Snowflake Earnings Beat Expectations, but AI Consumption Must Prove It Can Last

Snowflake earnings delivered a clear upside surprise, with quarterly revenue reaching $1.55 billion and product revenue growing 37% from a year earlier. The result exceeded the roughly $1.48 billion revenue consensus reported by LSEG. Snowflake also raised its annual product revenue forecast well above its previous guidance.

The headline beat matters because Snowflake operates a consumption-based business. Customers pay as they use computing, storage, and AI services, rather than through predictable seat-based subscriptions. Faster growth therefore suggests that enterprise workloads are expanding, not merely that customers signed longer contracts.

The deeper question is whether AI has changed Snowflake’s growth trajectory or simply produced a strong quarter. Databricks, Microsoft, Google, Amazon Web Services, and other data-platform providers are competing for the same AI workloads. Snowflake now needs to show that its acceleration reflects durable production use, not experimentation that customers later optimize away.

The Snowflake Earnings Beat Was Broader Than One Revenue Number

Snowflake exceeded expectations across current revenue, future commitments, and its full-year outlook.

Snowflake reported its fiscal 2027 second-quarter results on September 2, 2026. The quarter ended July 31, so the announcement covers activity completed before the final month of the calendar third quarter.

According to the company’s quarterly results, total revenue reached $1.55 billion. That represented 35% year-over-year growth. Product revenue reached $1.49 billion, rising 37%.

Those numbers must remain distinct. Total revenue includes product revenue and professional services. The product figure is especially important because Snowflake uses it to measure demand for its core platform.

External consensus data placed expected total revenue near $1.48 billion. Some reports rounded that estimate differently, but the essential result remains unchanged. Snowflake exceeded Wall Street’s average forecast by roughly $70 million.

Management described the period as its third consecutive quarter of accelerating product revenue growth. That pattern carries more weight than a single earnings beat. Snowflake exited fiscal 2026 with product revenue growth near 30%, then accelerated during the first half of fiscal 2027.

The customer indicators also moved in the same direction. Snowflake ended the quarter with 828 customers generating more than $1 million in trailing 12-month product revenue. That group grew 27% from the previous year.

Large-customer growth matters because enterprise data projects often begin with a narrow workload. They become strategically important when additional departments, datasets, applications, and AI systems move onto the same platform.

Snowflake also reported 829 Forbes Global 2000 customers. That number provides a separate measure of enterprise penetration, although it does not reveal how much each customer spends.

Net revenue retention was 126%. Net revenue retention compares current product revenue from an existing customer group with revenue from the same group one year earlier. A rate above 100% means expansion outweighed reductions and customer losses.

The 126% figure indicates that existing customers collectively increased their spending. It does not show whether that expansion was evenly distributed, however. A small number of rapidly growing AI workloads can materially influence consumption revenue.

Remaining performance obligations reached $9 billion, up 30% year over year. This metric represents contracted future revenue that Snowflake has not yet recognized. It offers visibility into demand, though recognition timing depends on customer usage and contract terms.

The company remained unprofitable under generally accepted accounting principles. It reported a quarterly net loss of approximately $192 million, improving from about $298 million one year earlier. That gap between adjusted performance and GAAP profitability remains part of the investment debate.

Taken together, these figures show a broad improvement. Customers committed more money, existing accounts expanded consumption, large-account numbers increased, and recognized revenue beat expectations.

That breadth explains why investors focused on more than the $1.55 billion headline. The quarter presented evidence that Snowflake’s growth engine accelerated across several connected indicators.

Snowflake Raised Its Forecast Far Beyond the Prior Baseline

The more consequential change was Snowflake’s decision to raise full-year product revenue guidance to $6.07 billion.

Snowflake previously expected fiscal 2027 product revenue of approximately $5.84 billion. Its new forecast adds about $230 million to that outlook and implies 36% annual growth.

The raised figure also exceeded the approximately $5.86 billion analyst consensus cited in early reporting. This was not a narrow guidance adjustment designed merely to match Wall Street’s expectations.

Snowflake expects fiscal third-quarter product revenue of approximately $1.59 billion. That outlook was above an analyst estimate near $1.51 billion in the original news report and implies that management expects momentum to continue beyond the reported quarter.

The market response reflected that forward-looking signal. Snowflake shares rose more than 20% in extended trading after the announcement.

Stock movements do not verify a company’s operating claims. They do show how investors interpreted the difference between the old forecast and the new one. In this case, the market treated the guidance increase as evidence that demand had moved beyond prior assumptions.

The scale of the revision matters because Snowflake had already raised expectations earlier in the fiscal year. Another increase suggests that the company observed stronger consumption after setting its previous forecast.

