQuery) all have competing data products; yet Snowflake’s cross-cloud portability (run workloads on any cloud, join data across clouds) is its primary differentiation — something none of the hyperscalers can offer since each is incentivised to lock customers into their own cloud
• Stock-based compensation at ~41% of revenue (~$1.5B) is the primary obstacle between Snowflake’s genuine operating progress and GAAP profitability; at this SBC level, Snowflake is essentially transferring value from shareholders to employees at a rate that inflates reported losses; the key question is whether SBC as a percentage of revenue declines toward industry norms (15–25%) as the company matures — every 5 percentage points of SBC ratio decline on a $4B+ revenue base represents ~$200M+ in annual GAAP loss reduction

Snowflake (SNOW) Business Model

Snowflake is a consumption-based data platform — an enterprise infrastructure software company whose revenue is structurally linked to customer data workload volume rather than headcount. The business model combines two economic characteristics that rarely coexist: the high gross margins of software (~73.5%) with the revenue variability of a utility (customers can immediately reduce spend by running fewer queries or optimising their usage).

How the Consumption Pricing Engine Works

Snowflake Credits are the fundamental billing unit. One Snowflake Credit represents a standardised unit of compute resource. The price per credit varies by:

• Cloud provider: Credits cost slightly differently on AWS vs. Azure vs. GCP, reflecting Snowflake’s underlying infrastructure costs with each
• Cloud region: US regions vs. European vs. Asia-Pacific regions have different pricing tiers
• Edition: Snowflake Standard, Enterprise, and Business Critical editions carry different per-credit prices (Business Critical is the highest, adding security features like HIPAA compliance and private connectivity)

Virtual Warehouses (compute): A virtual warehouse is a cluster of compute nodes that executes SQL queries, data loading, and transformation workloads. Customers can create multiple virtual warehouses of different sizes (XS to 6XL, with credit consumption scaling roughly linearly with size) and run them simultaneously on the same data. A size Medium warehouse consumes 4 credits/hour; a size X-Large consumes 16 credits/hour. Virtual warehouses auto-suspend when idle (saving credits) and auto-resume when a query arrives — this dynamic scaling is a primary source of Snowflake’s reputation for cost efficiency vs. always-on traditional data warehouses.

Managed Storage: Snowflake charges per terabyte per month for data stored in its managed storage layer (built on each cloud’s object storage — S3 on AWS, Blob Storage on Azure, GCS on GCP). Storage rates are typically $23–40/TB/month depending on region and cloud. Storage is the lowest-margin component of product revenue (Snowflake passes through most of the underlying cloud storage cost).

The economic incentive alignment problem: Because Snowflake earns more when customers use more, Snowflake is structurally incentivised to help customers run more workloads on the platform rather than optimise costs. This is the inverse of a SaaS model where the vendor earns the same regardless of usage. However, customers who believe Snowflake is too expensive will optimise aggressively (reducing Snowflake’s revenue) or migrate workloads to Databricks or a hyperscaler’s native warehouse (permanently reducing revenue). Snowflake’s pricing team therefore must balance maximising per-credit revenue against the risk of triggering customer cost-optimisation behaviour.

The Consumption Cycle: From Contract to Revenue

Snowflake’s largest customers sign capacity contracts — multi-year commitments to purchase a minimum dollar amount of Snowflake Credits annually. Capacity contracts:

• Are recognised as Remaining Performance Obligations (RPO) when signed
• Generate revenue as credits are consumed (not when the contract is signed)
• Carry “on-demand” overage pricing for usage above the committed capacity — typically at a higher per-credit rate than the contracted rate

The capacity commitment model creates the $6.9B RPO figure: this is the sum of all committed future spend from signed capacity contracts. RPO is a leading indicator because: (a) it grows when large enterprise customers sign or expand contracts; (b) it converts to revenue only as consumption occurs; (c) RPO growing faster than revenue means future revenue is building faster than current consumption, which is a positive signal.

Revenue is earned as credits are consumed: If a customer signs a 3-year, $30M capacity contract in Q1, Snowflake does not recognise $30M immediately — it recognises revenue quarter by quarter as the customer consumes credits. This creates the characteristic “deferred revenue” balance on Snowflake’s balance sheet.

