Sridhar Ramaswamy’s first full phase as Snowflake chief executive has produced the acceleration investors wanted and the strategic expansion customers expected. In the first quarter of fiscal 2027, product revenue reached $1.33 billion, up 34 per cent year on year, while total revenue rose 33 per cent to $1.39 billion. Net revenue retention was 126 per cent, and the company counted 779 customers generating more than $1 million in trailing product revenue. Snowflake raised full-year product revenue guidance to $5.84 billion and lifted its non-GAAP operating margin outlook to 13.5 per cent. These are strong numbers for a consumption platform already operating at substantial scale.
The question facing Ramaswamy is no longer whether Snowflake can attach itself to the enterprise AI cycle. It is whether the company can become the control plane through which agents securely find data, build workflows and take action, while preserving the openness that large organisations increasingly demand. Snowflake’s Summit 2026 announcements pushed into coding agents, knowledge-worker agents, governance, model access, open table formats and application connectivity. The breadth is deliberate. Ramaswamy is trying to ensure that AI activity creates more Snowflake consumption rather than moving value to model providers or separate application layers.
A product leader at a commercial inflection
Ramaswamy’s route to the top helps explain the strategy. He built his reputation in large-scale computing and machine learning, led Google’s advertising organisation, founded the search company Neeva and joined Snowflake when it acquired that business. Before becoming chief executive in 2024, he led Snowflake’s AI work. He therefore arrived with both technical conviction and experience of a market in which superior engineering does not automatically produce a durable business. Distribution, economics and user habit matter as much as model quality.
Snowflake’s consumption model makes that lesson unusually immediate. Customers commit capacity, but revenue is recognised when they use compute, storage and data-transfer resources. Product improvements that reduce the cost of a query can slow revenue in the short term unless lower prices stimulate more workloads. AI features can create demand, but only if they move from demonstrations to regular production use. Ramaswamy must manage an optimisation paradox: make the platform more efficient for customers while expanding the number, complexity and frequency of tasks they run.
The fiscal 2027 trajectory suggests the balance is improving. First-quarter product growth of 34 per cent exceeded the previous year’s pace, and management forecast 31 per cent growth for the full year. The company also generated $265.5 million of adjusted free cash flow in the quarter. Yet Snowflake continues to report a substantial GAAP operating loss, reflecting stock-based compensation and other costs. Ramaswamy must show that faster product growth can coexist with a cleaner operating model, particularly as acquisitions and new product teams widen the portfolio.
Agents need governed context
Snowflake’s central argument is that enterprise AI is constrained less by access to models than by access to trusted context. A general-purpose model can reason, but it cannot safely act on a company’s pricing, customer, supply-chain or clinical data without permissions, metadata and monitoring. Horizon Catalog is meant to provide that governance layer across data inside Snowflake, external lakes and compatible engines. Snowflake Intelligence and CoWork are aimed at knowledge workers, while CoCo gives builders a coding agent for data workflows and applications.
If these products work together, Snowflake can sit between models and business systems. An employee could ask for an analysis, an agent could retrieve governed information, and a workflow could produce or initiate an action without exposing raw data outside policy boundaries. That is a more ambitious role than supplying analytics infrastructure. It also creates more responsibility. Access controls have to remain consistent when agents chain tools, outputs must be traceable, and administrators need to understand which model or service touched which data.
Ramaswamy’s emphasis on openness is therefore strategic rather than cosmetic. Support for Apache Iceberg, Polaris and external engines helps Snowflake answer customer concern about lock-in. Enterprises want a coherent governance layer, but they do not want to surrender the ability to use different compute engines, clouds and AI models. Snowflake is trying to make openness a reason to centralise control on its platform. The challenge is to maintain meaningful interoperability without allowing rival services to capture the highest-value workloads while Snowflake bears the cost of storage and governance.
The partnership economy cuts both ways
Snowflake’s relationships with Anthropic, Amazon Web Services and other technology providers expand its reach. The five-year commitment to spend $6 billion on AWS compute and AI services signals confidence in workload growth and deepens joint go-to-market activity. Anthropic models within Cortex AI give customers direct access to advanced capabilities under Snowflake governance. Similar connections across the ecosystem allow buyers to choose models according to use case rather than rebuild data pipelines for each provider.