Consumption models can reveal demand changes faster than traditional subscription contracts. When customers run more queries, process larger datasets, or invoke more AI models, revenue can rise within the same quarter.

The same sensitivity creates risk. Customers can optimize workloads, reduce unnecessary queries, or delay projects. Snowflake experienced that pressure when enterprises scrutinized cloud spending in earlier periods.

This quarter indicates that expanding workloads outweighed those optimization efforts. It does not mean optimization has disappeared. Customers will continue looking for better performance per unit of computing, particularly as AI inference raises infrastructure bills.

Snowflake also guided toward a 14.5% non-GAAP operating margin for the full year. Its expected non-GAAP adjusted free cash flow margin was 23%.

These margin targets matter because AI workloads can be expensive to serve. Model inference, accelerated computing, networking, and storage create costs that differ from conventional analytical queries.

Management therefore faces two simultaneous tests. Snowflake must increase AI-related consumption while preserving acceptable gross and operating margins. Growth achieved through uneconomic computing would weaken the value of the revenue acceleration.

Snowflake’s long-term infrastructure commitments add another layer. The company signed a five-year agreement involving $6 billion of expected AWS infrastructure consumption during the previous quarter.

That commitment can secure capacity and improve economics. It also raises the importance of sustained platform demand because Snowflake must generate enough customer consumption to use the infrastructure efficiently.

The new forecast indicates that management currently sees enough demand to raise its expectations. The next several quarters will reveal whether that confidence survives changes in customer budgets and computing efficiency.

AI Is Turning From a Feature Story Into a Consumption Story

The central reversal is that AI appears to be increasing demand for Snowflake’s core data platform instead of bypassing it.

The bearish argument around enterprise AI was straightforward. New models might let companies interact directly with information across existing systems, reducing the need for a separate cloud data warehouse.

Snowflake is presenting the opposite outcome. Its platform sits between enterprise data and AI applications, providing storage, computing, governance, search, and access controls. AI applications can create additional queries and data-processing activity inside that environment.

CFO Brian Robins said the quarter included “a meaningful step-up in AI revenue.” He also connected the result with continued strength in Snowflake’s core data-platform business.

That distinction is important. AI is not operating as a completely separate product category. It can increase demand for existing warehouses, data pipelines, governance systems, and application infrastructure.

Snowflake Intelligence provides one example. The product lets business users ask questions across enterprise data through a conversational interface. Underneath that interface, Snowflake still needs to identify authorized data, retrieve relevant information, and run governed computations.

Cortex AI provides services for building and operating generative AI applications within Snowflake. Cortex AI SQL brings model-assisted operations into SQL, the common language used to query structured data.

A Snowflake-authored technical paper describes how Cortex AI SQL treats model inference cost as part of query planning. The system can route easier tasks toward less expensive models and reserve more capable models for uncertain cases.

That mechanism connects AI adoption directly with Snowflake’s economic challenge. Customers want useful model outputs, but they also need predictable latency and manageable computing costs.

If Snowflake reduces the cost of an AI query, customers might spend less on each request. Yet lower costs can also make additional production use cases practical, increasing total query volume.

This dynamic resembles earlier cloud optimization cycles. More efficient infrastructure does not automatically reduce a vendor’s revenue. Lower unit costs can stimulate greater consumption when customers find more workloads worth running.

Snowflake’s existing data position helps. Enterprise AI systems require more than access to a language model. They need current business information, permission controls, quality checks, lineage, and monitoring.

Those requirements become especially important when an AI agent can take actions. An incorrect summary creates inconvenience. An incorrect automated action can affect financial records, customers, inventory, or regulated processes.

Snowflake is trying to become the governed execution layer for those systems. Its opportunity is not limited to selling access to proprietary models. Customers can use different model providers while keeping data and control policies inside Snowflake.

That approach also explains why the platform’s core business accelerated alongside AI revenue. Before a customer deploys an AI agent, it may need to consolidate information, improve data quality, and establish access rules.

Those preparatory workloads generate conventional data-platform consumption. AI can therefore expand revenue before the final application reaches broad production use.

The company’s earnings discussion also emphasized migrations from older data systems. These migrations remain a separate growth source, but AI urgency can accelerate them.

An enterprise may tolerate fragmented data when employees manually assemble reports. The same fragmentation becomes a larger obstacle when an AI system needs reliable, immediate access across departments.

Snowflake’s strongest argument is therefore broader than “customers want AI.” It is that useful enterprise AI increases demand for organized, governed, and computationally accessible data.

Knowledge workers face a similar problem at a smaller scale. A model cannot reliably answer questions when relevant notes and documents remain scattered. A searchable personal knowledge base applies the same basic principle to individual work.