New Revenue Lines: Snowflake’s Consumption Expansion Strategy

Beyond traditional SQL warehousing, Snowflake is building new product lines that generate incremental credit consumption:

Snowflake Cortex (AI/ML): Cortex is Snowflake’s embedded AI platform — offering LLM inference, semantic search (vector search over embeddings), text-to-SQL (natural language queries against structured data), document AI (extracting structured data from PDFs/images), and ML model training functions, all running natively inside Snowflake without data leaving the platform. Cortex uses Snowflake Credits for AI inference calls — generating incremental consumption revenue on top of traditional SQL query consumption.

The strategic logic: enterprises have their production data in Snowflake. If they can run AI on that data without extracting it to a separate AI infrastructure (a GPU cluster, a dedicated vector database, an external LLM API), they (a) save integration costs, (b) avoid data movement latency and compliance risk, and (c) simplify their data architecture. Cortex’s value proposition is “bring AI to your data” rather than “move your data to an AI system.” Key Cortex products: Cortex Analyst (LLM-powered natural language to SQL), Cortex Search (vector search over Snowflake data), Cortex Fine-tuning (fine-tune open-source LLMs on Snowflake data).

Cortex is at an early stage — management has not disclosed Cortex-specific revenue — but AI consumption was cited as a meaningful incremental contributor to product revenue in FY2025.

Snowpark: Snowpark enables Python, Java, and Scala code to execute directly inside Snowflake’s compute environment, eliminating the need to extract data to external notebooks or compute infrastructure for data science workloads. A data scientist who previously exported Snowflake data to a Jupyter notebook running on their laptop or on Databricks can now run the same code inside Snowflake — generating credit consumption. Snowpark is the direct response to Databricks’ strength in Python-based data engineering.

Apache Iceberg Tables (open table format): Iceberg is an open-source table format that allows data to be stored in open file formats (Parquet) rather than Snowflake’s proprietary internal format. Iceberg Tables reduce data lock-in — customers can access the same data from Databricks, AWS Athena, or other engines — but Snowflake bets that it can win usage on performance, ease of use, and governance features even when data is portable. Iceberg Tables are strategically defensive: by supporting open formats, Snowflake removes “vendor lock-in” as a Databricks sales argument, competing on merit rather than migration barriers.

Snowflake Marketplace: A data marketplace where third-party data providers (financial data, weather data, geolocation data, healthcare claims data, ESG datasets) list their data products for Snowflake customers to purchase and access. Revenue model: Snowflake takes a commission on marketplace transactions and earns compute credits when customers query marketplace data. Currently a small revenue contributor but strategically significant — a rich Marketplace increases the value of being on Snowflake for data-hungry enterprises.

The Hyperscaler Relationship: Partner, Tenant, and Competitor

Snowflake’s relationship with AWS, Azure, and GCP is among the most complex in enterprise software:

Partner: All three clouds list Snowflake on their marketplaces, making it easy for cloud customers to purchase Snowflake on their existing cloud billing relationships and count Snowflake spend toward cloud commitment drawdowns (important for enterprises with large cloud capacity commitments).

Tenant: Snowflake pays each cloud for the compute and storage it uses to run the platform — approximately 36% of product revenue is cost of revenue, primarily cloud infrastructure costs. As Snowflake scales, it negotiates better unit economics with each cloud — this is the primary driver of product gross margin improvement over time.

Competitor: AWS Redshift/Athena, Azure Synapse/Fabric, and Google BigQuery all compete with Snowflake for analytics workloads. A customer who migrates analytics from Snowflake to BigQuery loses revenue for Snowflake and gains revenue for Google Cloud.

Snowflake’s cross-cloud advantage: No hyperscaler can credibly offer cross-cloud data sharing or analytics — AWS won’t connect natively to Azure, and GCP won’t share data with AWS workloads. Snowflake’s ability to join data across clouds (a financial firm running some workloads on AWS, others on Azure, can query across both in Snowflake) is a structural competitive moat that hyperscalers cannot replicate without undermining their own lock-in strategies.

Snowflake Competitors

Databricks — the primary competitive threat

Databricks (private, ~$62B valuation as of its last funding round) is Snowflake’s most consequential competitor and the one that generates the most anxiety among Snowflake investors. Databricks’ “Data Lakehouse” architecture combines data warehousing (SQL analytics), data engineering (ETL pipelines), and machine learning (MLflow, Mosaic AI) on a single unified platform using open formats (Delta Lake, Apache Iceberg). Where Snowflake historically led: SQL analytics, BI/reporting workloads, structured data warehousing, data sharing. Where Databricks historically led: Python-based data engineering, ML model training, unstructured data processing, streaming. The two companies are converging: Snowflake added Snowpark (Python in Snowflake) and Cortex AI; Databricks added SQL Warehouse and BI capabilities. Every enterprise evaluating a modern data platform now pits Snowflake against Databricks as the primary choice. Databricks’ private status means it reports limited financial data, but management commentary suggests $2.4B+ in FY2024 revenue growing rapidly.