These alliances also place Ramaswamy inside a difficult margin equation. Cloud infrastructure is a major input cost, model inference can be expensive, and customers expect continuing price-performance improvements. Snowflake must package agentic services so that customer value rises faster than its own compute bill. The company can defend economics through optimisation, volume commitments, premium governance and a growing set of native services. It cannot assume that rising AI usage will automatically translate into attractive margins.
Acquisitions add another layer. Observe extends Snowflake into observability, while the proposed acquisition of Natoma targets secure connectivity for agents. Each deal can increase the platform’s usefulness, but every addition competes for engineering attention and risks confusing the product story. Ramaswamy must integrate capabilities into shared governance, billing and developer experiences. A collection of acquired tools will not create a control plane. Customers must be able to move from data to agent to operation with fewer seams than they encounter across specialist products.
Enterprise adoption is the real benchmark
Snowflake’s large-customer growth provides an encouraging base. Organisations spending more than $1 million a year increased, and sectors such as pharmaceuticals, financial services, retail and media are moving AI workloads onto governed data. Ramaswamy needs those customers to extend beyond analytics into development and operational workflows. The decisive metric will not be how many users test an assistant. It will be how often agents run, how much governed data they access and whether the resulting work changes revenue, cost or service outcomes.
That shift will require a different sales conversation. Data teams understand warehouses and lakehouses; business leaders care about cycle time, risk and return on investment. Snowflake’s sellers and partners need to connect platform capabilities to a profit-and-loss line while avoiding inflated promises. Ramaswamy has argued that AI projects should link to measurable business value. His organisation must make that standard operational, helping customers prioritise workflows with sufficient data quality and repeatability to justify production deployment.
Asia offers both scale and friction. Banks, manufacturers, telecommunications groups and digital platforms hold rich data but face sovereignty rules, fragmented cloud estates and uneven AI readiness. Snowflake has expanded operations and partnerships across the region, yet it must localise governance, support and industry solutions. Ramaswamy’s Indian heritage adds visibility, but execution depends on regional engineering, channel depth and trust with regulators and enterprise buyers. A global product will not win by architecture alone.
Snowflake’s own use of its products will be another useful test. A company selling agents to large organisations should be able to show how its finance, engineering, support and sales teams redesign work under the same governance standards offered to customers. Internal deployment can expose permission gaps and weak hand-offs before they reach the market. It can also give sellers credible operating evidence. Ramaswamy should publish outcome measures carefully, separating genuine cycle-time or quality gains from activity that merely shifts effort between teams.
Speed must not outrun coherence
The volume of Summit 2026 announcements demonstrated product velocity. It also exposed the leadership risk of this phase. Snowflake is simultaneously a warehouse, a data platform, an application environment, a governance system, an AI service and an agentic interface. Customers may welcome integration, but they need a clear path through the portfolio. Ramaswamy must impose a strong product hierarchy, simplify commercial packaging and ensure that every new name maps to a recognisable customer problem.
He also has to protect trust. Agents magnify the consequences of poor permissions and incorrect context because they can act, not only answer. Snowflake’s ability to log access, enforce policy and isolate data is a potential advantage. One high-profile breach or uncontrolled action could slow adoption across the category. Security, evaluation and observability must therefore advance at the same pace as model and interface features. In the agentic enterprise, governance is part of the product experience rather than a compliance layer added later.
Ramaswamy has moved Snowflake from an AI narrative to an AI operating agenda, and the financial momentum gives him permission to invest. The next proof will come from consumption quality: more production workloads, stronger retention, expanding margins and customer outcomes that persist after the novelty of agents fades. If Snowflake becomes the place where governed data turns into repeatable action, Ramaswamy will have enlarged both the company’s market and its strategic authority. If usage remains concentrated in analytics while agent value accrues elsewhere, the platform will have expanded faster than its economics.