For Snowflake, the revenue opportunity comes from applying that principle across large organizations. Every production AI system can create new demand for data preparation, retrieval, security, and monitoring.

That is why this quarter is more meaningful than an isolated AI product announcement. Snowflake reported evidence that AI-related activity is now affecting recognized revenue and its annual forecast.

Databricks and the Cloud Giants Keep the Pressure High

Snowflake’s results strengthen its position, but they do not settle the contest for enterprise AI data workloads.

Databricks remains the clearest direct opponent. The company developed its reputation around data engineering and machine learning, while Snowflake initially became known for cloud data warehousing.

Both companies have expanded beyond those starting points. Snowflake added developer tools, machine learning services, open table support, and AI products. Databricks moved deeper into analytics, SQL workloads, governance, and business intelligence.

That convergence means buyers increasingly evaluate them against the same requirements. These include data engineering, analytics, model development, governance, application deployment, and support for open formats.

Snowflake’s revenue acceleration suggests that Databricks has not prevented it from expanding within large enterprises. The reverse conclusion would also be too strong. Many companies use both platforms and assign different workloads to each.

Microsoft, Google, and AWS create another form of pressure. Each controls a major cloud environment and offers integrated data, analytics, and AI services.

Microsoft can combine Azure infrastructure, Fabric, Power BI, and its relationships with OpenAI. Google offers BigQuery, Vertex AI, and its Gemini models. AWS connects its infrastructure footprint with Redshift, Bedrock, SageMaker, and related services.

These companies can package data and AI products with broader cloud commitments. They also control underlying infrastructure that Snowflake uses to deliver its service.

Snowflake’s response is multi-cloud availability and a consistent platform across providers. That can appeal to enterprises that want to reduce dependence on one cloud vendor or manage data across several environments.

However, multi-cloud architecture introduces its own costs. Moving information between regions or providers can create networking expenses, governance complexity, and latency.

Open data formats have also changed the competitive landscape. Apache Iceberg, an open table format, lets multiple computing engines work with the same underlying data files.

Snowflake supports Iceberg because customers increasingly want separation between data storage and a single vendor’s computing engine. Databricks promotes comparable openness through technologies associated with the lakehouse approach.

This shift gives customers more flexibility. It also weakens the traditional advantage of locking data and computing inside one proprietary environment.

Snowflake must therefore compete through performance, governance, developer experience, and integrated services. Simply storing a customer’s data is no longer enough.

AI increases that pressure because model providers and application developers want access to data without moving it unnecessarily. The platform that provides the easiest governed access can capture more downstream computing.

Snowflake’s second-quarter numbers show that customers are currently bringing more activity onto its platform. Remaining performance obligations and large-customer growth indicate that the company is also securing longer-term commitments.

Yet those commitments do not guarantee exclusive use. An enterprise can sign a large Snowflake contract while expanding Databricks, Microsoft Fabric, or Google BigQuery elsewhere.

Customer concentration deserves attention as well. Growth among accounts spending more than $1 million is encouraging, but large enterprises possess substantial bargaining power. They can negotiate discounts and shift workloads when economics change.

Snowflake’s consumption model makes competitive changes visible through usage. If a customer redirects machine learning training or AI inference to another environment, Snowflake may feel the effect before a subscription contract expires.

That sensitivity is partly why investors reacted strongly to the quarter. Accelerating consumption provides a relatively immediate signal that Snowflake is winning enough workloads to overcome competitive and optimization pressure.

The company’s previous annual filing identified competition, infrastructure dependence, security, and customer optimization as continuing risks. A strong quarter changes current evidence, not those structural conditions.

Snowflake has earned a stronger position in the debate. It has not earned exemption from it.

What the Numbers Still Do Not Prove

Snowflake has shown that AI contributed to faster growth, but it has not fully disclosed the durability or economics of that contribution.

Management said AI revenue increased meaningfully. The public earnings release did not provide a complete standalone AI revenue figure, customer cohort, or gross margin for those workloads.

That omission limits outside analysis. Investors can observe faster product revenue growth, but they cannot precisely separate AI consumption from migrations, conventional analytics, storage, pricing, or other platform services.

Customer adoption metrics also require context. An account can activate an AI feature without moving it into a critical production process. Trials, demonstrations, and limited departmental deployments can all count as usage.

Production workloads tend to behave differently. They process more data, require stricter security, and operate repeatedly. They also face closer scrutiny from finance, legal, and infrastructure teams.

The next question is therefore not whether customers can access Snowflake’s AI products. It is whether those products become recurring components of business operations.

Consumption volatility remains another uncertainty. AI development often creates an early surge as teams prepare data, test models, and compare architectures. Usage can decline when a project ends or reaches an optimization phase.