Oracle — the legacy data warehouse and cloud incumbent

Oracle competes through Autonomous Data Warehouse (ADW) and Oracle Cloud Infrastructure (OCI) as alternatives to Snowflake for enterprises with existing Oracle database investments. Oracle’s advantage: deep existing enterprise relationships, on-premise deployment options for regulated industries that cannot fully move to public cloud, and aggressive pricing for Oracle-to-Oracle migrations. Oracle is not Snowflake’s most common head-to-head competitor in greenfield deals but is a factor in any enterprise with Oracle’s significant installed base.

Alphabet (Google BigQuery) — the hyperscaler data warehouse

Google BigQuery is a serverless, fully managed data warehouse with consumption-based pricing — the closest architectural analogue to Snowflake among the hyperscalers. BigQuery’s advantage: deeply integrated with Google Cloud ecosystem, competitive pricing for pure analytics workloads, and Vertex AI for ML. BigQuery’s disadvantage vs. Snowflake: it only runs on GCP, limiting cross-cloud flexibility; data sharing across organisations is more complex than Snowflake’s native data sharing. For enterprises that are 100% on GCP, BigQuery is a strong alternative to Snowflake.

Palantir — the AI/ML analytics competitor for government and enterprise

Palantir competes in a different layer of the analytics stack — less on raw data warehousing and more on AI-powered analytical applications (AIP, Foundry, Gotham) layered over data infrastructure. Palantir and Snowflake can coexist in the same enterprise architecture (Snowflake as the data warehouse, Palantir as the operational AI layer on top), but enterprises building AI applications may choose Palantir’s integrated approach over building their own stack on Snowflake + a separate AI framework. The competitive overlap is greatest in government/intelligence analytics and in enterprises seeking pre-built AI operational workflows rather than DIY data platform construction.

Datadog — a comparable cloud SaaS growth benchmark

Datadog is not a direct competitor to Snowflake but is the closest financial comparison in the enterprise cloud software space: also consumption-based, also high NRR (120%+), also growing 20%+ at scale, also investing heavily in AI features (Bits AI, LLM Observability) layered on top of core monitoring. Comparing Datadog’s margin trajectory (reached consistent non-GAAP profitability earlier than Snowflake, lower SBC ratio) against Snowflake’s is the relevant benchmark for Snowflake’s profitability path. Both companies demonstrate the operating leverage inherent in high-gross-margin cloud software platforms as they scale.

Revenue Breakdown

Financial data sourced from Snowflake SEC Filings.

Product revenue growth of 28.6% is driven primarily by existing customer expansion (NRR 127%) rather than new customer acquisition — Snowflake added approximately 700 net new customers in FY2025, but the majority of revenue growth came from existing customers running more workloads. The largest customers dominate the revenue mix: customers spending over $1M annually (the “Global 2000” enterprise tier) grew approximately 30%+ and represent the majority of product revenue. Professional Services growth (15.9%) is intentionally slower — Snowflake de-emphasises its own professional services in favour of partner-led implementation, keeping services margins low and services revenue as a small percentage of total.

Revenue Trend (3-Year)

The 3-year trend shows consistent revenue growth with a decelerating rate (from 36.4% to 27.8%) as the law of large numbers applies to a $3.6B base — and consistent gross margin expansion (65.9% → 70.7% → 73.5%) as Snowflake negotiates better infrastructure pricing with the clouds at scale. Net loss has been remarkably stable at approximately -$830M for three consecutive years despite revenue growing from ~$2B to $3.65B — demonstrating that while the business is not GAAP profitable, the losses are not expanding; operating leverage is emerging in the non-SBC cost base. If SBC normalises from 41% toward 25% of revenue over 3–4 years (as the employee equity compensation pool matures from the IPO era), Snowflake’s GAAP profitability could improve rapidly.

Snowflake (SNOW) Income Statement

Financial data sourced from Snowflake SEC Filings.

Note: Blended gross margin (63.8%) is lower than product gross margin (73.5%) because professional services carries approximately 25–35% gross margin, diluting the blended rate. The non-GAAP operating margin improvement (4.4% → 7%) is the most honest representation of Snowflake’s underlying operating leverage — it excludes SBC but shows genuine margin expansion as revenue grows faster than non-SBC costs.

Snowflake (SNOW) Key Financial Metrics

• Product Gross Margin: 73.5% — Has expanded from ~65.9% (FY2023) to 73.5% (FY2025) — a 760 basis point improvement over two years as Snowflake’s scale enables better unit economics with AWS/Azure/GCP. The long-term target is approximately 75–78%, consistent with mature cloud platforms. Each point of product gross margin improvement on $3.4B in product revenue represents approximately $34M in additional gross profit, making the cloud cost negotiation a material value driver

• Net Revenue Retention Rate: 127% — The most important leading indicator of Snowflake’s revenue health. At 127%, existing customers are growing spending 27% annually through expanded data workloads. This NRR has been decelerating: from 158% (FY2023) → 131% (FY2024) → 127% (FY2025). The deceleration reflects both a tougher macro environment (enterprises have been cost-optimising cloud spending) and the law of large numbers (sustaining high NRR on a larger existing customer base is harder). Watch for NRR stabilising above 125% as a sign that the deceleration has plateaued

• Remaining Performance Obligations: $6.9B — Approximately 1.9x annual revenue, providing strong revenue visibility. RPO grew ~47% in FY2025, significantly faster than the 27.8% total revenue growth — indicating that customer commitment to Snowflake (through multi-year contracts) is accelerating even as near-term consumption growth moderates. This divergence (RPO growth » revenue growth) is a positive signal that consumption acceleration may follow commitment acceleration in future quarters

• Free Cash Flow: ~$1.0B (FY2025) — Snowflake is FCF positive despite GAAP losses, because SBC is a non-cash expense and working capital timing adds cash from customer capacity contract prepayments. ~$1B FCF on $3.65B revenue represents a 27% FCF margin — genuinely strong. FCF is the most accurate indicator of Snowflake’s cash-generating ability

• Stock-Based Compensation: ~$1.5B (~41% of revenue) — The primary disconnect between GAAP and economic reality. SBC at 41% of revenue is among the highest ratios in enterprise software. This is a consequence of Snowflake’s IPO-era equity grants to senior engineers and executives maturing through the P&L. At current revenue growth, SBC as a percentage of revenue should naturally decline toward 25–30% by FY2027 even without reducing absolute SBC spend — but shareholder dilution is real and ongoing

• Customer Count and $1M+ Customers: Snowflake had approximately 10,600+ total customers (FY2025) and approximately 580+ customers spending $1M+ annually. The $1M+ customer cohort drives the majority of product revenue and NRR — a $1M customer growing at 127% NRR adds $270K in incremental annual revenue; tracking the growth in this cohort is more strategically relevant than total customer count

Is Snowflake Profitable?

Not on a GAAP basis — Snowflake reported a net loss of $833 million in FY2025. However, the GAAP loss is almost entirely explained by ~$1.5 billion in stock-based compensation — a non-cash expense that transfers economic value to employees without consuming cash. On a non-GAAP basis (excluding SBC), Snowflake achieved approximately 7% operating margins and generated approximately $1 billion in free cash flow.

The profitability question is therefore: when does Snowflake reach GAAP profitability, and does it matter? On the first question — SBC as a percentage of revenue should mechanically decline as the IPO-era grant pool amortises and revenue scales. If SBC holds at ~$1.5B while revenue grows to $5B+, SBC falls to 30% of revenue; at $7B revenue, to ~21%. Combined with continued non-GAAP margin expansion, GAAP profitability at modest levels is a 3–4 year trajectory. On whether it matters: institutional investors who value Snowflake on FCF or non-GAAP earnings already see a profitable business; retail investors and analysts using trailing P/E see deep losses. The SBC dilution is real dilution regardless of accounting treatment.