Snowflake’s raised third-quarter forecast reduces concern about an immediate reversal. It does not establish a multi-year pattern.

The relationship between efficiency and revenue also needs careful interpretation. Snowflake continually improves computing performance so customers can complete more work with fewer credits.

That improvement strengthens customer value and competitiveness. In the short term, however, it can reduce revenue from an unchanged workload. Snowflake needs new workloads and higher activity to offset those efficiency gains.

AI may supply that additional demand. The quarter supports that view, but future results must show that workload creation consistently exceeds optimization.

Margins provide a second test. Snowflake expects a 74% non-GAAP product gross margin for the full year, alongside the 14.5% non-GAAP operating margin.

Those figures indicate that management expects substantial profitability despite increased AI activity. Investors should still watch whether model inference and accelerated computing place pressure on product gross margin.

GAAP profitability remains unresolved. Snowflake’s roughly $192 million quarterly net loss improved, but it shows that adjusted earnings do not capture every material expense.

Stock-based compensation remains an important difference between GAAP and non-GAAP results. It does not require the same immediate cash payment as wages, but it can dilute existing shareholders when companies issue additional shares.

The market reaction creates its own challenge. A higher valuation places more weight on future execution. Snowflake now needs to meet an annual forecast that is substantially above its earlier plan.

Enterprise budgets add uncertainty. AI remains a management priority, but many organizations are still determining which use cases provide measurable returns.

Security and reliability can slow deployments. Companies may allow an AI assistant to summarize internal documents before permitting an agent to update databases or trigger operational workflows.

Data quality presents another limitation. An AI system can retrieve information quickly while still producing unreliable answers if the source data is inconsistent, incomplete, or poorly governed.

Snowflake can provide infrastructure for governance, but it cannot automatically correct every organizational process that created weak data. Successful deployments require technical work and operational discipline from customers.

These constraints do not negate the quarter. They define the standard Snowflake must meet next. The company needs to convert early AI interest into recurring production consumption without sacrificing margins or trust.

Three Signals Will Show Whether Snowflake’s Acceleration Can Continue

Third-quarter consumption, AI economics, and competitive workload wins will determine whether this quarter marks a durable change.

The first signal is Snowflake’s fiscal third-quarter product revenue. Management expects approximately $1.59 billion, comfortably above the prior analyst estimate cited in initial reports.

Meeting that target would show that the second-quarter acceleration continued after July. Another guidance increase would strengthen the argument that management still underestimated demand.

A miss would not automatically invalidate the AI thesis. Consumption can shift between quarters because of customer schedules and seasonal activity. A meaningful shortfall would still raise questions about the durability of recent workload growth.

Net revenue retention should be read alongside product revenue. A stable or rising rate would indicate that existing customers keep expanding. A decline would suggest that new customer activity is doing more of the work.

The second signal is the relationship between AI revenue and margins. Snowflake must demonstrate that increased model usage can coexist with its gross-margin and adjusted free-cash-flow objectives.

Investors should watch product gross margin, infrastructure costs, and management’s comments about inference efficiency. If margins hold while AI consumption rises, Snowflake’s platform economics look stronger.

If margins weaken, the market will need to determine whether the pressure reflects temporary investment or a structurally more expensive workload mix.

The third signal is evidence of competitive production wins. Customer announcements matter most when they describe deployed systems, repeated use, measurable consumption, and expansion across departments.

Generic partnership announcements offer less evidence. A company can announce access to a model or integration without generating material platform revenue.

Migration activity will also reveal competitive strength. When organizations move older warehouses or fragmented data systems into Snowflake, they create a foundation for future analytics and AI consumption.

The strongest cases will combine migration with production AI. That combination supports Snowflake’s argument that AI expands the core platform instead of replacing it.

Databricks, Microsoft, Google, and AWS will continue releasing competing services. Their responses will help show whether Snowflake has created a distinct governance and consumption advantage or simply joined a crowded feature race.

Enterprise buyers should compare platforms at the workload level. Data location, governance requirements, open-format support, model choice, performance, and total operating cost matter more than a single quarterly headline.

Developers should watch how easily Snowflake’s AI services move from prototypes into monitored applications. Deployment speed matters, but reliability and cost control determine whether those applications survive budget reviews.

Knowledge workers should expect more conversational access to enterprise data. They should also ask which sources support an answer, what permissions apply, and whether the system can distinguish verified information from generated text.

Snowflake’s second-quarter results provide the strongest evidence yet that AI is increasing its platform consumption. Revenue, retention, commitments, and guidance all moved in a favorable direction.

The remaining test is repetition. Watch the $1.59 billion third-quarter product target, AI-related margin commentary, and named production deployments. Those three signals will show whether Snowflake earnings marked a lasting acceleration or an unusually strong quarter.