Snowflake (SNOW): What to Watch

1. NRR stabilisation above 125% — Net Revenue Retention Rate decelerating from 158% to 127% over two years is Snowflake’s most-discussed risk. The bear case: NRR continues declining toward 115–120% as customers optimise cloud spend and Databricks wins incremental workloads. The bull case: AI workloads (Cortex, vector search, ML features) re-accelerate consumption growth as enterprises move from evaluating AI to deploying it at scale. Watch each quarterly NRR disclosure — stabilisation at 125%+ would significantly reduce investor concern about structural share loss; any quarter showing NRR below 120% would validate the bear case

2. Cortex AI consumption ramp — Management has identified Cortex as the primary incremental consumption growth driver. Watch for any management quantification of Cortex-related revenue or consumption (currently not broken out), as well as commentary on Cortex customer adoption rates. Key milestones: Cortex Analyst (text-to-SQL) reaching meaningful adoption in the Global 2000, Cortex Search (vector search) displacing external vector databases for enterprises running RAG (retrieval-augmented generation) applications on Snowflake data. If Cortex becomes a meaningful consumption driver (5%+ of product revenue) by FY2026, the re-acceleration thesis becomes credible

3. RPO growth as a leading indicator — Remaining Performance Obligations growing 47% while revenue grows 28% creates a gap that, if sustained, suggests future revenue acceleration. Watch quarterly RPO growth — if RPO growth rate decelerates toward revenue growth rate (both at ~25-28%), it means new contract signings are in line with current consumption; if RPO continues outpacing revenue growth (35%+ RPO growth vs. 25-28% revenue growth), it suggests near-term consumption acceleration as contracts convert to usage. This is the most forward-looking indicator available in Snowflake’s reported metrics

4. Databricks competitive win/loss in enterprise RFPs — Snowflake does not disclose competitive win rates, but channel checks (from third-party surveys, earnings call commentary, and partner disclosures) provide directional data on Snowflake vs. Databricks competitive positioning. Watch for: (a) any major enterprise announced departures from Snowflake to Databricks (or vice versa); (b) management commentary on competitive environment on earnings calls — any shift in language from “competitive dynamic unchanged” to “increased competition” is significant; (c) Databricks IPO preparations, which will require public financial disclosures and provide the first direct revenue comparison between the two companies

5. SBC ratio decline toward 25% — Stock-based compensation at 41% of revenue is the largest single obstacle to Snowflake’s GAAP profitability. Watch the SBC-to-revenue ratio each quarter — declining below 35% would be a meaningful milestone, below 30% would be a material positive for GAAP loss trajectory. The driver is simply revenue scaling faster than SBC dollar growth; at $5B revenue with flat SBC of $1.5B, the ratio drops to 30%. Any absolute SBC reduction (which would require reducing equity compensation to employees — unlikely until growth decelerates further) would accelerate GAAP profitability significantly

6. Product gross margin progression toward 76–78% — Product gross margin has expanded 760 basis points over two years (65.9% → 73.5%) through cloud cost negotiation and engineering efficiency. Watch for continued expansion: management has guided toward ~75% near-term product gross margin. Each 100 basis points of product gross margin expansion on $3.5B+ product revenue represents approximately $35M in additional gross profit falling through to operating income. If product gross margin stalls at 73–74% (hitting the ceiling of what cloud cost negotiation can deliver), it suggests Snowflake’s margins are approaching their structural ceiling absent fundamental infrastructure changes

7. Iceberg Tables adoption and strategic open-format bet — Apache Iceberg Table adoption is Snowflake’s bet that winning on merit (performance, governance, ease of use) in an open-data-format world is better than winning through proprietary lock-in. Watch for management commentary on Iceberg Table adoption rates and any associated consumption economics (do Iceberg Tables carry the same or lower gross margins than proprietary format tables?). If Iceberg adoption accelerates without reducing consumption economics, it validates Snowflake’s “openness as a moat” thesis and removes one of Databricks’ most effective competitive arguments

Snowflake (SNOW) Financial Summary

Snowflake (SNOW) generated $3.65 billion in total revenue in fiscal year 2025 (+27.8% YoY) with a product gross margin of 73.5%, ~$1B in free cash flow, and a non-GAAP operating margin of ~7% — a genuinely strong underlying business obscured by ~$1.5B in annual stock-based compensation that drives a GAAP net loss of -$833M. The NRR of 127% and $6.9B RPO confirm Snowflake’s enterprise data platform position; the question is whether AI workloads (Cortex) and the expanding platform (Snowpark, Iceberg, Marketplace) can re-accelerate consumption growth toward 30%+ before Databricks captures a structurally larger share of enterprise data platform spending. For the consumption-based cloud SaaS growth comparison, see How Datadog Makes its Money. For the data analytics and AI platform contrast, see How Palantir Makes its Money.